VIEWS: 26 PAGES: 93 POSTED ON: 7/9/2011
DATE 3 December 2008 Dnr 08-9887/23 Hybrid Model Documentation v6.1 Table of contents 1 Introduction ...........................................................................................................1 1.1 Background .......................................................................................................1 1.2 Changes to model documentation.....................................................................1 1.3 Summary of results............................................................................................2 1.4 Structure of this document ................................................................................2 1.5 Overview of each model ....................................................................................2 1.5.1 Consolidation model ............................................................................................ 4 1.5.2 Core model .......................................................................................................... 4 1.5.3 Access model ...................................................................................................... 6 1.5.4 Co-location model................................................................................................ 7 1.6 Definitions and principles common to all models ..............................................7 2 Consolidation model ..........................................................................................12 2.1 Definitions and assumptions ...........................................................................12 2.2 Structure of consolidation model .....................................................................12 2.3 The main functions of the consolidation model ...............................................13 2.4 Annualisation assumptions..............................................................................14 2.4.1 Cost of Capital ................................................................................................... 14 2.4.2 Annualisation options......................................................................................... 15 2.4.3 Annualisation parameters .................................................................................. 17 2.5 Working capital ................................................................................................19 2.6 Functional area costs ......................................................................................19 2.6.1 Functional areas ................................................................................................ 19 2.6.2 Size of each area............................................................................................... 21 2.6.3 Cost of each area .............................................................................................. 21 2.6.4 Allocation of the cost of each area..................................................................... 22 2.6.5 Overhead costs.................................................................................................. 22 2.6.6 Expensed vs. annualised costs.......................................................................... 23 2.6.7 Number portability (IN/NP costs) ....................................................................... 23 2.6.8 Billing of transit traffic (kaskadavräkning) .......................................................... 24 2.7 Other common costs .......................................................................................24 2.7.1 Building costs..................................................................................................... 24 2.7.2 Shared facility costs........................................................................................... 25 2.7.3 Other shared facilities ........................................................................................ 26 2.8 Boundary between core and access ...............................................................26 2.9 Treatment of other services incl. retail ............................................................27 2.10 Allocation of costs to services .........................................................................27 2.10.1 Allocation for core services ................................................................................ 28 2.10.2 Allocation for access services............................................................................ 29 3 Core model ..........................................................................................................30 3.1 Definitions and assumptions ...........................................................................30 3.1.1 Increments ......................................................................................................... 30 3.1.2 Network structure – PSTN switching ................................................................. 30 3.1.3 Allocation Issues – PSTN Switching .................................................................. 33 3.1.4 Networks structure – Transmission.................................................................... 34 3.1.5 Network structure – Core Trenching .................................................................. 35 3.1.6 Network structure – IP Routing .......................................................................... 37 Post- och telestyrelsen i 3.1.7 Allocation issues – IP Routing ........................................................................... 38 3.1.8 Network structure – North and South................................................................. 38 3.1.9 Design timeframe (horizon) ............................................................................... 39 3.2 Structure of core model ...................................................................................40 3.3 Technical and volume inputs...........................................................................42 3.3.1 PSTN and broadband/Bitstream volumes.......................................................... 42 3.3.2 Additional network volumes ............................................................................... 44 3.3.3 Non-PSTN volumes ........................................................................................... 44 3.3.4 How the volumes are used ................................................................................ 44 3.4 Cost inputs.......................................................................................................45 3.5 Network Design Rules .....................................................................................46 3.6 Switching .........................................................................................................48 3.7 IP Network equipment .....................................................................................48 3.8 Transmission and infrastructure ......................................................................50 3.9 Routing factors ................................................................................................51 3.10 Shared costs in the core model.......................................................................52 3.11 Core model calculations ..................................................................................52 3.11.1 Volume calculations........................................................................................... 52 3.11.2 Routing table calculations .................................................................................. 53 3.11.3 Adjustment of traffic to RSMs ............................................................................ 53 3.11.4 Traffic profiling ................................................................................................... 53 3.11.5 Building calculations .......................................................................................... 53 3.11.6 Switching calculations........................................................................................ 54 3.11.7 Transmission ..................................................................................................... 55 3.11.8 Core Trenching .................................................................................................. 57 4 Access model......................................................................................................58 4.1 Definitions and assumptions ...........................................................................59 4.1.1 Geotypes ........................................................................................................... 59 4.1.2 Sampling............................................................................................................ 59 4.1.3 Main assumptions.............................................................................................. 61 4.2 Structure of access model ...............................................................................62 4.3 Modelling the access network .........................................................................62 4.3.1 Calibration ......................................................................................................... 63 4.3.2 LRIC costs of each resource.............................................................................. 64 4.3.3 Tätort ................................................................................................................. 64 4.3.4 Copper cables and nodes .................................................................................. 65 4.3.5 Copper cabling: (i) PDPs ................................................................................... 66 4.3.6 Copper cabling: (ii) SDPs and street level network, from SDP to NTP .............. 67 4.3.7 Copper cabling: (iii) use of cables on poles ....................................................... 68 4.3.8 Copper cabling: (iv) summary of design parameters to be optimised ................ 69 4.3.9 Fibre .................................................................................................................. 70 4.3.10 Trench, Duct, Miniduct, and route sharing ......................................................... 71 4.3.11 Fixed Wireless access ....................................................................................... 72 4.3.12 Final Network Design – in particular, the decision of where to deploy FWA ...... 73 4.3.13 Output to consolidation model ........................................................................... 73 4.4 Modelling the access network: equipment at the scorched node and links to island sites................................................................................................................74 4.5 Shared costs in the access network................................................................75 4.6 Network elements for the access network.......................................................75 4.7 Bitstream access .............................................................................................76 Post- och telestyrelsen ii 5 Co-location model...............................................................................................77 5.1 Definitions and assumptions ...........................................................................77 5.2 Structure of co-location model.........................................................................77 5.3 Modelling co-location services ........................................................................78 5.4 Direct costs in the co-location model...............................................................80 5.4.1 Installation costs related to location of equipment ............................................. 80 5.4.2 Installation costs and annual costs relating to cable products ........................... 80 5.4.3 Power consumption costs .................................................................................. 81 5.5 Shared costs in the co-location model ............................................................81 5.6 Common costs in the co-location model .........................................................82 5.7 Other service costs included in the co-location model ....................................82 5.7.1 Bitstream Operator Access ................................................................................ 83 5.7.2 Raw copper........................................................................................................ 83 5.7.3 Shared raw copper and Bitstream access ......................................................... 84 5.7.4 Regional and local POI ...................................................................................... 84 5.7.5 Interconnection capacity .................................................................................... 85 5.7.6 Information requests .......................................................................................... 85 5.7.7 DACS/line conditioner removal .......................................................................... 86 5.7.8 Change of operator or shared to full access ...................................................... 86 5.7.9 Suspension/Suspension Removal of shared access ......................................... 86 Appendix 1 List of abbreviations used in this report .............................................87 Post- och telestyrelsen iii 1 Introduction 1.1 Background This hybrid model is the result of a process initiated by Post & Telestyrelsen (PTS) in March 2007 to revise the original hybrid model developed in 2002/2003. At an early stage of the original model development in 2002, PTS released for discussion a proposed timetable showing how each of the main activities would be conducted. This timetable was discussed and accepted by the parties. The key dates (together with dates of subsequent revisions and updates, including this one) were as follows: May 2002 – September 2002: Structuring the process August 2002 – March 2003: BU modelling August 2002 – June 2003: TD modelling June 2003 – November 2003: Reconciliation and hybrid modelling November 2003 – December 2003 Cost results and pricing methodology October 2004 – December 2004: Updated cost results for 2005 October 2005 – December 2005 Updated cost results for 2006 September 2006 – November 2006 Updated cost results for 2007 June 2007 – December 2007 Revised BU modelling December 2007 – April 2008 Revised Hybrid modelling Updated cost results for 2008 August 2008 – December 2008 Updated cost results for 2009 Consultations with the Swedish telecommunication industry have been conducted at all stages of the process and opportunities have been given to influence the model structure and features. The current version of this documentation refers to version 6.1 of the Hybrid model, published in December 2008. This is an update of the hybrid model version 5.1 which was published in April 2008. 1.2 Changes to model documentation The model documentation has been updated and revised to reflect the changes made to the version 6.0 of the hybrid model. The following sections have been updated or changed. 2.4.1 Cost of Capital 5.3 Modelling co-location services Post- och telestyrelsen 1 1.3 Summary of results The results produced by the hybrid model are presented in the Output and Results worksheets of the Consolidation model. 1.4 Structure of this document This document describes the structure and principles behind the Hybrid Long Run Incremental Cost (LRIC) model developed by PTS. The hybrid model consists of several individual models that are inter-related and work together. Together, they form one model. This document uses the word model to mean both one of the individual models as well as the overall model that is formed by the combined individual models. This document does not describe in detail how the models actually work and the nature and role of each of the spreadsheets – this is described in the User Guide. The document describes each of the four main models: • Consolidation. This combines results from each of the models below and calculates the resulting service costs. • Core. This calculates the cost of core network services and systems. It also includes some access costs. • Access. This calculates the costs of access services such as raw copper. • Co-location. This calculates the cost of services areas that may be used by other operators to co-locate equipment at TeliaSonera sites. An overview of each model is provided in Section 1.6 below. Subsequent main sections describe each model in greater detail. In practice, the hybrid model is a revised version of the original bottom-up model. All the changes (in summary form) made to the bottom-up model following the reconciliation process are documented in the “Changes” sheet in each model. 1.5 Overview of each model This section provides a general guide to each of the main models, which together form the overall hybrid model. As mentioned previously, the hybrid model consists of four components that are linked together: consolidation; core; access; and co-location models. These are shown in Figure 1 below. Post- och telestyrelsen 2 Figure 1: Overall structure of the models Input Co-location data model Consolidation model: Input Core - transfer of costs data model - annualisation - validation - - costing services - Input Access data model The models have some shared data – common data that is used by each model. These have been linked so that the original inputs are made in one model only. The consolidation model has a verification function (validation) that enables checks to see that the different models are using the same data. For greater transparency, the main inputs shared between different parts of the model are shown in a dedicated interface sheet in each feeder model (I_Interface). Each model is largely self-contained – it carries out almost all of the calculations associated with its services. To a certain extent, this enables each model to be developed independently. However, where appropriate the models are inter-linked to ensure internal consistency between product names and volumes. The final results are calculated in the consolidation model. It is here that data from each model is collected and processed into the final service costs. Outside the actual models, some additional analytical work has been required for some areas. These are referred to as “off-line” calculations, as they do not form a part of the models. They are, however, no less important to the model since the off-line calculations create input values for the models. The main off-line analysis concerns core and access street km analysis. This analysis has been made using Microsoft Autoroute and maps from TeliaSonera's wholesale website. This work produced information about the street km distances and hence provides the basis for trench and cabling calculations in both the core and access models. Cost inputs are often confidential. Some versions of the models will have disguised values. The source data is identified by comments fields. The source name is identified by a code. The code can be traced back to the name of a source using a code sheet (this is confidential). Due to camouflaging of some input parameters, the final result presented in the public version of model will differ slightly from the result obtained when running the model with confidential inputs.1 1 The actual price setting is based on the original data and not the camouflaged data. Post- och telestyrelsen 3 1.5.1 Consolidation model The consolidation model brings together the outputs of each of the main models. The consolidation model collects the cost data from each model. Additional numerical data about the number of services, and other technical values, are also collected. The data is linked into the consolidation model. The cost data is annualised using annuities. This annualisation converts the capital costs of equipment into average annual costs, based on the equipment lifetimes, price trends and scrap values. In addition, a capital charge is added using the cost of capital. Each cost item is given an allocation – this defines what network element or service the cost relates to. The cost is then allocated and network elements are transformed into service costs using a routing / allocation table technique. The allocation allows cost items that were calculated in one model to be transferred to other services (thus access line cards or main distribution frame costs are calculated in the core model but are used as part of access service calculations in the consolidation model). The routing tables are constructed using data supplied from the main models. The routing tables define how each product uses the network elements and, along with product volume data, they enable the services to be costed from the network element cost information. Additional inputs and calculations in the consolidation model allow an initial allocation of operating costs (in the core and access model) using mark-ups to be transformed to operating costs using a functional area approach. Furthermore, indirect and common costs that do not constitute part of the other models are input and allocated to the final service costs. A Sensitivity Analysis sheet is included within the consolidation model. This compares updated outputs with outputs from a previous run of the model so that users can see the impact on final results of any changes they have made to key inputs in the model. The sensitivity sheet also provides a network element breakdown of the per unit cost of certain key products and an Aggregated Outputs section summarising overall totals for GRC, Annualised GRC and Opex across a number of areas. 1.5.2 Core model The core model calculates the network systems and associated costs that are needed for a network operation of the scale of TeliaSonera. It calculates the cost of switching and transmission systems, both for the PSTN and an Ethernet/IP Network. The core model therefore deals with the element costs that are driven by traffic (PSTN call/minute volumes and broadband/bitstream usage) in contrast to the access model, where costs are driven by number of customers. The core model should be understood to have a different boundary to the core network. The core model also calculates costs of line cards and the main distribution frames (MDF). These costs are part of the access network, but are included within the core model calculations. These access-related costs are allocated to access service costs in the consolidation model. Furthermore, since the allocation of costs between access and core relies on the boundary between access and core as stipulated by TeliaSonera’s Post- och telestyrelsen 4 network, some core model costs related to transmission are allocated to access in the consolidation model. The starting point for the core model is the volume data for each service. The PSTN call services, IP network services and other non-PSTN services define the overall network size. The dimensioning of the network for PSTN is done through a routing table that defines how each service uses the network, while a separate routing table defines IP network usage by broadband and bitstream products. Taking account of how remaining non-PTSN services such as IPTV and leased lines use the network gives an overall dimension for the network. The equipment needed to create the overall network is calculated next – technical design rules are used to calculate the numbers and sizes of each element. The final costs of the many network elements needed for the PSTN or broadband/bitstream services are finally exported to the consolidation model along with routing table data and volumes data. Off-line calculations are used to estimate inter-site distances and building costs per square metre. Key features of the core model are: • All products are now by default assumed to run over a hierarchical Ethernet/IP network. The hierarchical layers comprise DSLAMs, aggregation/Metro Ethernet switches, edge IP routers and core IP routers, with additional equipment included specifically for voice telephony services (softswitches and media gateways); • trench lengths are based on actual route lengths for transit-transit, transit-local and local-local links using Autoroute mapping, and on estimated route lengths based on sampling for remote-local links. All inputs are shown in the I_Trenching input sheet. An assessment has been included of the need for repeaters to cater for long routes (those in excess of 80km); • logical/physical rings are used to provide resilience. Optical systems are usually used, but microwave can also be used; • some sites are connected via spurs, due to local geographical features such as valleys. If the NGN selector is switched off then voice telephony products are assumed to run over a non-IP network comprising: • three layer switch network assumed with telephony servers providing some of the functionality currently provided within local exchanges and tandem switches; • additional international gateway and IN platforms as required; • circuit-switched voice technology is assumed for PSTN services, based on the Ericsson ENGINE solution; • transmission for voice telephony products based on SDH (Synchronous Digital hierarchy), whereas for broadband/Bitstream and IPTV services still using the IP Network it is based on direct Ethernet and/or IP links. Post- och telestyrelsen 5 1.5.3 Access model The access model calculates the equipment and costs needed to create an access network for Sweden with the scope of services and demand as faced by an operator with SMP such as TeliaSonera. The access model calculates the amount of cables and equipment needed to connect from the Main Distribution Frame (MDF) to the customer premises. The main items in the access network are copper cables, ducts and trenching. Distribution points and splitting points are also required. Fibre is also used in the loop and hence fibre distribution links are also included. The access services are calculated in the consolidation model. This stage allows for additional costs that are calculated in the core model to be added to the cost calculated in the access model. Figure 2: Overall scope of the access model (PSTN viewpoint) LE LE Demux RSS MDF site Line RSM Line RSM cards cards Access FAM Access Network Network PDP PDP Model Model PDP PDP SDP SDP SDP SDP SDP SDP Customer Site NTP Fibre Copper The diagram shows access model calculated costs from the MDF via a tree- and branch style network to customer premises. Primary distribution points (PDP) may be used to split the cables. Secondary distribution points (SDP) can also be used to split the cables to the final drop-off points. These final drop points or final splits (not shown) link the customer site to the cables in the street – over the final drop. Not all links will have the need for primary and secondary splits as well as the final split – the majority need only one distribution point. Fibre access Multiplexers (FAMs) are used in the street (where needed) to provide optical systems links to copper final delivery to customers. There is also a facility to Post- och telestyrelsen 6 use Fixed Wireless Access (FWA) systems in areas where they might prove more cost effective. The costs of the cables and equipment are calculated using data about the populations and node sizes. Much of this analysis is carried out by geotype and is done as an off- line analysis. 1.5.4 Co-location model This model calculates the systems and costs required to equip space in TeliaSonera buildings that is suitable for co-location space services. The main components of the co-location services modelled at different sites in the SMP operators network are: • Location of equipment; • Installation and mounting of equipment; • Station wiring; • Placing; and • Power, cooling and ventilation. Co-location is relevant in relation to switched interconnection, access to unbundled local loop, and other potential purposes. The model only explicitly considers the co- location costs of the unbundled loop. However, it takes account of sharing of costs between other co-location services and other increments. Unlike services in the core and access network, co-location services consist of relatively few cost categories. They are mostly standalone “sub-products” that may be combined by the operator who demands co-location. Therefore, although the co- location model is simpler in structure compared to both the core and access models, cost inputs are more detailed in order to capture costs at a sufficiently granular level. The main co-location cost is the cost of space. It is assumed that space in buildings is, in the long run, an incremental cost – hence the building size is variable in the long run. Without this assumption the building costs would be fixed. For practical modelling purposes the co-location model also calculates the costs of other services, including interconnection capacity, regional and local POI and installation of services such as voice telephony lines, Bitstream, raw copper and shared raw copper. 1.6 Definitions and principles common to all models Some concepts and definitions are used throughout this document and it is useful to understand the main ones. Scorched Node. This is a TeliaSonera site that has a voice switch or concentrating equipment (Remote Subscriber Stage – RSS) that has been used to replace a voice switch. The location and number of these nodes cannot be altered. The equipment within each may be altered (or “scorched out”). A scorched node may contain several different types of equipment and it is typically a building varying from a small hut in a village to a large exchange site in a city. Nodes with fewer than 30 lines in use and no Post- och telestyrelsen 7 DSLAM are deemed to be part of the access network and have been removed from the core network definition. MDF-site: This is the location where the Main Distribution Frame (MDF) is located. The MDF is located together with a switch, concentrator or Remote Subscriber Multiplexer (RSM) and generally also a DSLAM. In principle, an MDF-site with only a small (<30 lines) RSM (no switching or concentrating functionality) and with no DSLAM is not considered the access/core demarcation point under the scorched node definition, as defined by the MRP. As one of the main purposes of the access model is to calculate the cost of copper access, however, and copper access is defined from the MDF to the customer premises, PTS has decided to take the existing MDF locations in TeliaSonera’s network for given as well. Geotype. Each site can be classified to be in one of several geotypes. A geotype depends on the density of subscribers per km2. Costs of services may therefore vary by geotype. Geotypes enable the model to represent the diversity of areas in Sweden, whilst avoiding the need for detailed analysis and estimation for every one of the 8,000 or so switch/concentration zones. The geotypes used in the model are: • City: over 1,000 lines per km2 • Urban: 100 to 1,000 lines per km2 • Rural A: 10 to 100 lines per km2 • Rural B: 1 to 10 lines per km2 • Sparse: up to 1 lines per km2; at least one access network subscriber line • Empty: no access subscriber lines. Access Zone. Each MDF is assumed to have an access zone around it. The subscribers in the zone connect to the node via the local access network for the zone. Subscribers are generally connected to the scorched node that they are closest to (with a few exceptions where local geography makes it more cost effective to connect subscribers to another nearby zone due to an obstacle such as a lake or due to local clustering of households). A zone is typically a few km2 (city) up to a few hundred km2 in area (rural). The overall area covered by a node depends on the access technology used – the limits of copper cables set restrictions on the distance customers can be located away from the node. The scorched node assumption means that the zones are effectively fixed. Note that there are many parts of Sweden that require no access zone at all, as there are no customers in the “zone” – the zone has only lakes, forests and mountains with no population to service. These are allocated to the last of the geotypes listed above (“Empty”), and play no further part in the analysis of costs in the access network. Fibre access multiplexer (FAM). This is an item of electronics that multiplexes copper subscribers with fibre-accessed customers. The combined data is linked back to a scorched node via fibre optic link where the data is de-multiplexed. A FAM is typically in a street cabinet or larger customer-building basement. Remote subscriber multiplexer (RSM). This is a scorched node site that has multiplexing equipment. The RSM combines data and voice service from customers and transmits Post- och telestyrelsen 8 them over a fibre link to another scorched node site where they are de-multiplexed and linked to other systems such as voice switched and data systems. Any voice-switch site may be converted to an RSM and vice versa under the Scorched node rules. The RSM will have copper line termination cards. Core-access demarcation. For ease of modelling, the demarcation of the models and the services are different. The access service includes all equipment from the customer premises up to the scorched node, including the line cards in the scorched node. Thus, access includes copper terminating line cards in the RSS, voice switch and DSLAM. The MDF is also included in the access network costs. The model demarcation, however, is at the MDF site where the access model includes all the costs from the Network Termination Point (NTP) up to (and excluding) the MDF. The differences of core/IP and access models and networks are illustrated in Figures 3a and 3b. Please note that the core-access demarcation is defined in relation to TeliaSonera’s actual network. The access network is defined up to (and including) the line card located in the RSS or LE or DSLAM (existing small RSMs where there is no DSLAM will be located in the access network). In the hybrid model, a large number of RSSs are replaced by RSMs. The access-core demarcation remains at this site, even though the node in principle has no voice switching or concentrating functionality (though it generally does have broadband/Bitstream concentrating functionality). This is to ensure that the costs modelled correspond to actual services provided2. Figure 3a: Model demarcation – Core and access networks Core Network Transit Switch TS (TS) Model Core Local Network Exchange (LE) LE LE Demux RSS Line RSM cards RSM Line cards Access FAM Network PDP PDP Model Access Network SDP SDP SDP PDP PDP Customer Site SDP SDP SDP NTP Fibre Copper The core model includes the line card and MDF calculations, even though these are access network cost items. 2 If this were not the case, costs could be artificially transferred from core to access, simply by replacing an RSS with an RSM. The costs would be allocated to subscription, but not to copper access. Post- och telestyrelsen 9 Main distribution frame (MDF). The main termination point for copper access cables in a scorched node site. The copper pairs from the customer are linked from the MDF to the equipment in the scorched node. Spur. A path to one or more sites that has only one physical route. Ring. A logical or physical connection that has two paths that therefore can (optionally) provide alternative routes should one path round the ring fail. Logical versus physical link. The logical link (or links) is the path between two items of equipment. The physical path is the actual route taken by the data between the equipment. There may be two logical paths, but it is possible for these to be on the same physical path or on diverse paths. Capital versus operational costs. A fundamental feature of the model is the different determination of capital-related costs and operational costs. Capital costs (or capex) are a result of the purchase and installation costs of the equipment. The total capital cost of a particular type of item is the cost of one item multiplied by the number of items required. The lifetime of the equipment, price trends and scrap values are also capital-related data. They are used, along with the purchase costs, to determine the average cost per annum – the annualisation calculation is carried out in the consolidation model. The depreciation is calculated in the consolidation model (and is only related to the capital cost). Operational costs are a result of maintaining, operating and repairing the equipment once bought. This is an ongoing or annual cost. Operational costs are derived through the functional area approach (see below), hence the capital cost and operational costs are not directly related. Figure 3b: Model demarcation – IP and access networks Post- och telestyrelsen 10 IP Core Router Network (CR) CR Model ER IP Edge Router ER Network (ER) RSS DSLAM Line cards Access Network PDP PDP PDP PDP Model Access SDP SDP SDP Network SDP SDP SDP Customer Site NTP Fibre Copper Functional areas and operational cost calculations. Operational costs are based on a definition of functional areas. These are areas of the business that are needed to carry out a set of related functions such as operational work on (say) switches. Functional areas are defined to have a variable number of staff dependent on a staff driver and subject to a minimum number. The costs of this functional area are therefore indirectly dependent on the capital cost. The models allocate the cost of the functional area to the capital items in proportion to a cost factor. The cost factor is defined as a percentage of the capital cost – the percentage is an initial estimate of the operating cost as a fraction of the purchase cost. Costs are allocated pro-rata. Therefore if the capital cost of equipment varies, then the total operational cost is still fixed (it is defined by the functional area size). However, if the quantity of the equipment varies then this might well vary the total operational cost, depending on the drivers being used. The allocation of the total operational cost will also vary slightly. However, if every item’s capital cost increases similarly, the allocation will remain the same. Post- och telestyrelsen 11 2 Consolidation model 2.1 Definitions and assumptions The consolidation model uses the costs relating to core, access and co-location that are produced by each of the separate models. The costs are brought together in the consolidation model. The consolidation model produces the final cost of each service. It also undertakes some checks for consistency between the other three models. As well as annualised costs, any cost may be expensed, hence the cost is recovered as a one-off payment. Some cost items that are relevant to access or co-location services are typically treated in this way. Core PSTN or IP network costs would not normally be expensed. The model therefore calculates service costs as a mixture of one-off and annualised costs. The choice of annualisation or expensing is essentially a pricing decision. 2.2 Structure of consolidation model The consolidation model structure is captured in the navigation map. This is reproduced in the diagram below. Figure 4: Consolidation model - navigation map The consolidation model has three main stages – input, calculations and output. The main functions of the consolidation model are the calculation of service costs from the cost category inputs of the core, access and co-location models, and the integration of functional area costs, including common business costs. Post- och telestyrelsen 12 2.3 The main functions of the consolidation model The consolidation model essentially carries out three key functions: • annualisation of capital costs to give an annual cost. Optionally the cost item may instead be expensed (treated as a one-off cost); • allocation of operational costs from the functional area approach to network elements; and • calculation of service costs. These functions are discussed in more detail in the sections that follow. The consolidation model contains come central data values and calculations, but its prime purpose is to consolidate all of the calculated costs from each model and calculate the service costs from these cost inputs. The design is based on disaggregated cost data. This approach, based on a collation of cost data, enables all costs that contribute to each service to be identified individually. The collated input costs from each model are annualised (or expensed as appropriate). This involves adding together different cost types such as equipment costs, installation costs, operating costs, indirect costs and common costs to derive a single cost (annual or one-off) that represents the long-run cost that must be recovered. The assumed cost of capital, price trends, lifetimes and scrap values are combined in this calculation. The disaggregated approach also allows the user to use alternative annualisation formulae for any cost category. A number of different annualisation options have been provided. These are discussed in more detail in the next section. However, the default method used by PTS is tilted annuities. The annual cost of each cost category is allocated to network elements or services. The user is allowed the flexibility to define the allocation to use. The resulting element costs are then processed into service costs. Routing tables and other allocation/combinatorial techniques are used as required. The final cost of services is then subject to an “uplift” for common business costs. A feature of the approach taken is that the costs, when input to the consolidation model, can be altered (or overwritten) to give another value and the results directly evaluated. This approach to sensitivity analysis is not a normal action (it invalidates the true results), but it is easy to carry out. In addition, a sensitivity analysis sheet in the consolidation model allows the user to take a copy (“snapshot”) of the original results before any changes are made. These are then compared with the altered results so the user can assess how the revised inputs impact on the model results. Another approach to sensitivity analysis is to alter the allocation of the cost category. If a cost category is not given an allocation (by removing or deleting the network element allocated to the particular cost category) then the cost is taken out of the service cost. It is easy to compare the new value with the normal value and hence see the sensitivity of any service to a particular cost category. This type of analysis is also referred to as the “delta” method – the user can quickly see the difference or delta caused by any one cost. This means that the contribution of any one cost to a service can be easily be evaluated. Post- och telestyrelsen 13 2.4 Annualisation assumptions 2.4.1 Cost of Capital The cost of capital measures the opportunity costs of the sources of capital (debt and equity) invested in the company. PTS originally estimated the cost of capital for a fixed SMP operator in Sweden in accordance with the guidelines determined by PTS on 15 October 20033. A review of the cost of capital was conducted in 2007 and the results of that review are now included within the hybrid model. 18.104.22.168 Overall approach In line with the adopted guidelines, PTS has applied the same cost of capital for core, access and co-location services and estimated the cost of capital on a nominal and pre- tax basis. The Weighted Average Cost of Capital (WACC) is calculated as the weighted cost of debt and equity: E D WACC = × Ce + × ( 1 − T) × C d , E+D E+D where E is the market value of equity, D the market value of debt, E+D is the market value of the company, Ce the cost of equity, T the effective tax rate, and Cd the cost of debt. The cost of debt, Cd, reflects the interest rate that lenders would require for lending their money, i.e. the risk free-rate adjusted to reward lenders for the risk that the borrower will default. The cost of equity has been estimated according to the Capital Asset Pricing Model (CAPM), according to which the cost of equity is calculated as Ce = E(Rj) = Rf + βj [ E(Rm) – Rf ], where E(Rj) is the expected return on asset j; Rf is the risk-free rate; βj measures how sensitive asset j is to movements in the market portfolio; and E(Rm) is the expected return on the market portfolio.[ E(Rm) – Rf ] is the market risk premium, in practice often referred to as the Equity Risk Premium (ERP). The calculations of the individual parameters are discussed below. 22.214.171.124 Cost of debt The cost of debt has been estimated as the sum of the risk free rate and a debt premium. 3 Konsultationsrapport – kommentarer rörande AMI:s rapport om kapitalkostnad för svenska SMP- operatörer. Post- och telestyrelsen 14 The risk free rate has been estimated at 4.28% using a 6-month average of a 10-year Swedish government bond4. PTS has estimated the optimal gearing for a fixed operator to lie between 30-50%. The debt premium has been estimated at 0.8% and 1.30% for a gearing of 30% and 50% respectively. Finally, PTS has used the corporate tax rate of 28% as proxy for the effective tax rate. 126.96.36.199 Cost of equity The cost of equity has been estimated using the perspective of a marginal international investor. The international market risk premium has been estimated to 4.75%. Beta has been estimated to 0.6. 188.8.131.52 Summary of calculations The calculations of the WACC using the above parameters are summarised in the table below: Table 1 Estimating the cost of capital for fixed SMP operator Low gearing High gearing Risk-free rate 4.28 4.28 Equity risk premium 4.75 4.75 Unlevered Beta 0.60 0.60 Levered Beta 0.79 1.03 Cost of equity 8.01 9.18 Debt premium 0.80 1.30 Cost of debt 3.66 4.02 Gearing 30 50 Tax-rate 28 28 Post-tax WACC 6.70 6.60 Pre-tax WACC 9.31 9.16 Range 9.31 – 9.16 Mid-point 9.2 On this basis, PTS has applied a cost of capital of 9.2% in the hybrid model. 2.4.2 Annualisation options The consolidation model offers a number of annualisation options. These are: • Straight-line depreciation 4 Up to and including December 2007. Post- och telestyrelsen 15 • Tilted Straight-line depreciation • Sum of digits depreciation (front loaded) • Standard annuity function • Tilted annuity function Straight-line depreciation divides the asset’s price by the asset’s life to produce an annual depreciation charge. To calculate the annualisation charge, a capital charge is added. The straight-line annualisation factor used in the model is: ⎛ SV ⎞ ⎛ 1 ⎞ ⎜ CV − ⎟×⎜ + CoC ⎟ , ⎜ ( 1 + CoC ) AL ⎟ ⎝ AL ⎠ ⎝ ⎠ where CV is the capital value of the asset, SV the scrap value of the asset, AL the asset life and CoC the cost of capital. Tilted straight-line depreciation takes account of expected price changes for assets. It will result in a steeper depreciation profile when prices are falling than unadjusted straight- line depreciation. The tilted straight-line annualisation factor used in the model is: ⎛ SV ⎞ ⎛ 1 ⎞ ⎜ CV − ⎟×⎜ + CoC − PT ⎟ , ⎜ ( 1 + CoC ) AL ⎟ ⎝ AL ⎠ ⎝ ⎠ where PT is the price trend. The sum of year digits (SOYD) is a simple method for generating a front-loaded depreciation schedule. It may be a useful approximation if the asset’s operating costs are expected to rise or its price or the revenue it generates is expected to fall. The sum of years digits annualisation factor used in the model is5,6: ⎛ SV ⎞ ⎛ 2 ⎞ ⎜ CV − ⎟×⎜ + CoC ⎟ ⎜ ( 1 + CoC ) AL ⎟ ⎝ AL + 1 ⎠ ⎝ ⎠ The annuity approach calculates both the depreciation charge and the capital charge. A standard annuity calculates the charge that, after discounting, recovers the asset’s purchase price and financing costs in equal annual sums. In the beginning of an asset's lifetime the annualisation payment will consist more of capital charges and less of depreciation charges; this reverses over time resulting in an upward sloping depreciation schedule. The increase in the depreciation charge over time exactly counterbalances the decrease in the capital charge with the result that the annualisation charge is constant over time. The standard annuity function used in the model is: 5 Note the formula used is a simplified version of the sum of digits (front loaded) annualisation formula. This simplification is possible since the costs we are modelling are those for the first year in an asset’s life. 6 Note that sum of years digits depreciation may also be back-loaded. This is the reverse of the depreciation under sum of years digits front-loaded. Post- och telestyrelsen 16 ⎛ SV ⎞ CoC ⎜ CV − ⎟× ⎜ ( 1 + CoC ) AL ⎟ AL ⎝ ⎠ ⎛ 1 ⎞ 1− ⎜ ⎟ ⎝ 1 + CoC ⎠ A tilted annuity calculates an annuity charge that changes between years at the same rate as the price of the asset is expected to change. This results in declining annualisation charges if prices are expected to fall over time; for a large enough tilt the slope of the depreciation profile will also be negative. As with a standard annuity, the tilted annuity should still result in charges that, after discounting, recover the asset’s purchase price and financing costs. The tilted annuity function used in the model is: ⎛ SV ⎞ CoC − PT ⎜ CV − ⎟× ⎜ ( 1 + CoC ) AL ⎟ AL ⎝ ⎠ ⎛ 1 + PT ⎞ 1− ⎜ ⎟ ⎝ 1 + CoC ⎠ The choice of the depreciation methodology should ideally be the one which best reflects economic depreciation. This implies that holding gains and holding losses, which follow from changes in asset prices, should be taken into account. Compared to the tilted straight line depreciation formula, the tilted annuity approach has the advantage that the annualisation charge is independent of the age of the asset. The fact that the hybrid model is (artificially) modelling new assets therefore becomes less of an issue7. PTS has therefore decided to annualise costs on the basis of tilted annuities8. This approach was also used in the bottom-up model and is supported by TeliaSonera. Note that the majority of equipment costs and installation costs are passed on to the consolidation model, where they are annualised. However, there are a few exceptions. These are building costs and common site costs that are annualised in the core model. 2.4.3 Annualisation parameters Price trends, residual (scrap) values and equipment lifetimes are specified for all cost categories and are used in the calculations. These inputs are made in the separate core, access and co-location models. The consolidation model merely imports these values and then uses them in the annualisation process. The hybrid model uses the economic life of equipment to measure asset lives. An overview of the asset lives and (forward-looking) price trends used are provided in the table below: 7 Otherwise, it could be argued that the bottom-up model should not model the costs in year 1 but rather in year 3 or 5 for example. 8 Note that tilted annuities do not work well when asset prices are declining rapidly (understating costs compared to economic depreciation) or where asset prices are rising over time (overstating costs compared to economic depreciation), although less so than many of the other methodologies. Hence, using tilted annuities may result in understatement of core costs and overstatement of access. Post- och telestyrelsen 17 Table 2 Asset lives (years) and price trends in the hybrid model Cost category Asset lives Price trends Access Trench (incl. mini-duct trench) 40 +2% Access Duct (including mini-duct) 40 +2% Core trench 40 +2% Core duct 40 +2% Poles 20 +2% Copper cable 25 +6% Line cards 10 -3% Fibre cable 20 -5% Cabinets/distribution points 15 +1% Manholes 40 +1% MDF 15 0% NTPs (copper) 20 0% NTPs (fibre) 20 -5% Frame unit 10 -2% Switchblock unit 10 -4% Processor unit 10 -5% Software 10 -4% Port unit 10 -3% ODF 10 -5% ADM 10 -5% STM Multiplexers 10 -5% STM Cards 10 -3% Synchronization 10 0% DSLAMs 5 -5% Aggregation switches 5 -7.5% IP routing equipment 5 -7.5% Media gateways 5 -7.5% Softswitches 5 -7.5% Cross-connects 10 -4% Signalling points 10 -4% Submarine cable links 20 -4% Microwave 15 -8% IN 10 -5% Power supply unit 10 -2% Back-up power 10 -2% Air conditioning unit 10 1% Security system 10 1% Site preparation 15 0% Buildings 40 +2% Post- och telestyrelsen 18 Although the model has the functionality to use scrap values, these are assumed to be zero for almost all cost categories. The only cost category using a scrap value as an input is building costs. Here the scrap value is set at 30% of today’s building value. 2.5 Working capital Working capital costs are the costs of maintaining balances of physical or financial stocks (assets and liabilities). The cost of working capital is calculated by multiplying the cost of capital with the calculated working capital. Based on empirical evidence from the top-down model, the cost of working capital has been set to zero. 2.6 Functional area costs The approach to modelling operating costs and indirect costs relies on a functional area approach. The approach uses information on allocations performed by mark-ups in the separate models to allocate costs of each operational (or functional) area that is needed. Each of these areas is dimensioned in order to define the total operational costs expected for an efficient operator (using average cost per staff member). These results define the expected total operating and indirect costs required to run the domestic PSTN and IP network of an operator with SMP in Sweden. This approach consists of several stages: • Define the operational areas and staff/non-staff drivers; • Calculate the size of each area; • Calculate the cost of each area; and • Allocate the cost of each area to network elements, such that the total is equal to the sum of the functional areas. The latter allocation of costs to network elements is performed by means of an allocation table and weights provided by an initial allocation of costs using mark-ups in the core and access models. 2.6.1 Functional areas The FA costs in the I_FA_Costs sheet have been grouped into four different major cost categories: • Network costs; • Non-network costs; • The IC and Access specific costs; and • Overhead costs. These costs have been subdivided into the following areas: Post- och telestyrelsen 19 Figure 5: Detailed depiction of the FA cost categories Costs Network costs relating to ongoing costs* Non-network costs IC and access specific costs Switched network management IP Network management Corporate Overheads Customer oriented costs Switched network maintenance IP Network maintenance Human resources Billing Switched network planning IP Network planning Finance Debtor handling Network management system IP Network management system Support systems Other IC specific Core infrastructure management DSLAM management Admin Areas Core infrastructure maintenance DSLAM maintenance Core infrastructure planning DSLAM planning Core transmission equipment Access infrastructure management management Access infrastructure maintenance Core transmission equipment Access infrastructure planning maintenance Access equipment management Core transmission equipment planning Access equipment maintenance Access equipment planning * the hybrid model caters for additional categories of network costs related to ongoing costs that are not used. These include: ‘Site management (excl. HQ)’, ‘Field Engineering – core’, ‘Field Engineering – access’ and ‘others - to be added if needed’. Note that some of the cost categories stated above may have no staff costs, only non-pay costs. While it is likely that all categories will comprise elements of pay and non-pay, the underlying input data used in the off-line analysis was not sufficiently detailed to achieve such an allocation in all cases. Post- och telestyrelsen 20 2.6.2 Size of each area For each of the identified functional areas above, the model includes inputs for staffing requirements, i.e. the number of staff required for each area and the non- pay costs (broken down into three types of non-pay cost: Outsourced Staff; Systems; and Other). The total staffing requirements and non-pay items determine the size of each area. Note that the inputs to staffing requirements are strictly based upon head-count numbers. Subsequently, the allocation between the different staff types: managers, technician and support is made by means of a user defined allocation profile. Distribution to different staff types is done at a later stage in the calculation, cf. below. Inputs to each FA are a result of an off-line analysis of various data sources, in particular input data from the top-down model and inputs received during the various consultation processes that have taken place. The inputs used are based on an evaluation of the requirements for an optimised network with the scope and size of the SMP operator and the actual numbers in Skanova as stated by the top-down model from TeliaSonera9. As an entry point, the costs in the hybrid model are based upon the numbers found in the top-down model and figures provided by TeliaSonera. However, these costs have subsequently been analysed and modified where appropriate to ensure consistency with the underlying cost inputs. In addition, a number of sanity checks have been applied to the off-line analysis to ensure that the cost levels and allocations (between categories) are appropriate. Generally, three main types of corrections have been made: 1) Technological. The hybrid model has a different network structure than that of TeliaSonera’s actual network, particularly with the NGN Selector being switched on as the default. 2) Retail. Specific retail costs should not form part of the FA costs. Retail costing has been carefully removed from all costing elements, ensuring the correct input values are used in the model 3) Non-PSTN. Core costs should be input net of non-PSTN (but including IP) costs, while access FA costs should include non-PSTN costs. 2.6.3 Cost of each area The cost of each area is calculated using staffing profile assumptions – %-manager, %-support and %-technical (in-house and outsourced) – and the average annual cost of the different staff types. 9TeliaSonera has provided PTS with an efficiency study of 8 October 2003, in support of TeliaSonera’s claim that its actual costs reflect the costs of an efficient operator. Post- och telestyrelsen 21 The staff costs include social and pension costs, training and education, fringe benefits, company cars, healthcare etc. and are based on information provided by TeliaSonera to ensure consistency with other FA inputs. The non-pay costs can be defined in terms of three types of non-pay cost: Outsourced Staff; Systems; and Other. Outsourced staff requirements are calculated in a similar manner to in-house staff requirements, using appropriate drivers. This method has also been extended to IP related support systems. 2.6.4 Allocation of the cost of each area In order to allocate the costs using the functional area approach to services, the model relies on two different approaches: • allocation to network elements and hence to services; and • allocation directly to services using a mark-up approach. Direct network costs are allocated to network elements, while the remaining functional area costs (overheads costs comprising indirect costs and interconnect and access specific costs) are allocated using mark-ups. In order to allocate the direct network costs to network elements, the model utilises an allocation table and the operating costs already allocated to different network elements by using operating cost mark-ups in the core and access model. The allocation table consists of zero (do not allocate costs) and one (allocate costs). Using this allocation table and the costs allocated to each network element by using mark-ups, the model calculates the functional area costs to each network element. The formula used to allocate costs is: FAi × NE opex × α ij FA j = ∑ j , i ∑ NE j opex j × α ij where • FAj = the FA cost of network element j • FAi = the FA cost of area i • NEj = operating cost allocated to network element j using mark-ups in the core and access model • αij = allocation key for network element j and area i This methodology has been developed as an attempt to overcome some of the shortcomings of relying on mark-ups over equipment costs as an estimate of direct network operating costs – shortcomings that were discussed during consultations with the industry but also in the MRP. The mark-up approaches are discussed in the following section. 2.6.5 Overhead costs Overhead costs are split into common business costs, costs related specifically to access and interconnection and transit uplifts. Post- och telestyrelsen 22 In the model, common business costs are defined as: “Costs that are required by an efficient operator with SMP in Sweden, with the scope of services similar to TeliaSonera”. These costs are common to the businesses of core, access, co- location and other retail services. Note, however, that the inputs used are only those related to wholesale services (PSTN and IP network related), excluding any non-PSTN/non-IP network costs. The costs are indirect non-network costs that are required to make the business function and hence not directly related to the services or the network. Examples are the chairman’s office, recruitment costs, legal department, audit fees etc. Clearly these are required in any business, are common to the entire business and do not vary directly with service or network costs. The costs are allocated directly to the final services using a (multiplicative) mark-up approach, where the total common business costs are calculated as a fraction of the total costs of core, access and co-location. The access and interconnection specific costs have been defined as management information/support system, bad debts10, consultancy services, product development costs, charges by PTS and “other” costs. These are split by access and interconnection services and allocated to these specific services using a (additive) mark-up approach. That is, costs are calculated as a fraction of the demand (per line, minute or Gigabyte) and simply added to the final service costs. Note that this mark-up is only applied to the specific wholesale services to which they relate, not all the services modelled in the hybrid model. The transit uplifts have been included to cover costs for the required 2 Mbps interface and the cascading costs (the proportional cost paid to other operators for the leased line between TeliaSonera and the interconnection operator). 2.6.6 Expensed vs. annualised costs The model calculates the mark-up for common business costs based on the annualisation of all costs. Therefore expensed costs are allocated a share of the common business costs by calculating the common business costs mark-up as if all costs where annualised and applying this mark-up directly to the expensed unit costs. This may be regarded as a pragmatic solution to ensuring that expensed costs receive a proportion of common business costs. The underlying assumption is that the total pot of annualised costs, for those costs that are expensed, may be taken as a proxy for the proportion of annual one-off costs. 2.6.7 Number portability (IN/NP costs) The model includes the costs for number portability (IN/NP costs). These costs are in I_FA_Costs and consist of annualised CAPEX and OPEX. The cost per call is calculated and then added to all call-related costs as an additive mark-up. 10Note that the costs of bad debt, like any other cost, should be considered in a forward-looking perspective of an efficient operator. Post- och telestyrelsen 23 2.6.8 Billing of transit traffic (kaskadavräkning) PTS has included costs for maintaining functionality that enables cascade billing and actual billing of cascade transit traffic. It is calculated as an additive mark-up in the Consolidation model and added to IC single transit and IC double transit in C_Services. The mark-up consists of both capital and operating expenses. The CAPEX is the costs for development of the billing system and both external and internal resources are taken into account. The total CAPEX is then annualised with an asset life of 7 years, price trend of 0 percent and a scrap value of 0. The OPEX consists of pay costs for staff to carry out support of this billing function and a compensation factor for general overhead and management plus non-pay operating costs. Annualised CAPEX and OPEX are then added and divided by the total number of transit calls to give the final mark-up. 2.7 Other common costs Common business costs were described above. This section concerns other common costs such as building related costs and shared costs such as ducts and cables. These are the network costs of shared network equipment that are necessarily incurred if access and interconnection services are provided and are not avoided if interconnection and access services are no longer provided. These costs are not specific to the consolidation model, but are described here because they relate to all models. 2.7.1 Building costs Building space and common building-related costs are an input in the core model. Common building-related costs (or site costs) include site security, power supply units and air conditioning. Costs related to toilets, storage etc. are captured in the common costs. The model attributes each site type (RSM, RSS, LE and TS) a common site cost. This cost is then divided by the average size of these sites to obtain a per square metre value. The average site size is assumed to consist of the area used for PSTN equipment, IP equipment, non-PSTN equipment, and co-location. The average site size for co-location is an input value from the co-location model. The raw annual building space cost per square metre for each geotype is an input to the model. This value has been derived from publicly available sources. Common site costs are annualised and added to the annual building costs to achieve a cost per square metre for each geotype. These values are then converted to a per square metre value for each site type. This value is also used in the co- location model. Each cost category has a defined accommodation area of occupancy. Thus local exchanges will obtain some building costs, and transmission equipment will also occupy some space. The areas occupied by each piece of equipment are user inputs. Post- och telestyrelsen 24 The total common site costs (equipment, operational costs and any allocated building costs) are next allocated to the equipment within the sites. This is done in proportion to the area occupied. Thus the area occupied in a local exchange building by a local exchange switch determines the amount of common shared building costs that are allocated to the network element or service. Accommodation costs remain disaggregated through the individual models into the consolidation model. 2.7.2 Shared facility costs A number of shared facilities exist (excluding buildings). While it is relatively straightforward to identify the network elements used by services other than PSTN, the potential difficulty is determining the ‘usage’ of the network elements by other services. The two major categories of shared facility costs discussed in this section are: trench facilities and core network systems. 184.108.40.206 Trench facilities There are two key sharing aspects of trench: • The physical amount of trench in km shared with other utilities; and • The physical amount of trench in km shared with the access network. In addition to the physical amount of sharing there is a separate issue, namely how these costs are shared. The model allows for user-definable inputs that specify the amount of shared digging length and other inputs to calculate how the costs are apportioned. The physical amount of sharing is a technical design factor that is an output of the model calculations. The amount of cost sharing, however, is a more subjective decision, as there is no clear cost driver that can be used to allocate these costs. 220.127.116.11 Core network systems Core network systems (transmission equipment, switches, routers etc) are shared by many services. These costs are allocated to the services based on the primary cost driver being: • capacity (Megabits per second or Mega packets per second) used by the service for transmission/routing; and • call volumes for voice switching equipment. These allocations are carried out by routing factor techniques or by splitting of the network element’s costs, based on the capacity consumed by the other services. For example, as leased lines are dedicated links, there are no measurable minutes of use of network elements as there are for PSTN calls. As a result, the use of network elements by leased lines must be proxied by other measures such as leased line capacity. That means the costs of shared network elements must be allocated between leased lines and fixed PSTN and other services on the basis of the capacity of the equipment used to provide the services. Post- och telestyrelsen 25 2.7.3 Other shared facilities The switches are shared by core and access services. The model identifies the line card, MDF and frame unit as all access-related. Although central parts of a core switch, these costs are clearly not call dependent and are therefore allocated to access accordingly. 2.8 Boundary between core and access The hybrid model follows the modified scorched node assumption. This implies that the SMP operator’s existing number and location of its nodes are taken as given. However, the mix of equipment at each site (such as exchanges) may be changed. For example, a local exchange may be replaced by a remote subscriber stage. Multiplexers or similar equipment with no switching capability may also be placed at a node. A node is defined as a site with an exchange (including concentrators) and/or a DSLAM. This means that the costing boundary between the core and the access network is placed at the exchanges, concentrators and DSLAMs in TeliaSonera’s network. Table 3 Number of nodes (incl. RSMs) in the hybrid model compared to TeliaSonera’s network RSMs RSMs RSSs * Local Transit Total (Access)* (Core)* Exchanges Switches Hybrid 828 5,347 1,526 125 14 7,840 Top-down - 1,972 6,296 162 32 8,462 * Number of RSM (RSS) sites where the parent switch is not located at the same site As can be seen from the table, the number of nodes (incl. RSM sites) used in the hybrid model appears to differ from the corresponding number of node locations in TeliaSonera’s network. Both sets of figures are based on information provided by TeliaSonera. The figures used in the hybrid model are based on a thorough investigation of site and line data provided by TeliaSonera. Access RSMs are defined as RSMs with fewer than 30 lines in use and no co-located DSLAM. These have been excluded from the trenching calculations in the core network but have been modelled in the access network instead. PTS therefore uses the figure of 7,012 sites or scorched nodes (7,840-828 in the core model). The number of sites used in the core model: RSM RSS LE TS Total N. Sweden 1,416 203 13 4 1,636 S. Sweden 3,931 1,323 112 10 5,376 In the case of the 828 RSM (Access) nodes, both the line cards (note that the actual line card is at the RSS) and the de-multiplexing equipment is considered to be part of the access network. In addition, the transmission link from the RSM to Post- och telestyrelsen 26 the RSS is considered to be part of the access network. The rationale for this approach is that since the RSM is too small to have any real need to concentrate traffic, and there is no co-located DSLAM (which would concentrate broadband/Bitstream traffic), the equipment is dimensioned purely on the basis of the number and type of subscriber lines. In the case of the remaining RSM nodes, line card costs and MDF (multiplexing) costs are considered to be part of the access network, whereas de-multiplexing and transmission costs are considered to be part of the core network. The rationale for this approach is that line card costs and MDF costs would also be in the access network in TeliaSonera’s network. On the other hand, connection between the node and its parent node(s) will have concentrated traffic flowing through it (either because the RSM is large enough to, at least potentially, warrant concentration, and/or because there is a co-located DSLAM at the node). Turning now to RSS nodes, the MDF and line cards (and a part of the concentrator frame cost) are considered to be part of the access network since the quantity required of these items is driven by the number of subscribers. Other RSS costs and the transmission links between the RSS and LE are considered to be part of the core network. Although the 828 RSM (Access) nodes are considered part of the access network, these nodes are nevertheless modelled in the core model. The reason for this is that the Access model is designed to measure the costs of plant and equipment between the customer’s premises and the concentrator or RSM and is simply not suited to measure the costs of RSM equipment and the transmission links and infrastructure from this RSM equipment. By way of contrast the core model already models the equipment in core RSM nodes and the transmission links from this equipment. Therefore, the core model can easily be modified to cope with the equipment in other RSM nodes.11 2.9 Treatment of other services incl. retail Other retail services and the retail business are not calculated in the models. Where there are shared networks or facilities, the costs that are driven by these other services are calculated and due portions of network costs are excluded. Thus, for example, leased line transmission costs and common building space costs are calculated and the costs excluded as appropriate. 2.10 Allocation of costs to services For access and core services, service costs are calculated using the routing / allocation tables, using information on the attribution of cost category to network element. For co-location services the allocation is done directly without any allocation table. The attribution of cost category to network element is carried out in worksheet I_Cost_Category using user defined inputs for each cost category. Separate inputs can now be specified for when the NGN Selector is turned off and when it 11In fact, the model contains a specific section C_Transmission Section 10, which is designed for this purpose. Post- och telestyrelsen 27 is turned on. This is necessary as a few cost categories need to be attributed to IP related Network Elements when the NGN Selector is turned on whereas they are normally allocated to PSTN related Network Elements (Back Up power being a case in point). However, before costs are allocated to services, they are annualised (or expensed as appropriate). Annualised equipment costs and installation costs together with annual operating and indirect costs are added to derive a single annual cost. To allocate costs to core services the model offers five possibilities: • busy hour minutes of traffic; • minutes of traffic, • busy hour Gigabits per second (BHG), • busy hour Mega packets per second, or • a user defined allocation gradient (only used for broadband/Bitstream services) Using a “busy hour minutes of traffic” allocation, each service’s share of total minutes in busy hour is used as the cost allocation key. Using “minutes of traffic” as the allocation key, the service’s average use of a network element is divided by the total volume in minutes through the element. For “busy hour Gigabits per second” (and busy hour Mega packets per second), in the IP network, the model uses each service’s share of total Gigabits per second (or Mega packets per second) traffic in the busy hour as the allocation key. 2.10.1 Allocation for core services The busy hour allocation uses cost weights derived in the core model to dimension the network. These costs weights (CW) are calculated using the following formula: rf ij × BHTi CWij = , BH capacity j where • rfij = routing factor for service i and network element j • BHTi = busy hour traffic (BHE or BHCA or BHG) for service i • BH capacityj = busy hour capacity through network element j or ∑ rf ij × BHTi . i Note that ∑ CW i ij = 1. For voice traffic, the model uses a conversion factor to convert the annual call minutes into busy hour erlang. This conversion factor is: BHE = annual minutes/52/6/10/60. The factor of 52 reflects the number of weeks in the year while the factor of 6 is used to convert into daily values (weekend traffic is Post- och telestyrelsen 28 assumed to be the same as one weekday). The factor of 10 implies that 10% of traffic occurs in the busy hour. Finally, the factor of 60 is to convert from minutes to hours, cf. section 3.3.1 for more information on how this figure is used. In the case where the conversion factor is the same for all services the results of using a busy hour allocation will be the same as using a minute allocation – the cost weights will be the same, i.e. rf ij × BHTi rf ij × Traffic i = BH capacity j ∑ rf i ij × Traffic i For broadband and bitstream products, GigaBytes per month per subscriber are converted into busy hour Gigabits per second via the following formula: BHG = annual GB per subscriber/52/busy days per week/busy hours per day/3600*8*number of subscribers where 52 denotes the number of weeks per year and 3600*8 converts GigaBytes per hour to Gigabits per second. Busy days per week is a user input (set at 7 for private broadband/bitstream usage and 5 for business usage in the current model). Equally, busy hours per day is a user input (set at 4 for all broadband/bitstream traffic in the current model). The model also has conversions that enable Mega packets per second to be calculated from Gigabits per second, with the exception of VoIP traffic where the packets per second is an input parameter dependent on the selected Codec. 2.10.2 Allocation for access services For the access network, costs are allocated to services using the allocation table. For each network element, allocation factors are attributed to each service using the network element in question. The allocation factors reflect the relative usage of the network element (cost causation principle). These factors are weighted against the volumes of each service. Post- och telestyrelsen 29 3 Core model This section describes the main features and rationale behind the core model. 3.1 Definitions and assumptions 3.1.1 Increments The core model is a bottom up model of the equipment and systems required to carry the services defined in the MRP with the required level of service quality. It therefore calculates the cost of both wholesale (interconnect) and retail PSTN and IP Network increments. The core network model is defined to include all systems and equipment contained in scorched nodes, including links between the nodes. It does not include links from the node to the customer (these links are in the access model), except in the case of customers located on islands with no scorched node and no overland connection to the mainland (these are re-allocated to the access network in the consolidation model). The core network (c.f. core model) does not include line cards and the MDF. The core model includes additional costs required to give the extra capacity needed to support other services than those defined in the MRP. These other services, or non-PSTN increment, include IPTV, leased lines and (non- broadband/Bitstream) datacom services. The costs of servicing and supply of dedicated equipment for these increments are not included. The transmission capacity required to support the non-PSTN increment is included and a portion of the total transmission cost is later taken out of the calculation. This portion represents the cost of the non-PTSN increment. Non-PTSN increment costs are therefore not calculated in total, but the capacity effect on the PSTN/IP Network increment is taken into account. Due to the effect of cost-volume relationships, increased capacity reduces the average cost per unit. 3.1.2 Network structure – PSTN switching The PSTN switching network is based on the traditional three-layer network but uses Modern Equivalent Assets, as described in further detail below, rather than traditional switching technologies. In addition, there is a further equipment class of Telephony Servers. If the NGN Selector is set to active/on (now the default state), then all voice telephony traffic is routed via the IP Network, the DSLAMs are assumed to contain POTS line cards, the call switching is carried out via a number of Softswitches, and interconnection with other networks is achieved via Media Gateways. The three traditional layers are: • Remote subscriber switch (RSS) layer – although, as discussed below, Engine access Ramps are used instead of RSSs. The RSS is sometimes referred to as a remote concentrator. RSSs concentrate traffic that is then Post- och telestyrelsen 30 sent back to a parent Local Exchange (LE). The LE completes a call by sending it to another RSS or to another core switch. The RSS has access line cards that link to the customer. • Local exchange (LE) layer – although, as discussed below, Multi-Server Gateways (in combination with telephony server functionality) are used instead of traditional local exchanges. One LE may parent many RSSs. It is assumed that RSSs are located at scorched node sites and that an LE does not have access line cards. An LE scorched node site may have one or more RSS(s) located at the same site to enable access line cards to link to the customers in the LE node zone. • Transit switch (TS) layer – although, once again, the model uses Multi- Server Gateways (in combination with telephony server functionality) instead of traditional tandem switches. The TS links to other TSs and to LEs. A site with a TS may have an LE and RSS to access the customers located in the TS site zone. Calls are routed from a customer connected to an RSS to a parent LE. The call may then be sent to another RSS. Alternatively, the call may route to another RSS parented on another LE. To get to the other LE the call may be passed via the TS layer. It is also possible for the LE to send the call to another LE. A further element is included in the switching architecture. RSMs (Remote Subscriber Multiplexers) may substitute for a RSS (or vice versa). The RSM does not concentrate or switch traffic, but carries the traffic over 2Mbit circuits with one channel required for each customer. The RSM traffic passes through another de-multiplexer at another site to connect into the RSS switch. RSMs do not have access line cards, but the onward transmission connects into line cards at the corresponding RSS site. Post- och telestyrelsen 31 Figure 6: RSMs may replace a RSS Boundary between 1. Existing RSS Access and Core Access Core NODE site 2M LL (Fi bre) Leased line system Small building 64k 2M Customer LE RSS 64k n x 2M 64k 64k SDH SDH 64k ISDN30 64k Other Mux 32M 32M Datacoms A TM 2. New RSM instead of RSS Access Core Boundary between Access and Core NODE site 2M LL (Fibre) Leased line system Small building 64k 2M 2M LE Customer 64k n x 2M MUX 64k 64k 2M 2M 8M or STM1 64k Core Node-to-node fibre IS DN30 OLE OLE 64k Fibre 32M 32M 32M Datacoms ATM The diagram shows how an RSM may replace a RSS. The numbers of RSMs and RSS sites can be adjusted (noting that the total number of sites is fixed by the scorched node assumption). This adjustment can be made in table 1.2 of I_Demand_data. Adjusting the RSM/RSS site mix is a key feature of model optimisation. The relative numbers of RSS/RSM is defined by the model user. The numbers of each type of node is the main optimisation option for the model. The inputs for this are defined in Table 1 of sheet I_Demand_data. This covers the numbers of other site types as well as RSS/RSMs. The options selected depend on: • engineering approach toward multiplexing; • costs – the values selected should reduce the costs; • resilience – a few very large nodes may put too much traffic through a single point of potential failure. A further distinction is made between RSMs deemed to be part of the core network (RSM_Core) and RSMs deemed to be part of the access network (RSM_Access). For the purposes of this model, Access RSMs are defined as remotes with fewer than 30 lines in use and no DSLAM co-located at the site. The input totals used are based on actual site data provided by TeliaSonera, up to date as of August 2007. Switching in the model is based upon Ericsson engine type technology. The model uses EARs (Engine Access Ramps) instead of traditional concentrators, MSGs (Multi-Server Gateways) and Telephony Servers. This technology reduces overall costs compared to traditional PSTN devices: while EARs are more expensive, the cost savings made in relation to LEs and TSs are sufficient to Post- och telestyrelsen 32 reduce overall costs. The cost reductions arise since LEs and TSs are replaced with simpler – less “intelligent” – gateways for which the processing intelligence is centred in a few Telephony Servers. In the model cost inputs, account has been taken of the fact that there is less processing required in the MSG equipment. In order to come up with realistic cost inputs, the modellers discussed with Ericsson the price information which had previously been provided by operators. The final cost inputs reflect Ericsson’s assessment of the information provided by operators and its own views on prices for the engine solution. Cost data has been entered that covers the costs of the MSG equipment and the estimated accommodation space required. The MSG engine solution is based on the structure of “basic” EAR switches that have the line cards connecting to customers. These parent onto LEs (i.e. MSGs) that have the call switching engine. The EAR has the functionality to switch calls, and, since routing factors are based on information provided by TeliaSonera, is assumed to do so in the model. An EAR can be located remotely or at the same site as the LE MSG. Those customers which are connected at an LE (MSG) site are assumed to be connected to an EAR rather than directly to the LE (MSG). The LEs (MSGs) also connect to parent transit switch MSGs. The MSG call- control is carried out by telephony servers. Relatively few telephony servers are required for a network (although some resilience is needed as these control many switches). The telephony servers supply signalling information to the MSGs. The required number of EAR switches depends on user inputs of the RSS/RSM numbers (based on actual TeliaSonera site data). The number of EAR units at a site depends on the customer numbers (there is a limit to the number of customers per RSS unit). LE and TS numbers are user inputs – these can be defined in a manner that reduces costs, but takes into account the engineering philosophy taken for the network. Again, inputs for numbers of LE and TS nodes are based on actual TeliaSonera site information. Note, it is assumed that the trunk interface cards in all switches are the same cost no matter the switch type. International gateways are also included in the model. The cost of these gateways reflects the costs of large MSGs. 3.1.3 Allocation Issues – PSTN Switching A number of allocation decisions are required in relation to the switching network. Firstly, since RSMs do not actually concentrate traffic, an issue arises as to whether these should be classified as part of the access or the core network. The RSS and RSM sites are now considered as a single set of sites with the site being deemed an RSS where there are more than 500 lines in use. For smaller sites, a distinction is made between Access RSMs and Core RSMs. Where an RSM is defined as an Access RSM (<30 lines in use, no co-located DSLAM), the RSM itself and the transmission between that RSM and the LE is considered to be part of the access network. On the other hand if the RSM is defined as a Core RSM (>30 lines in use or a co-located DSLAM) or RSS in TeliaSonera’s network, it is assumed to be part of the core network in the model. Post- och telestyrelsen 33 Secondly, since EARs (or RSSs) in the modelled network act as switches (whereas the original EAR technology could not switch internally the current version is able to do this), the frame cost for these EARs is assumed to be predominantly driven by the core network. However, some EAR costs are allocated to access reflecting, in particular, the fact that the capacity of each EAR is constrained by the number of subscribers making subscriber numbers as well as traffic volumes a cost driver. 3.1.4 Networks structure – Transmission The overall core network is designed to link all the scorched node sites. Transmission technology for PSTN and non-PSTN products is based on SDH. Link sizes for optical links vary from STM1 up to STM 64. Broadband and bitstream services are modelled to run over a pure IP/Ethernet network and are assumed not to use the SDH transmission network. If the NGN Selector is set to active/on (now the default state), then all voice traffic will be routed directly over the IP Network and the SDH transmission network is removed from the calculations. Microwave systems may be used up to STM 1 size. Most of the network is designed using a ring structure to provide resilience. Some sites may also be linked on spurs. The typical connection of sites is shown in the diagram below. Figure 7: Ring and spur structure Fibre pairs use RSS Logical connection RSS Logical connection same physical path Fibre pair R SS S RS RS S Fibre pair RSS RSS RSS RSS S RS R SS RSS LE LE RSS (big ci ty) RSS (big city) Co-located Co-located RSS and LE RSS and LE Ring structure used to connect most sites Sites connected via a spur link Rings structures are also used to connect LEs to other LEs and to connect LEs to TSs and TSs to other TSs. In addition, a back-up point to point network has been developed for TS to TS links to provide an additional level of resilience. The ring provides resilience – if the capacity factors are set correctly, there is at least double the total capacity of the ring on each link (or “hop” between nodes), Post- och telestyrelsen 34 and the total traffic may be routed round the ring in the opposite direction. The model allows for 30% additional resilience between remote subscribers stages and local exchanges; 50% between local and tandem exchanges one of which is in the relevant LE-TS, the other in another ring12) and 100% resilience at the inter- tandem switch level. In this case the ring network is presumed to be dimensioned to carry the required capacity with the point to point network also being dimensioned to carry this capacity. Some sites may be connected via a spur (as shown in the diagram above right), where there is no physical diversity. This is typically appropriate for some islands and some sites that are in remote areas or along valleys. Spurs provide lower level of resilience, but this is still satisfactory for smaller sites as the probability of link failure is low and the total number of customers affected is low. Spur links may be provided on cable or microwave connections. Any spur links to islands for the access network are calculated similarly. See Table 6 of sheet: C-Transmission. Note that transmission capacity for delivering a call to the nearest POP has not been included in the model. These costs are considered to be marginal and without influence on the final results. 3.1.5 Network structure – Core Trenching The model assesses the requirement for core trenching separately from the SDH ring topology analysis. A study was conducted of the core trench routes required to interconnect the various core sites within TeliaSonera’s network in Sweden. The study was based around a dataset of the node locations provided by TeliaSonera (TS_Sitemaster) and used Microsoft Autoroute. For the transit-transit and local-transit routes a complete set of routes necessary to interconnect those locations was compiled (with physical trench “rings” incorporated for resilience). The diagram below illustrates the results of the Autoroute-based analysis for inter- transit (blue lines) and additional local-transit (red lines). 12 In order to reach the other tandem, which is used as a backup in the model the traffic will need to use the inter-tandem network. This has been reflected in the dimensioning of inter-tandem rings. Post- och telestyrelsen 35 Figure 8: Transit and Local level trench km analysis For the remote (concentrators and multiplexers) to local routes, due to the much larger dataset, a sample was taken with the aim of covering at least 20 percent of the available nodes. The actual percentages covered by the sampling are shown below: Figure 9: Remote to Local sampling Direct LIC in Total LIC in use use Total sites LX RSS/RSM-A RSM-B Qty Qty Qty Qty Qty Qty Totals 888,351 2,010,275 2,163 64 1,786 313 Percent Sampled 57% 41% 28% 46% 26% 38% The conversion of trench km to trench “types” (such as “Trench H; asphalt / tarmac”) uses allocation matrices. These took as their starting point the overall Post- och telestyrelsen 36 allocation matrix used in version 4.1 of the Hybrid model, with various adjustments made (which are commented on in the worksheet). The model allows for trench sharing between the core and access networks and also between TeliaSonera and utilities. This issue is discussed in further detail in Section 3.9. 3.1.6 Network structure – IP Routing The IP Routing network is based on a four-layer network: • DSLAMs. The DSLAMs represent the layer at which the customer typically connects (via a DSLAM line card). • Aggregation switches and Metro Switches. These are assumed to be managed Ethernet switches, used to group together a number of DSLAMs and to form high capacity rings in dense (typically business) urban areas. • Edge Routers. These represent the first level of IP routing within the network and would typically be installed in rings, with a number of Aggregation/Metro switches feeding into two (for dual parenting reasons) Edge Routers on a ring. • Core Routers. These represent the second level of IP routing within the network and again would typically be installed in a ring (or more likely two ring) configuration with a number of Edge Routers feeding into two (for dual parenting reasons) Core Routers. Although there is no specific requirement for an equivalence between the differing levels of the IP Network and the PSTN network, the DSLAMs could be thought of as roughly corresponding to the RSS/RSM layer, the Edge Routers to the LE layer and the Core Routers to the TS layer. Equipment costs for xDSL and IP routing equipment, including two sizes of DSLAMS, Aggregation Switches, Edge and Core Routers, are specified in a separate input sheet, I_IP_Routing_Costs. Repeaters have also been included within this sheet and are used where the inter-node distances are in excess of 80km. The assessed number of repeaters is inputted within worksheet I_Trenching on a route-by-route basis. Additional equipment has also been included here (specifically, Edge and Core Media Gateways and Softswitches) as these are required when the NGN Selector is active (such that all traffic is assumed to flow over the IP Network). Assumptions regarding equipment costs, asset lifetimes, price trends and maintenance/installation mark-ups for IP network equipment have been based on typical values for the industry as of 2007. When the NGN Selector is switched on, now its default state, additional DSLAMs are provisioned to cater for the need to originate and terminate voice calls over the IP network. These are modelled distinctly within the Core model, accepting that in practice an operator might well choose to mix voice line cards and xdsl line cards on the same DSLAM. Post- och telestyrelsen 37 3.1.7 Allocation issues – IP Routing Allocation of the IP Network across the various broadband and Bitstream services has been carried out in a similar manner to that of the PSTN services via an extension of the existing routing tables. This allows specifically for a cost-based distinction between Bitstream products interconnecting at differing levels of the network hierarchy. One important allocation issue exists with both the broadband and Bitstream products. This is concerned with the relative allocation of costs between services identical in all but access speed (for example, 500Kbps versus 2Mbps or 8Mbps). With PSTN-based services, the allocation is made in accordance with usage (typically, minutes of traffic), with both the retail and wholesale services generally tariffed on this basis. However, the industry norm in Sweden is currently for broadband and Bitstream services to be tariffed on the basis of (a) access speed and (b) whether or not the product is for private or business use. The assumption is presumably that customers with a higher access speed will use the service more than those with a lower access speed, and similarly business customers will use the service more than private customers or require a more strict Service Level Agreement (SLA). However, with regards to usage, there is no guarantee that this will be the case, and even if it is the case, there is unlikely to be a stable, definable relationship between access speed and usage. The model provides for two alternative methods to allocate costs across broadband/Bitstream services. The first of these allocates on the basis of the assumed average usage per month (measured in Gigabytes of data downloaded) based on information provided by TeliaSonera. With the usage patterns typical of 2007, two impacts of this are that the cost of the products for private use are higher than those for business use (as the usage is higher), and that the cost of the highest speed products are in excess of the current retail prices. The second method allocates the costs according to a user defined allocation gradient. This could, for example, be set in accordance with the retail price gradient, or adopt some other relationship between access speed and cost/price. The gradient is currently set on the basis of a logarithm formula: Gradient = ln (connection speed / minimum speed) where the minimum speed is set to 64 Kbps (equivalent to a normal POTS line). 3.1.8 Network structure – North and South The differences in topographical features in Sweden mean that the country has been divided into two regions – North and South. North Sweden is defined for the purposes of the model as above the 6,750 latitude, as shown in Figure 10, where the grid reference for all switching zones in Sweden have been plotted. There is one transit switch structure for the North and a second for the South. This reflects: • The larger area in the North has low total traffic and low total numbers of subscribers. The North has many zones, each with few customers. This is Post- och telestyrelsen 38 because each zone has a finite size limited by the maximum physical distance from the node to the customer. The distance limit is set by copper technology. • The South has more traffic and a higher density of population and subscriber lines. The dual treatment enables the structures of transit switches in the North and South to be adjusted independently, but based upon the same design principles. The rings and point to point links in the PSTN transmission network have been dimensioned to take account of the traffic flowing from North to South and vice versa. In addition, the rings in the South of Sweden have been dimensioned to take Figure 10 Boundary between North and South account of the traffic flowing between the of Sweden rings. In order to calculate point to point 7.700,0 capacity assumptions each tandem exchange node was given a number, with 7.500,0 numbers being assigned as follows: • 1-5 to the southern ring in Southern 7.300,0 Sweden; 7.100,0 • 5-10 to the northern ring in Southern Sweden (node 5 is on both rings); 6.900,0 • 11-14 to the ring in Northern Sweden. 6.700,0 Following this, for each pair of nodes a route was determined. For adjacent nodes, such as 1-2, the route is simply 1-2; for 6.500,0 non-adjacent nodes the route involves a number of stages, e.g. 1-3 is 1-2 and then 6.300,0 2-3. In addition, it was assumed that intervening multiplexing was used 6.100,0 1.000,0 1.200,0 1.400,0 1.600,0 1.800,0 2.000,0 wherever possible. The number of times a particular path is used (such as 1-2, 2-3, 3-4) was then calculated and using average traffic flow between nodes (a figure which varies between Southern and Northern Sweden) the total capacity requirement was calculated for each traffic path, including the links between Southern and Northern Sweden. 3.1.9 Design timeframe (horizon) The core model has a design timeframe for different cost elements. Any cost category can be allocated with a time horizon. Thus some elements can be dimensioned for the year 1 demand, others for years 3 or 5 demand. This means that costs need not reflect only today’s volumes, but those of the future, weighted by the need to provide for future equipment (some equipment can be provided just in time, others must be planned for further in advance). Non-PSTN volumes can be adjusted separately to allow for future growth and these inputs are in Section 2 of I_Demand_data. Post- och telestyrelsen 39 The model has a macro that in effect runs the model 3 times to gain the different data. This macro is located as a button on the Output sheet and should be run once the model has completed all other calculations. The user must enter the time horizon to use for all different equipment categories (1, 3, or 5) – see the growth year factor in the sheet: Output. The macro should be run after the core model has been altered and/or the setting of the NGN Selector or other important parameters has been changed– only if the macro is run will the correct results be linked to the consolidation model. 3.2 Structure of core model The core model structure is shown in the diagram below. Figure 11: Core model structure The diagram shows the model structure taken from the model map. The model map is a diagram in the Excel model that allows the user to navigate through the model. There are eight main stages in the calculation: • Input volumes, technical data and routing table data • Trenching data • Transmission dimensioning • Switch dimensioning • IP routing dimensioning • Cost determination Post- och telestyrelsen 40 • Sharing with other services and access network • Export to consolidation. In addition to these stages there are some offline calculations that have been used to process benchmark data for use in the model. These calculations are not part of the Excel model. The starting point is the total traffic for all the services (PSTN, broadband/ Bitstream and remaining non-PSTN). This is used to calculate the size of the network needed – both the size of transmission links and the switching and IP routing systems. The network is dimensioned to meet the demands of the average annual input traffic and the peak traffic demand. These traffic volumes are based upon the latest available data from TeliaSonera and adjusted to account of future growth so that the network represented in the model reflects the actual considerations of network planning for an SMP operator. In addition, for PSTN traffic, the network has been dimensioned to allow for unpaid holding time (between ringing and answering a call), based on information provided by TeliaSonera and for differences in busy hours over the day of the week and the month of the year. The peak traffic is spread across a number of different node sizes and links. The model assumes the traffic is profiled or spread across a variety of small, medium and large sites according to the number of lines for PSTN traffic and accounting for the profile of non-PSTN traffic across different parts of the network, as detailed in the rest of this section. Separate usage profiles are applied to broadband and bitstream traffic, routed over the IP network where again the peak traffic requirement has been assessed based on the annual usage. The average traffic per site (or link) is determined by the number of such sites. The profiling data and the number of sites and traffic calculations are taken from the tables of input volumes and technical data. The model then uses the sizes of each element to determine the cost. The cost- volume relationships are defined by cost tables and an assumed linearity. Input costs for all items are specified. The input cost values for different sizes of equipment are based upon expected corresponding volumes. The actual volume adjusts the variable part of the cost. The result is the actual cost of the small, medium or large etc. element. The total cost is the cost of the element times the total numbers of each element. This generic process is used for PSTN switching, IP Routing, and transmission. The operational costs are determined and combined with the equipment costs. The operational costs are defined by an assessment of the expected cost for the different functional areas of the SMP operator and the cost of employment. The allocation of operational costs is based on an initial estimate of the annual operational cost (as a percentage of the purchase price). The resultant initial values are used as a “weighting” to allocate the costs determined by the functional area analysis (cf. section 2.6). Post- och telestyrelsen 41 Accommodation costs are calculated separately based on the cost of rented space per geotype and for the mix of geotypes expected in the nodes of the core network. 3.3 Technical and volume inputs In this section a description is provided of the technical and volume inputs in more detail. 3.3.1 PSTN and broadband/Bitstream volumes PSTN The input tables in I_Product_List define the total traffic volumes in the base year for each product. These inputs are extended in I_Product_Data to specify volumes for: • call minutes; • calls made; • growth percentages for future years; and • holding times and unsuccessful call attempts. Each of these values can be specified for each of the products. The model converts the minutes data into Busy Hour Erlang (BHE) values. This conversion is defined as a simple formula (the BHE value is related to the total annual minutes and how the product is weighted towards peak demand). The conversion should be defined as the fraction of the annual call minute traffic that occurs in the network busy hour. The network busy hour is the peak time of day averaged across all the network. A rule of thumb conversion states that BHE = annual minutes/52/10/6/60. This assumes that every week is similar, the peak busy hour in a weekday has one tenth the daily traffic, weekend traffic is in total the same as one weekday and (of course) 60 minutes per hour. Other formulae may be user-defined to reflect actual measures on traffic profiles. This is an offline calculation that may be calculated based on data from Swedish operators, but cannot be conducted without detailed operator data. The above formula has been multiplied by 1.2 to take account of estimated variations between traffic flows by day of week and month of year. The holding times and unsuccessful call attempts cause an effective inflation on the call volumes compared to the measured call volumes that are input as the base volume driver. TeliaSonera has provided information on the extent of this inflation and this has been incorporated into the traffic dimensioning process. Note that the call volumes should be the measured call lengths (measured on a per second basis) and should not be rounded up as is common in some billing systems and with some switch systems that use pulse period type measurements. Note also that the volumes include ISDN2 and ISDN 30 calls – they are not treated any differently in the model from PSTN calls. Partial ISDN 30 lines have not been modelled. Post- och telestyrelsen 42 Additional inputs are provided in section 7.12 of worksheet I_Design_rules to allow for the calculation of average throughput (in kbps and packets per second and prioritisation adjustment) for when the NGN Selector is switched on and voice traffic is assumed to flow over the IP network. Broadband/Bitstream For IP network traffic, the initial inputs in I_Product_List specify the numbers of subscribers for each service (broadband, bitstream) as well as average usage in terms of GigaBytes per month per subscriber. A usage profile is then applied to calculate BusyHour Gigabits per second requirements for each product. The usage profile takes account of the concentration of traffic into a certain number of “busy days” per week and also into a certain number of “busy hours” per day. The busy days per week have been assumed to be 7 (weekdays and weekends) for private usage and 5 (Monday-Friday) for business usage. The number of busy hours per day has been set at 4 for all products. These parameters are all user inputs and can be amended if required in worksheet I_Product_Data. In addition, there is a conversion parameter (based on average packet size) in section 7.13 of worksheet I_Design_rules so that the traffic can be measured in Mega packets per second as and when required. Bitstream backhaul products are also included within the model. These products aggregate the traffic to/from a number of DSLAM locations and present it to the OAO as a single data feed. The products are distinguished by the bandwidth of the aggregated feed, measured in Mbps on the presumption that the OAO purchases the right to use the stated bandwidth to its maximum 24 hours per day, 7 days per week. On this basis, the products are treated within the model in a similar way to a non-contended leased line – that is, the “sold” capacity is reserved within the network for the customer and thus the cost of that bandwidth is allocated accordingly. The average packet size for these products is assumed to be the same as for the underlying Bitstream/Broadband products. IPTV Now that there are a considerable number of IPTV subscribers, the network usage related to IPTV traffic has been explicitly modelled – based on an assumption that the traffic flows through the network using multicast techniques. The use of multicast means that, for each TV channel available: • a single IP flow is needed from the TV Head End equipment to each of the Core Routers (thus the total number of flows is equal to the total number of Core Routers) • a single IP flow is needed from the relevant Core Router to each of the Edge Routers (thus the total number of flows is equal to the total number of Edge Routers) • a single IP flow is needed from the relevant Edge Router to each of the DSLAMs supporting IPTV (thus the total number of flows is equal to the total number of DSLAMs supporting IPTV) The inputs relating to IPTV are contained in section 2.11 of worksheet I_Demand_data in the Core model. These allow the user to specify which Post- och telestyrelsen 43 scorched node sites are assumed to provide an IPTV service and also the number of channels (and bit rate per channel) provided. As with broadband, there is a conversion parameter (based on average packet size) in section 7.14 of worksheet I_Design_rules so that the traffic can be measured in Mega packets per second as and when required. 3.3.2 Additional network volumes Additional volume data is assumed based on inspection of map data and information provided by TeliaSonera to account for parameters such as: • number of sites accessed via submarine cables; • number of sites access via microwave; • number of sites on spurs; and • total number of sites by site type. The total number of sites is profiled to be spread over RSM, RSS, LE or TS. Note that a LE site may contain an RSS, but is termed an LE site – the naming convention is that the site is called after the highest level switching system on the site. Similarly, a TS site is assumed to cover both LE and RSS functionality. 3.3.3 Non-PSTN volumes The network transmission is dimensioned to cope with extra capacity needed for remaining non-PSTN services. The total volumes are inputs – this is in Mbits/s – for each of four “urbanisation” categories, and can be found in section 2 of worksheet I_Demand_Data. These are the same four categories used in the core trench analysis detailed in I_Trenching. Thus category A is for Stockholm, B for other large cities, C for the remainder of North Sweden, and D for the remainder of South Sweden. The inputs were based on an analysis of non-PSTN product data provided by TeliaSonera, along with assumptions on the likely distribution of these products across the four urbanisation categories (for example, it was felt very unlikely that high speed private circuits would occur in the smaller towns and villages but would instead by concentrated in the larger cities). An additional routing table allocates the total volume inputs across the three major hierarchical levels: remote to local, local to transit (including local to local), and transit to transit. As with broadband, there is a conversion parameter (based on average packet size) in section 7.15 of worksheet I_Design_rules so that the traffic can be measured in Mega packets per second as and when required. 3.3.4 How the volumes are used The predicted volumes are uplifted by the future years’ demand. There is a weighting of future demand so that the network can be selected to be dimensioned for one of the future years depending on the weighting factor. This weighting factor is included at the top of the Product data input sheet. Post- och telestyrelsen 44 This means the base year data is used with the growth factors to dimension a network larger than needed for today’s traffic (assuming growth is positive). However, since negative margins for growth cannot be ruled out for certain services, there is a possibility that the demand per network element including growth is lower than the demand excluding growth. In order to take account of this potential problem the model does not allow the demand per network element including growth to be lower than the demand excluding growth. The volumes are used with the routing table to define the total demand through each network element. The routing table is described in more detail below. In summary it defines how each product uses the various network elements (RSS, LE, TS, LE-TS transmission etc.). The volumes – busy hour Erlangs (BHE) and busy hour call attempts (BHCA) for PSTN and busy hour gigabits per second (BHG), or busy hour Mega packets per second (BHMP), for IP traffic – inflated for growth and the routing table derive the total traffic through each network element. The traffic is then split: north Sweden – south Sweden. This is a user-defined split and is based on the location of sites. The demarcation of North and South Sweden is conducted on the basis of the number of lines per site as a first indicator. The split of network elements (LE, RSS etc) is a user defined input for each region (north and south). The total number of elements is defined by the scorched node assumption. The average traffic per element is the total traffic divided by number of elements. The model, however, takes a more realistic approach and profiles the traffic per network element – it is spread to different sized elements, with the average traffic per element being the same. The result is that we obtain the traffic for a number of different sized elements. The driver for some elements is BHE or BHCA, for others BHG. The values of BHE, BHCA, BHG or BHMP are used to drive transmission capacity and PSTN switch equipment and DSLAM/IP Network equipment component sizes. The volumes are also used to calculate the final product costs. The volumes are exported to the consolidation model. In the consolidation model, the volumes (minutes, call attempts, Gbps or Mega packets per second) are used to drive the cost of the network elements into the products. Also, in consolidation the cost per minute, call, Gigabyte, Gbps, or Mega packets per second of each network element can be calculated (if desired). 3.4 Cost inputs The costs of equipment must be defined for each item. Cost data includes: • Capital cost. The purchase price in the base year. • Installation cost. Additional one-off costs that are needed to set up and install the equipment. This might be additional payments to the equipment vendor to install the system. Post- och telestyrelsen 45 • Lifetime. The average lifetime of the equipment before it is replaced. • Price trend. This is the price trend in nominal terms (if the price is 5% less than last year and this is expected to continue over the equipment lifetime then the trend is -5%). • Operational costs. This is the estimated annual operation cost for the item as a percentage of the capital value. Note that the actual operational costs are defined in separate calculations. • Scrap value. The value of the equipment when it is scrapped as a percentage of today’s purchase value. Assume the equipment is scrapped today (in the base year, not in x years’ time). Additionally some other factors are needed for equipment such as wastage factors. This defines the percentage of the equipment that is consumed when it is installed and not available for use (typically some lengths of cables are wasted at each end). Modularity is also defined. This is the number of connections or services on the equipment item. The cost of one item must be defined, but the item might have 10 customers connected, and an alternative vendor might have a different cost but 12 customers connected to it. The modularity is very important to enable the cost per unit to be calculated. Cost inputs are critical to the results. The values must be checked with data sources to ensure they are valid. The model does not check that the input values are within limits. Cost inputs are in SEK. There are three separate cost input sheets for Switching, IP Routing and Transmission equipment. 3.5 Network Design Rules The network is dimensioned using design rules. These dimension the network using engineering/economic factors. The basis for the engineering design rules are mostly self explanatory in the formulae of the model or are explained in the text comments to the model. General points about the technical design are: • Three layer switch network assumed in the PSTN with telephony servers providing some of the functionality currently provided within local exchanges and tandem switches. • For the IP network, a four level hierarchy of equipment is assumed: DSLAMs, Aggregation Ethernet switches, Edge Routers and Core Routers. The Edge Routers are assumed to be roughly where the existing local switches are, with the Core Routers located where the modelled 14 transit switches are located. Repeaters are assumed to be needed for long routes (over 80km). The associated design assumptions are included in section 7 of worksheet I_Design_Rules in the Core model and the cost assumptions in worksheet I_IP_Routing_Costs. Post- och telestyrelsen 46 • Connections between sites are predominantly based on physical/logical rings. • SDH transmission systems are the main transmission method in the PSTN. • Systems (rings and switches etc) are profiled so that there are variations in size (with the correct average size). • Systems sizes are rounded up to the next available equipment size. Since equipment is only available with standard sizes, it is assumed that the next size up is always used. • Practical engineering rules are incorporated. Thus for test and prudent engineering safety factors, some headroom in capacity is assumed. Systems are not designed to be used “at 99.9% of the physical limit.” They are used up to the limit defined by the technical design rules. Erlang calculations are a key calculation since the traffic demand in erlangs is used to dimension many network elements for voice traffic. The erlang formula is approximated by a formula based on Gaussian statistics. This formula could be changed by the user to an alternative erlang calculation (though it might slow down the model calculations). The accuracy of the Gaussian approximation is sufficient for the model. Only at very low traffic levels are errors significant – and these can be reduced by adjustment of the erlangs_conversion_factor. In a carrier network, traffic levels tend to be large. When round-ups to the nearest 2Mbit/s interface levels are considered, the Gaussian approximation is shown to be even more valid (the errors for low [<20 erlang] traffic levels do not matter). The accuracy is shown in the graph below. Figure 12: Comparison of model calculation of capacity needed and exact erlang formula. 1200 60 1000 #circuits needed (model) 50 #circuits needed (Erlang) Number of voice circuits required Number 2Mbit links required #2Mbit links (model) 800 #2Mbit links (Erlang) 40 600 30 400 20 200 10 0 0 10 100 1000 Busy hour erlangs Post- och telestyrelsen 47 3.6 Switching Conventional circuit-switched equipment is assumed for the PSTN, but incorporates the latest technology available. The Multi-Service Gateway solution differs from the most traditional systems in that some of the intelligence is removed from the switches themselves. The core processing power of the switches is moved into a new network element, the Telephony Server. Signalling is sent up to the Telephony Server where call routing is determined and instructions sent back down to the relevant switches. Switches are assumed to have a core switch block that is traffic (BHE) dependent. Some parts of the switch are fixed (invariant with demand). Others are assumed to be BHCA driven (such as the variable processor element). It is assumed that the costs of processing power at the switch level are significantly reduced compared to the more traditional switching solutions, since much of this processing power is now located in the Telephony Server. The fixed processor element is split between BHE and BHCA. Additional accuracy is ensured by assuming a variety of switch sizes, each with different cost assumptions. The model, therefore, does not assume “one size fits all.” A switch has trunk cards (2Mbit/s) that link it to other switches – the number of which is driven by the BHE traffic to other sites. Only EAR units are able to have line cards to customers. An LE site may have co- located EARs. Thus LEs (MSGs in the model) do not have direct customer line card interfaces (it should be noted that this assumption depends on the switch vendor; however, LEs that do have customer line cards as well as hosting RSSs have very similar overall functions to a LE host plus RSSs located in the same site). An LE (MSG in the model) has access interfaces (2Mbit/s) to link to the RSS units that the LE parents. The LE also has an additional cost per TS by which it is parented. TS sites will typically have LE and RSS co-located at the same site. A TS will have a trunk line card for every 2Mbits of traffic from an LE that it parents. International gateways are assumed to be functionally similar to transit switches. 3.7 IP Network equipment The IP Network comprises three distinct types of equipment: DSLAMs, managed Ethernet switches and IP routers. The DSLAMs are assumed to be current generation ones utilising Ethernet rather than ATM-based backhaul. Two different sizes have been included, one capable of around 250 customer connections, the other, larger one, capable of around 750 customer connections. Where more customer connections are required, the model simply assumes, as is current industry practice, that multiple DSLAMs will be located together at the same site. No cost distinction has been made for the capacity of the backhaul connection from each DSLAM as (a) the default Post- och telestyrelsen 48 backhaul is likely to comprise 1 Gbps ports and (b) the cost impact of higher speed (10 Gbps) ports was not felt to be likely to be significant, if indeed necessary, given current usage patterns. The DSLAMs are grouped together via managed, Layer 2 Ethernet “aggregation” switches, with the average number of DSLAMs per managed switch being a design input (based on the number of scorched node sites per Ethernet switch). The switches in turn are connected to the Edge, Layer 3 IP routers, which essentially comprise the first level in the IP network hierarchy where per- connection traffic routing is feasible (permanent “routing” is possible at the managed Ethernet switch layer via the use of VLANs, but such routing is more akin to leased line routing than traffic routing). The aggregation switches could be connected to the Edge Routers in a “point to point” configuration or in a ring configuration (the latter utilising the spanning tree algorithm), though the cost impact of either design choice is unlikely to be significant and as such has not been modelled distinctly. The Edge Routers are assumed to be configured as rings to allow for routing resilience and also to be connected to two Core Routers (larger, higher specification routers) at the core of the network. Likewise, the Core Routers are also assumed to be configured in a ring configuration (or more likely as at least two rings) for resilience reasons. The design assumes that the IP Network does not require additional transmission equipment (such as an SDH network). Whilst it is acknowledged that some operators do overlay the IP network over an (existing) SDH network, it was felt that this was largely done for legacy reasons since there is no overriding technical reason for it to occur. The design thus assumes that the IP network equipment uses its own fibres in the trench rather than, with the NGN Selector switched off, sharing capacity on the fibres used to convey PSTN (and remaining non-PSTN) traffic. Given that there will be IP equipment at each of the nodes where there is PSTN/SDH equipment, it has been assumed that, when the NGN Selector is turned off, the demand for fibres will be roughly equal for the PSTN/SDH network as for the IP network at all levels in the hierarchy. However, as this is a design input, it can easily be modified to incorporate other fibre usage patterns. Since there is significant scope for variation in the precise configuration, supplier and hence cost of the equipment at the managed layer 2 Ethernet switch and layer 3 IP router levels, the model does not attempt to prescribe precise configurations or manufacturers. The reasoning for this is strengthened by the significant variation in supplier discounts that can be available (discounts of up to 75 per cent of list price are not unheard of for such equipment, though 40 to 50 per cent is perhaps more likely for large scale operators). What has been assumed as inputs to the model therefore are typical costs seen across a variety of operators across Europe for equipment at the respective levels in the network hierarchy. The impact on the outputs of fairly large variations to the costs of the equipment are unlikely to be that significant given the low-cost nature of IP equipment compared to more traditional PSTN/SDH equipment (for example, the switching/routing equipment currently comprises only around 3.5% of the total cost for the 2Mbps shared access Bitstream level 2 product). Post- och telestyrelsen 49 3.8 Transmission and infrastructure Transmission in the PSTN is based on SDH fibre technology. STM 1 - STM 64 systems may be used and multiple systems may be deployed to give the required capacity. Some microwave systems can also be used. The capacity of the SDH systems depends on the total traffic in a ring or on a spur (total of traffic from all nodes). This total traffic excludes IP network traffic but includes PSTN and remaining non-PSTN traffic (though only when the NGN Selector is turned off). This capacity allowance enables all the traffic to be routed in any direction round a ring – giving a fault tolerance. SDH system costs are not based on just the required capacity, but on the total capacity to account for prudent engineering – systems are not expected to be close to maximum capacity. The total traffic before another system is required is a user-defined input. Fibre infrastructure is assumed – the cables can be dimensioned in different ways. With high speed systems, it is unlikely that there would be a need for cables with very large numbers of fibres (such as 96 pair), since even a 20 pair cable with STM 64 on each pair has a very significant capacity. The user can select the mix of deployed fibre cable for each route type in table 3.3 of I_Design_Rules. Additional meshing between nodes is allowed – this enables additional inter-site capacity to supplement the rings, giving further resilience to cable breaks. This might typically be introduced in the core inter-transit links. Cross connects are placed at the intersection of LE and TS rings and at approximately 20% of the LE sites. The physical SDH network is based on the logical demands. These logical links and their capacity are set by the PSTN and non-PSTN demands. RSSs are logically linked to a parent LE. LEs link to two TSs. LEs are also inter-linked to some degree to reduce the need to send calls to TS sites. TS sites are logically fully meshed (there should be no need for a call that uses 3 TS sites). However, LE-LE links are provided within the same rings as LE-TS transmission. The use of the LE-LE transmission (within LE-TS rings), TS-TS transmission and LE-TS transmission etc. are defined by the routing factors. As was described in an earlier section, the previous method of assessing the required amount of trenching in the core network is based on a detailed assessment of the required street km to interconnect a dataset of the node locations provided by TeliaSonera. For the transit-transit and local-transit routes a complete set of routes necessary to interconnect those locations (with physical trench “rings” incorporated for resilience) was compiled. For the remote (concentrators and multiplexers) to local routes, a sample was produced with the aim of covering at least 20 percent of the available nodes (27.5 percent have actually been covered). The conversion from street/trench km to trench “types” (such as “Trench H; asphalt / tarmac”) uses allocation matrices. These took as their starting point the overall allocation matrix used in version 4.1 of the Hybrid model, with various adjustments made (these are commented on in the worksheet). The calculations for the core trenching can be found in a separate worksheet, I_Trenching. Post- och telestyrelsen 50 3.9 Routing factors Routing factors (in the routing table) define how each product uses the network. The model has the capacity for sub-routes to be defined for each product. The model reflects TeliaSonera’s estimates of the probabilities of particular types of routes, since any other approach, aside from one based on detailed analysis of traffic flows within TeliaSonera’s network, is subject to too large a margin of error. However, it can be noted that since the network structure being modelled is not the same as TeliaSonera’s own network structure, the routing factors in the modelled network will differ to some degree from those in TeliaSonera’s network. This factor has been taken into account in relation to the remote concentrator to local exchange routing factor (higher than in TeliaSonera’s network) and in relation to the routing factors for international calls – in most cases there is a link between the tandem switch and the international switch (classified as an inter- tandem link in the model). The likely impact on other routing factors was examined but it was concluded that the available data was too limited to estimate this impact with any degree of accuracy. The model includes a feature that provides a basic automatic adjustment of route probabilities, although this feature is not invoked in the current version of the hybrid model. The routes and probabilities are combined to define an average route probability. This average route probability is used in the model. The route table is used with product volume data to define the network element sizes (capacity) and hence costs. It can be noted that a call that uses two RSSs and one LE may use LE-RSS transmission, but in some cases such a call might not use the RSS-LE transmission – in the case where one customer is connected to an RSS that is co- located at the LE site. This is an important issue which has been considered in defining the routing table, since it ensures that the capacities of the transmission links are correctly dimensioned. The issue is also relevant when it is recalled that LEs may be located at TS sites, hence not all calls use the inter-switch transmission elements (they only use tie cables within the building – these are not dimensioned in the model). The amount of traffic that uses LE-TS transmission is a percentage of the total traffic that connects LE switches to TS switches. These percentage factors are defined in the model (the formulae can be over-written with values). Similar percentages adjustments are determined for RSS to LE transmission. As regards the Metro service, the routing transmission routing factor has been adjusted to take account of the fact that RSS-LE routes tend to be shorter in Metropolitan areas. The routing table has been expanded not only to allow for the inclusion of broadband/Bitstream products carried over the IP Network related equipment but also (when the NGN Selector is active/on) to allow for existing PSTN traffic to be routed over the IP Network. The average route table is exported to consolidation to enable the products to be costed from the network element costs. Post- och telestyrelsen 51 3.10 Shared costs in the core model The core model has two shared costs. The main items are: • Core access sharing; and • Other utility sharing. The first item arises because most streets require some access network (they have some customers to service). Therefore an efficient network might share this duct and trench with the core network. Only in out of town areas would the duct be used purely for the core network transmission. The user may specify the amount of sharing in percentage terms in table 2.5 in I_Design_Rules. The cost split is subjective as there is no direct cost driver (the duct and trench is common since it would be required for either core or access cables). A 50-50 split has some logic, but the decision is user-defined and is essentially a pricing decision. Sharing of digging (trench) costs with other utilities (electricity, cable TV other operators etc.) is usually possible (and in some areas it may be encouraged or enforced by the local municipality). The amount of sharing can also be defined by the user. The cost sharing is a user input and the same comments as core-access splitting also apply. The initial assumption used in the model assumes that the likelihood of sharing with utilities is greater on RSS links, which may be nearer to the end-user, than on trunk links between, say, TS sites. Other costs in the core model that are not finally allocated to core include the line card and MDF costs. Where possible, input data has been sourced from Sweden. However, in some cases, the line card costs have been supplied per line, rather than per line card. Thus, in some instances, the modularity of a line card may appear as 1, when in fact it is known that a line card has a modularity of, say, 16. The input cost values need, therefore, to be managed very carefully. 3.11 Core model calculations In the following, a more detailed description of some of the primary calculations used in the model is provided, to supplement the more generic description above. Please note that the calculations are all open in the model – there are no hidden calculations. Some descriptions/guides are included within the model itself. The model is therefore the ultimate source for all calculations. 3.11.1 Volume calculations Call minutes, numbers of calls made and Gigabytes per subscriber are converted to values for use in network dimensioning. Busy hour erlangs, busy hour Gigabits per second (BHG), and busy hour Mega packets per second (BHMP) are derived by simple numerical relationships to annual minutes, Gigabytes per subscriber, and Gigabits per second respectively. The relationship depends on the time of day/day of week profile. For PSTN products, uplifts are applied where appropriate. This adds in additional volume demands for unsuccessful calls. Post- och telestyrelsen 52 The result is shown as BHE, BHCA, BHG, or BHMP values, depending on the type of product. 3.11.2 Routing table calculations The routing table for PSTN products is based on information provided by TeliaSonera because it was considered that no other approach could provide sufficiently robust information. However, the user is able to enter different inputs if required. For retail broadband products, the routing currently assumes that all traffic flows from the DSLAM through to the top level of the IP Network hierarchy, and for Bitstream that it leaves the network at the appropriate level (directly at the DSLAM for level 1 Bitstream and at the Edge Router for level 2 Bitstream). Again, the user is able to enter different inputs if required. The average network usage is defined – number of network elements times the probability. These values are summed up to give a weighted average usage of the network by each product. In sheet C_Route_table, the average usage is combined with the BHE, BHCA, BHG, or BHMP values to obtain two tables – one that defines the total usage of each element in BHCA, BHE, BHG or BHMP, which is used to dimension the network (table 1). The other (table 2) defines the cost driver – this shows the relative amount of sold products using each network element. 3.11.3 Adjustment of traffic to RSMs The total traffic from RSSs is mapped to RSM sites, based on the number of sites converted to RSMs (see sheet: I_Demand_data). Proportional rules are applied (see table 2) based on numbers of nodes. This defines the total traffic through the elements. 3.11.4 Traffic profiling The numbers of lines are profiled to sites of different sizes, as is the PSTN traffic (table 3 of I_Demand_data). The profiling does not alter the total traffic, but spreads it across sites so that the model does not calculate costs based on “one size fits all.” The profiling method need not be absolutely accurate since the effect of slight inaccuracies is a second order effect – the prime cost driver is the total traffic, and this is derived from the inputs of PSTN volumes, broadband/ Bitstream volumes, and non-PTSN volumes. 3.11.5 Building calculations Building values, site values, lifetimes and cost of capital are used to define the average cost per square meter (using an annualisation method). However, the model uses as default an entered value (SEK/m2 per annum) based on market data. The main site cost items (power, air conditioning etc) are annualised and added to the raw (building) cost. Average building costs are defined based on a distribution of sites across geotypes. Post- och telestyrelsen 53 Site space requirements is a critical input as the average cost of the main site items is divided by the average site areas. This cost (per unit area) is added to the basic floor space costs to give a total cost per m2. This cost of space is multiplied by the space demands (area) of each type of equipment (as derived in the equipment calculations sheets). The area is defined by the area needed per item of equipment times the total number of equipment items. 3.11.6 Switching calculations The voice traffic demand in BHE or BHCA for each switch network is profiled initially to the different sized switches (table 1 of C_switching). The total traffic is then divided by the number of nodes to get the traffic per node or per port interface (i.e. the link to another node). The traffic per node defines the number of 2Mbit interface cards needed (table 1.8). This is calculated using a Gaussian formula which is a simple approximation of the classic erlang formula, as detailed in section 3.5 of this document. The total number of port cards is derived by multiplying the cards per node with the number of nodes. Table 2 simply links in the source equipment costs. Table 3 multiplies the cost (table 2) times the volumes (derived in table 1) to get total costs of switch elements. The elements of the switches are split between core and access. This allows the model to allocate some common costs of switches to access – the percentage is a user-defined input (table 8 of I_switching_costs). For many parts of a switch, the cost will be considered to be part of core (call dependent). The model assumes that only line cards are part of access, in part because information from TeliaSonera suggests that such elements as keyset receivers and ringing generators which can also be considered as part of access are very small in cost terms. It can also be argued that it is quite reasonable to consider some switching costs as common and hence allocate a percentage to access. The allocation of a switch cost to access (other than line cards and elements such as keyset receivers) is worthy of discussion. There are often some common costs of a switch – items needed both for the core switch and for the hypothetical case where there are only access line cards. If we accept common costs exist, then there is a mark-up or allocation issue. The costs could be allocated to core or access. The choice is essentially a pricing decision. The model has a % allocation. This is user defined, but it could be derived using analysis of the costs allocated directly, so that an equal mark-up value can be defined. The method to do this for RSSs would be: • set the allocation 100% to core and set the RSS switch costs to zero; • measure the costs of the core RSS element. This defines the RSS incremental core costs, excluding switch; • set the RSS access lines cards to zero and measure the change in access costs. This is the incremental cost of access; Post- och telestyrelsen 54 • compare the incremental core and access costs and define the allocation % to be the same ratio; and • return all the cost values to normal. This has not been carried out – the % allocation entered is a nominal value. The costs of the switch elements are entered into I_Switching_Costs. Note that there is a choice of the tables to fill in. Total costs for switches can be defined and these broken down by percentage factors to the sub-elements of the switch (table 2) or the costs of the sub elements can be directly entered (table 3.3). Table 3.3 is the most normal location for cost data. The allocation of RSS (EAR) frame units to core or access has a significant impact due to the large number of RSSs in the network. The model assumes that almost all the costs are allocated to core. This is the assumption used by TeliaSonera and is broadly consistent with the fact that RSSs are used for switching within the model. The total costs of each part of the switch are collated at the end. The total costs are divided by the unit cost of the same element to get the total numbers of each element used. This is used to get the building space required for the element. This is defined as the number of units times the area used per unit. The area used per unit is defined as an input. The values for these calculations are derived using an index look-up technique (a method commonly used throughout the model). The area used is multiplied by the cost per area (derived earlier in the building calculations). See tables 6 and 7 of C_switching. The costs of core and access from switch calculations are collated in table 8 for export to the output stage. 3.11.7 Transmission The inter-site transmission in BHE is derived from the I_Demand_data worksheet that profiles the total traffic across different site sizes. The LE to RSS site traffic is split into the LE to RSM site traffic and the LE to RSS site traffic (see table 1 of C_Transmission). Note that the BHE values used for LE-RSMs are not used – the size of the transmission from RSM sites to LEs is based on the number of lines, not the BHE values (this is done in table 1.3). The number of lines to RSM sites defines the number of Mbit required (each PTSN line requires 64kbit/s or 2Mbit/s if the line is a primary rate ISDN line). The BHE traffic on the inter-switch site links, or the number of Mbit if to an RSM, is divided by the number of switch node sites. The traffic per link (BHE) is converted to 2Mbit circuits using a Gaussian based formula. The number of 2Mbit circuits per link is multiplied by total sites to get the total number of 2Mbit circuits required. The non-PSTN traffic is added to the PSTN traffic. The total is increased to allow for logical diversity (additional links to other nodes) see Tables 1.6 to 1.9. The logical diversity (as used, say, for parenting two LEs for every RSS) is also applied to non-PTSN services. Although this could be deemed over-engineering, it has Post- och telestyrelsen 55 been included to ensure all services have a high quality of service. It is possible to modify the formulae so that only PSTN has additional logical diversity. Note that the diversity factors have a significant influence on costs. Since broadband/Bitstream traffic is now being modelled explicitly, it was necessary to include a calculation for the remaining non-PSTN traffic at the different levels of the hierarchy. This has been based on an assessment of non- PSTN data provided to us by TeliaSonera, coupled with a number of assumptions about network usage and network routing (specifically including an assessment of the use of the trench network by high capacity circuits in the more rural areas of the country). These calculations have been included in an expansion of section 2 of worksheet I_Demand_Data in the Core model. The resultant percentages are fed into section 17.1 of worksheet C_Transmission in the Core model. The total traffic is next split between north and south Sweden (based on a user defined split %). This is required because the traffic levels and network design in the north are significantly different to the south – the north has large areas, many sites but low traffic levels (see Table 2). Further, in the case of tandem to tandem switch links, account is also taken of traffic between North and South Sweden. The node numbers in each region are divided by the number of rings in a node to get the number of rings required in the network. The model has profiles of rings – not all rings are the same size. The ring sizes (nodes per ring) and % of rings that are small, medium etc are user inputs. The model assumes the primary transmission technique is to use SDH rings to give resilience. These technical assumptions can be adjusted in table 4 of I_Design_Rules. The next part of the model builds up the links from RSMs and RSSs that are needed to form SDH rings. This is done in a sequence of stages in table 3. Table 4 of C_Transmission calculates the links between LEs that are required to form the LE rings. Much of the LE ring distance can be shared with the RSS ring structures since this will account for some saving of trench and ducting costs. It is assumed, for engineering prudence, that the RSS and LE ring structures do not share the same fibre cables, although in some cases this may be desirable, for example in more remote areas. Table 5 shows the TS-TS calculations. Many of the details underlying these calculations are shown in Section 7 of I_Design_Rules. As noted above, in addition to the basic ring structure there is a back-up point to point structure. Both the ring and the back-up point to point structure are designed to carry all required capacity in the network. Table 6 calculates the requirements for special links such as spurs and island links. The numbers of these are derived from user inputs (table 1.2 in I_Design_Rules). The size of the links is based on the average link size (note this is worst cased – it is assumed the largest of the north or south Sweden average is taken). Note that some islands without nodes are included in these calculations. These costs are then passed to the access model to account for the phone lines not located on the mainland or on an island with its own node. Post- och telestyrelsen 56 The total traffic in the link is the traffic per node times the average number of nodes on the link (a user-input) – this ensures the total spur capacity is able to cope with all nodes on the spur. Table 7 calculates the equipment numbers required using the previous table calculations. 3.11.8 Core Trenching The model assesses the requirement for core trenching separately from the SDH ring topology analysis. A study was conducted of the core trench routes required to interconnect the various core sites within TeliaSonera’s network in Sweden. The study was based around a dataset of the node locations provided by TeliaSonera (TS_Sitemaster) and used Microsoft Autoroute. For the transit-transit and local-transit routes a complete set of routes necessary to interconnect those locations was compiled (with physical trench “rings” incorporated for resilience). For the remote (concentrators and multiplexers) to local routes, due to the much larger dataset, a sample was taken with the aim of covering at least 20 percent of the available nodes. The results of the analysis have been included in a new worksheet: I_Trenching. This worksheet consists of four main sections. The first three cover the three main hierarchical levels of the trench network: inter-transit routes (that also include nearby local switch locations); additional local-transit routes; and remote- local routes. The final section brings together the data from the first three sections and arranges it into a format suitable for insertion into the C_Transmission worksheet. For the first two sections of I_Trenching, each individual route is identified separately, and includes an assessment of the need for Repeaters for the IP network (where the links are longer than 80km). For the third section (remote-local), each local exchange area is identified separately, and for those which were sampled the results of the Autoroute analysis are included. The analysis includes an assessment of the amount of trench km that would be shared with the two higher levels of the trench hierarchy. This section then scales the results of the sampling to account for the remainder of the local exchanges. The scaling is carried out on the basis of grouping the local exchanges into one of four groups: Stockholm, Gothenburg and Malmö; other large cities; remainder of North Sweden; remainder of South Sweden. Should additional local exchange areas be sampled at a later date then these can be easily inserted, with the scaling adjusting itself accordingly. The conversion of trench km to trench “types” (such as “Trench H; asphalt / tarmac”) uses allocation matrices and also adjusts for trench sharing with the TS- TS and TS-LS routes. The matrices took as their starting point the overall allocation matrix used in version 4.1 of the Hybrid model, with various adjustments made (which are commented in the worksheet). The results from Section 4 of worksheet I_Trenching are taken forward to drive the tables in section 11 (section 17 in version 4.1) of worksheet C_Transmission. This has allowed a number of previous sections of that worksheet to be removed. Post- och telestyrelsen 57 4 Access model This section describes the main features and rationale behind the access model. The access model calculates the resources required to build that part of the network that is below the existing RSS/LE Switch site, as far as the Network Termination Point (NTP) at the customer premises. The main purpose of the access model is to calculate the cost of copper access defined as access to the copper line stretching from the MDF to the customer premises. The model does not calculate access network costs at the switch site – as the following diagram shows: Figure 13: The scope boundaries of the access model LE LE Demux RSS MDF site Line RSM Line RSM cards cards Access FAM Access Network Network PDP PDP Model Model PDP PDP SDP SDP SDP SDP SDP SDP Customer Site NTP Fibre Copper The main components of the access model are: • Trench, Duct, and Poles used by the access network; • Copper cables and distribution points; • Final drop miniduct (duct to subscriber building from street), associated digging works, and Network Termination Points (NTPs); • Fibre in the access network; and • Fixed Wireless Access network (FWA). After calculating the resources needed to build this part of the network, it uses the results to calculate the capital cost of those resources, and the annual operating cost of such a network. Post- och telestyrelsen 58 4.1 Definitions and assumptions 4.1.1 Geotypes The main definition used in the access model is based on the segmentation of Sweden into a subset of similar areas or zones called geotypes. All switch zones in Sweden are allocated to one of five geotypes, illustrated below: Figure 14: Illustrative maps of the different geotypes served by the access network • City: more than 1,000 lines per km2: • Urban: from 100 to 1,000 lines per km2: • Rural A: from 10 to 100 lines per km2: • Rural B: from 1 to 10 lines per km2 • Sparse: up to 1 line per km2. Uninhabited areas of the country are not in any switch zone and do not have any access network costs associated with them. This definition of geotypes enables the model to represent the diversity of areas in Sweden, whilst avoiding the need for detailed analysis and estimation for every one of the switch zones. The optimal design of the network, and the mix of costs incurred, will be significantly different in each of these types of area. 4.1.2 Sampling The main purpose of the access model is to calculate the cost of copper access defined as access to the copper line stretching from the MDF in the SMP Post- och telestyrelsen 59 operator’s network to the customer premises. The model therefore takes the existing MDF locations in TeliaSonera’s network as given, even though some of these locations in principle are not scorched nodes, as the MDF may be co- located with a simple RSM without any concentrating functionality. Sweden has around 7,860 MDF zones (defined as the geographical areas connected to a given MDF site). In order to design and cost the network, the access model requires detailed estimates of the geography for each zone analysed. Clearly it is not practical to analyse every zone in the country. The model therefore does most of its analysis on a sample of 25 zones. The model allows some flexibility in the construction of this sample. It is possible to sample more heavily from those zones where a higher proportion of the access network costs are incurred. This needs to balance a number of considerations, such as: • those with the most customers, typically in geotype 1 and 2; Figure 15 Sampling from the • those with the most km of trench and duct (typically in geotype full set of zones 3 and 4); • those with the most pair-km (typically in geotypes 1-3); and • those with the greatest diversity in teledensity (typically geotypes 1 and 5). A different weight could be attached to each zone selected. For example: • If the geotype 1 has 100 zones, and the sample contains four zones from geotype 1, each sampled zone could be given a weight of 2513; or • If there are 3,000 zones in geotype 4, and six zones in the sample from geotype 4, then each could be given a weight of 500. An alternative way of weighting the samples could be on the basis of the area covered by the sample zones in each geotype compared with the overall area covered by all zones of that geotype. This clearly requires information on the area covered by each zone, something that was not available for previous versions of the model but was provided by TeliaSonera for this revision. This approach of using differential weightings allows an efficient sampling frame to be used, giving much more representative results than using a simple random sample of 25 zones. Differential weightings are used within geotype, e.g. to target finer sampling on the zones in the geotype that have higher teledensity (and therefore have more of the demand). 13Individual weights can be assigned to each zone. e.g. if ten of the 100 zones are very different from the other 90, then one of them can be selected and given a weight of 10, and three from the other 90 can be selected and given weights of 30 each. The model gives the best results if time is taken to make the data represent Sweden, e.g. by selecting zones of high/medium/low teledensity within each geotype. Post- och telestyrelsen 60 For bottom-up model v4.9, a new set of 25 sample areas was selected and an analysis carried out off-line to determine the street km within the new sample areas, using Microsoft Autoroute. The results of this analysis were fed into the access model in two areas, I_GIS_Routes, and I_Zones_Raw_Data, and are used for all subsequent versions of the model. Worksheet I_GIS_Routes has been adapted to take advantage of the new samples to provide a more detailed calculation of the necessary trenching in the access network. The adaptation utilises a multi-stage process: • An input of the assessment of street km for each of the 25 sample areas, including weightings to produce a national equivalent figure. Note that the definition of “inside Tätort” was based on the indications provided by Autoroute of conurbations, and includes, for example, “built up” areas of smaller towns and villages. • An input of the assessment of the necessary trench km taking account of the need for trenching on one or both sides of the street and the expectation that a proportion of the “streets” visible in Autoroute in the rural areas will not require trenching. • An input of the assessment of the type of trenching by sample and road type, with separate inputs for inside and outside the Tätort. This was done in two stages: initially, selecting between “high”, “medium” and “low” cost trenching, and then for each of those three categories selecting between the various trench types (including tunnelling and poles and direct buried cable) The various inputs within I_Zones_Raw_Data have also been updated to reflect the new set of 25 samples. 4.1.3 Main assumptions The main assumptions in the access model are as follows: • The access model does not deal with RSMs or RSSs. They are calculated in the core model. Multiplexing below the scorched node (Fibre access Multiplexers, or FAMs) is part of the access network: FAMs are not “sites” to be considered under the scorched node criteria. The models only allow for the use of FAMs to connect islands to the network. • The physical boundary between the two models is at the scorched node. For those components in the access network that are located at the scorched node, e.g. MDF and line cards, the costs are calculated in the core model and passed to the consolidation model to be added to access services when calculating the final access service costs. • The model allows for flexing of the number, size, and location of PDPs and SDPs. • In general customers are connected to the nearest RSS/RSM site, other than in a small proportion of cases where there are local geographical reasons (such as e.g. highways or lakes) to do otherwise to avoid unnecessary costs, i.e. much as at present. Post- och telestyrelsen 61 • Some sharing of trench and/or duct is possible with the core network and other increments and utilities. • There is no sharing of copper cable with the core network. • Microwave solutions (point to point) and FWA solutions (one hub to many customers) can be used for some islands and for sparsely populated areas where this is cost efficient (the model does now incorporate a limited amount of FWA, but only in geotypes 4 and 5). 4.2 Structure of access model The model includes a navigation map, which shows all of the worksheets and all of the links between them. This is reproduced in the diagram below. Figure 16: Access model – navigation map The arrows show the flow of information between the groups of calculations. Solid lines are used for main flows, dotted lines for minor flows. Grey areas show worksheets that are grouped – having related functions. Flows of information within each group are not shown in detail; they all go down the list. 4.3 Modelling the access network The access model has nine groups of calculation sheets: • calibration; • LRIC costs of each resource; • Tätort; • copper cables and nodes; Post- och telestyrelsen 62 • fibre; • trench and duct, including route sharing with core and with other utilities, also miniduct and associated digging for the final drop; • FWA; • final network design – in particular, the decision of where to deploy FWA; and • output to consolidation model. The approach taken to each of these is outlined below. 4.3.1 Calibration As a further check to ensure that the model is using data that is representative of Sweden, all of the input data for the sample zones is calibrated to agree with data for the whole of Sweden. Calibration is done for the following: • number of zones; • total area in km2; • number of lines (also by type of connection – POTS, ISDN, xDSL, 4 wire copper, and fibre); • number of customer sites; • house-to-road distance; and • total national street km. The model allows the user to choose between two ways of doing each of these calibrations: • for each geotype on its own; and • for the country as a whole, i.e. pooling the data across all geotypes. This part of the model also takes account of the year for which the network is to be dimensioned. Normally the network would be dimensioned to fit to the data for the base year. However, the model allows for the network to be planned for the volumes for a different year – up to three years ahead – using the growth estimates for number and mix of lines. Two points should be noted: • The model already allows for assumptions about prudent deliberate overprovision to cater for future growth, moves and changes, and bypassing of faulty lines. The idea is that these figures are sufficient to provide for normal growth. There is no need to choose, say, volumes for year three to ensure that the network has sufficient resources to meet normal needs over the forecast period. • If growth is large, then it could be argued that there would be an increase in street trench and also in miniduct, although each is affected differently: Post- och telestyrelsen 63 Table 4 Different types of growth and the impact on network equipment Type of growth: New town/suburb, new Infill building Existing sites, Network equipment site (“green field”) additional lines More cables SDP-PDP- Yes Larger cables, Larger cables, same RSS same cable km cable km More cables, final drop Yes Yes Yes (or larger cables) More Street trench and Yes No No duct More miniduct Yes Yes No The approach taken in the calibration of the model is to scale up the miniduct / final drop digging to reflect the growth, but to leave the street trench and duct unchanged, i.e. to treat miniduct correctly in the first two cases, to treat street trench correctly in the latter two cases, and avoid having any large bias. 4.3.2 LRIC costs of each resource In order to optimise the design of the network, the model needs to know the true LRIC cost of each of the resources that it is able to choose to deploy – e.g. different sizes of copper cable, and the various types of FWA equipment. This is calculated for use inside the access model for the purposes of designing and optimising the network. Final LRIC costs are calculated in the consolidation model. The cost of capital used here should be set to the same value as that used in the consolidation model14. 4.3.3 Tätort In most of the zones from the new 25 samples there is a main town or village (Tätort), usually near the centre of the zone, and usually close to (or including) the location of the switch. Typically this Tätort is very different from the surrounding area: • The Tätort may typically contain 50%-90% of the demand for lines but may cover just 1%-20% of the area of the zone: the teledensity inside the Tätort is often 10x to 100x that of the rest of the zone. • The mix of digging surfaces in the Tätort may be different from the surrounding area. The model therefore does much of the analysis taking account of this difference, by calculating results for the area inside the Tätort, then for the area outside, then aggregating of the results. 14It is possible to get the access model to optimise using a different cost of capital, and measure the effect of errors in this key parameter – to see its impact on the network design and on the final cost of access. It is likely that small to medium errors in the cost of capital, say, less than three percentage points, will have minimal impact on the design of the network. Post- och telestyrelsen 64 If the official Swedish definition of Tätort areas is followed (broadly, areas with at least 200 inhabitants and no more than 200m between buildings), then published data on each Tätort can be used as a starting point for preparing the data for the model. In this case, the distinction between areas inside the town and areas outside is of greatest importance in geotypes 3 and 2. It has less importance in geotype 1 (where normally the entire zone is inside a Tätort), geotype 4 (where few zones have villages of 200+ inhabitants) and in geotype 5 (which possibly have no villages of 200+ inhabitants). In version 4.9 and subsequent versions of the model, the definition of “inside Tätort” is based not on the published definition but on the indications provided by Microsoft Autoroute of conurbations, and includes, for example, “built up” areas of smaller towns and villages. This approach was consistent with the method used to calculate the length of street km in the 25 sample areas. 4.3.4 Copper cables and nodes The calculations for the copper network aim to make extensive use of the economies of scale from deploying large cable sizes. They aim to economise by using large sizes for the bulk of the pair-km, only using the smaller sizes for shorter lengths closer to the customer. The model allows the network in each switch zone to have the following: • Primary Distribution Points (PDPs) to allow breakdown from large cable sizes of (typically) 20-100 pairs. • Secondary Distribution Points (SDPs) to allow breakdown of cables to dropwire sizes, e.g. 2 pair. The model allows this second layer of distribution points to be omitted for some zones e.g. where there is a cost advantage. • Exit from street duct (EFSD) – the joint between the miniduct serving one customer site and the street trench and duct shared by all sites. No change of cable size at the EFSD – the drop wire runs from SDP to NTP. • Distinctions between the number of customer sites in the zone and the number of NTPs – typically there will be 1-3 NTPs per customer site, occasionally more (e.g. in blocks of apartments). These are shown in the following diagram: Table 5 Different parts of access network Typical distances and volumes for an “average” switch zone: NTP NTP RSS / LE ~ 2km ~0.5km ~0.1km Final Drop MDF PDP SDP EFSD EFSD EFSD EFSD ~50 ~250 ~10 NTP NTP 1 ~1,000 Post- och telestyrelsen 65 The model also allows a percentage of the zone to be served by poles and drop wires. Each of these areas is described in turn in the following subsections. 4.3.5 Copper cabling: (i) PDPs The model allows the user freedom to vary the number of PDPs in the zone. Increasing the number of PDPs usually has the following effects: • increases the total km of cable; • increases the costs of distribution points; and • decreases the total pair km. There is usually an ideal, where the overall cost is minimised. The user should experiment to find a good or even optimal design in each zone. This can be done in Table 5.1 and 5.2 in sheet I_Zones_Raw_Data in the access model. For zones that have no Tätort, the main decision is the number / size of PDPs to choose. For those zones that do have a Tätort, there are separate inputs to control the number of PDPs inside the Tätort and the number outside, as shown in the example below. Figure 17: Example with and without Tätort. Zone with Tätort Zone with no Tätort A: Switch is located inside Tatort D: Zone with no Tatort Switch zone Switch zone PDP PDP Tatort PDP PDP PDP PDP PDP Switch PDP Switch site site PDP PDP PDP PDP PDP PDP PDP PDP PDP PDP Example zone with four PDPs outside the Tätort and Example zone with no Tätort and seven PDPs seven PDPs inside If the switch is located outside the Tätort, the model allows a choice of two approaches: • a “super-PDP” to serve the entire Tätort; or • no Super PDP, cables run from each PDP in the Tätort to the Scorched node. Post- och telestyrelsen 66 These are shown in the diagrams below. Figure 18: Options for PDP distribution With a Super-PDP No Super-PDP B: Switch is outside Tatort, and network is C: Switch is outside Tatort, and zone network designed with a Super-PDP is designed without a Super-PDP Switch zone Switch zone PDP PDP PDP PDP Tatort Tatort PDP PDP PDP PDP Switch Switch Super- PDP site PDP site PDP PDP PDP PDP PDP PDP PDP PDP PDP PDP PDP PDP PDP The model also allows the user to vary the approach taken to locating PDPs and SDPs within each zone: • close to the centre of the PDP’s catchment area, to minimize the length of the shorter cables (PDP-SDP-NTP); or • close to the scorched node, to minimizes the length of the larger cables (PDP-RSS). The parameters for these inputs can be adjusted / varied over a wide range. 4.3.6 Copper cabling: (ii) SDPs and street level network, from SDP to NTP The choice of the number of SDPs in each zone and the strategy for their location follows the approach described above for PDPs. The only difference is that the PDP plays the role of the parent node. The model allows the user to omit a layer of the network e.g. by setting the number of SDPs equal to the number of PDPs. The diagram below shows the street level layout of the network at and below the SDP. Post- och telestyrelsen 67 Figure 19: Access network – street level network at/ below SDP Diagram of SDP zone: PDP SDP SDP SDP SDP Key: Street trench Copper cable ~ 10 pairs or more Copper cable ~ 2 pairs or fewer Customer building (house, SDP NTP SDP Miniduct trench SDP zone boundary Street-level analysis takes account of differences in resource requirements for miniduct, associated digging, and cabling. The model takes account of the ratios of: • customer sites / customer lines, • EFSDs / customer sites, and how these vary between zones, in determining the resources required. 4.3.7 Copper cabling: (iii) use of cables on poles The model allows for the use of poles and dropwire, especially for the connection from SDP to NTP. Typically this solution has the following effect: • reduction in street digging cost; • elimination of cost of miniduct and associated trench; • increase in cable cost per meter (cable can be more expensive); and • for final drop, shorter cable lengths where dropwire poles are used. The diagram below shows this. The use of dropwire poles should be estimated in such a way that it reflects the mix of digging and terrain types – there is some scope for optimising it, but it is often not economic or realistic to use it throughout a zone (e.g. if the dropwire from pole to customer building is too long then additional poles are required). Post- och telestyrelsen 68 Figure 20: Options for final delivery to customer Diagram showing layout of the street and customer site Customer Key: site Street trench Miniduct trench route of wire inside building (vertical component is not Building shown) NTPs Route of miniduct Route of dropwire Road and verge Road and verge Route of cable if buried SDP This choice of poles or buried cables is often constrained by local rules such as building restrictions and network design considerations. For this reason, the model does not optimise the choice between dropwire on poles versus buried cable. Instead, the model asks the user for some broadly reasonable estimates of the mix of cables buried versus cables on poles, and the model calculates the costs of the network based on that assumed split. 4.3.8 Copper cabling: (iv) summary of design parameters to be optimised The following is a list of the network design parameters that the user is at liberty to vary. After any major15 change to the input data that relates to the copper network (i.e.: cost of copper cables, cost of copper distribution points, or sample zones data on number of lines, area, centralisation, size of Tätort), these parameters should be adjusted for each zone to find the lowest cost solution possible. This re-optimisation ensures that the design of the network makes best use of the resources available, given the circumstances, i.e. the physical layout of the zone and the number and distribution of the customers. Failure to redo this optimisation could lead to the escalation of access costs, and if the underlying data has changed significantly then the overstatement of costs could be significant. Parameters to be optimised for each zone are as follows (1-6 only apply to zones that have a Tätort): 1. decision to use/not to use a Super PDP, and location of Super-PDP if used; 2. capacity of each PDP inside Tätort; 3. location strategy for PDPs inside Tätort; 15Suggested rule of thumb is that a “major” change is one where any data on unit costs, areas, or number of lines has changed by more than 20%. A re-optimisation is always required before a major new real-data release of the model, or before a major analysis of results. Post- och telestyrelsen 69 4. capacity of each SDP inside Tätort (and whether or not a layer of SDPs is to be deployed); 5. location strategy for SDPs inside Tätort; 6. capacity of each PDP outside Tätort; 7. location strategy for PDPs outside Tätort; 8. capacity of each SDP outside Tätort (and whether or not a layer of SDPs is to be deployed); 9. location strategy for SDPs outside Tätort. 4.3.9 Fibre The network to serve fibre customers has a very different set of drivers from that for copper: • the network design may be a mixture of rings and spurs; • capacity is not pro rata to the number of pairs; and • the savings from deploying smaller cable sizes are less dramatic. The diagram below shows the resources required for fibre, and how it is altered if the mix of rings and spurs is changed. Figure 21: Fibre cables to customers Diagram to show fibre cabling routes Customer site Showing what is required for: - tree and branch structure Customer Customer site - ring structure building Customer building D Scorched C node A B Road Customer site Key Fibres for tree and branch structure Additional fibre routes for ring structure NTP Customer building The distribution of fibre customers has a large impact on the cost of connecting them. The diagram below shows two cases: • a “naïve view,” with all customers (first divided into those inside Tätort / those outside Tätort ) spaced evenly; and Post- och telestyrelsen 70 • a more “typical case,” with customers in clusters and generally nearer the centre. The lines show a plausible cabling network for each. In the latter case, cabling is greatly affected by the distribution of customers. Figure 22: Distribution of customers Naïve view: customers are spread evenly More typical case: customers are in clusters and tend to be nearer to the centre Naïve view: fibre customers are distributed evenly Typical distribution of fibre customers Switch zone Switch zone Customer site Customer site Switch Switch Tatort Tatort The model focuses on getting an assessment of the degree of clustering and centralisation of fibre customers in the zone, and using this to estimate the cable and digging resources required for the network or fibre required to serve fibre customers. 4.3.10 Trench, Duct, Miniduct, and route sharing For access, the network of trench/digging will generally follow the road network. There is then a “final drop” digging (or dropwire from a pole) between the customer’s building and the trench that runs alongside the road. The model uses a set of conversion factors to estimate the km of street trench based on the mix of roads: • some road segments require trench on both sides; • much of the road network requires trench on one side; and • some road segments need no trench for access (although some may require trench for core). These factors have been estimated from the maps for the sample of 25 zones. To ensure a consistent definition of the boundary between “street trench” and “final drop digging”, the data used in the model is supported by more detailed breakdown of road kms in each sample zone, linked back to the different categories of roads shown on the maps. This mitigates the risk of double counting or of omissions. Post- och telestyrelsen 71 After the total km of trench has been estimated, it is broken down by digging surface. Adjustments are also made for any routes shared with: • core; • other service increments; and • other utilities. For the final drop, a separate set of inputs are taken. The total amount of final drop route is the number of customer sites (not lines; a site may have several lines, e.g. a block of flats) multiplied by the average length of the final drop. Although each final drop is quite short, there are several millions of them, so the costs incurred can be significant. As with the street trench, the final drop will be a mix of digging and poles and may be shared with other utilities and other increments in the SMP network (although not core). The model allows these inputs to be specified. A first estimate of the distance between house and road has been made on the basis of maps. PTS has then reconciled the results of this map analysis with results provided by Cartesius for TeliaSonera. Cartesius has estimated the house-road distance on the basis of house co-ordinates provided by TeliaSonera and road co- ordinates available to Cartesius. 4.3.11 Fixed Wireless access A large proportion of Sweden is very sparsely populated. In these areas, Fixed Wireless might well be less expensive than cable and digging. The model therefore evaluates the cost of deploying Fixed Wireless technology on a zone by zone basis. Figure 23: Sparsely populated zone – illustrative distribution of customers Switch zone Tatort Customer site Switch For some zones it may be attractive to deploy FWA outside the Tätort – the (main) town or village – and use cable and digging inside the Tätort. The model evaluates this approach too. Post- och telestyrelsen 72 4.3.12 Final Network Design – in particular, the decision of where to deploy FWA The model compares three options for the design of the network in each zone: Option 1: All fixed, no FWA Option 2: FWA for all Copper NTPs, fixed connections for any fibre NTPs Option 3: FWA for Copper NTPs outside Tätort: Fixed connections for Copper NTPs inside Tätort, and for all fibre NTPs. The model determines which of these is the most economically attractive, using the full LRIC cost of each solution rather than just the capital cost. The model adopts the most economically advantageous design for each zone (lowest LRIC cost solution that meets the design constraints) except where the user deliberately prevents it from doing so. The user is free to limit the choices to any combination of the three options, and can control this for each individual zone in the sample of 25, so for example it is possible to limit the deployment of FWA to zones of particular geotypes. In areas where FWA is deployed, it reduces the amount of trench that can be shared between core and access. The model allows the user to choose between two alternative approaches: • minimise cost to access; and • minimise combined cost to core and access. The latter approach is recommended. The final amount of shared trench, after allowing for cost minimisation due to the deployment of FWA, may be considerably less than the amount offered for sharing by the core network. For example, even if FWA only serves 3% of customers, it may cover 30% of the country, and hence may eliminate sharing of route for some 20% of the core network routes. This reduction needs to be reflected in the input data used in the core model. The recommended approach is as follows: 1. Run core model to determine the total km2 or route offered for sharing. 2. Run access to determine the impact of FWA. 3. Re-run core, with the revised figure for shared route km after the effect of FWA. Step 1 is only needed when the number of core sites, the average distance between them or the proportion offered for sharing is changed. These changes should be comparatively rare, so the amount of repetition is small. 4.3.13 Output to consolidation model The most important set of outputs is the breakdown of the resources used, and the associated costs, by up to 160 cost categories. The model provides the following data for each cost category: • Volume of resources used, with units. Post- och telestyrelsen 73 • Suggested allocation to Network Element (e.g. “copper cable 2 pairs” is likely to be allocated to “copper cable, SDP to NTP”). • Capex cost of equipment/materials. • Capex cost of installation. • Price trend, asset life, scrap value of equipment (and, separately, for installation capex). • First estimate of the annual operating cost associated with the cost category (these figures are calibrated in consolidation, to ensure that they sum to the total opex figure). The model also provides a summary breakdown of these costs by some 20 major cost categories. If required, the model can output the results for just one geotype. This allows comparison of the cost and resources used (and the resources per line) for different types of zone. Other outputs are as follows: • list of network elements; • list of access network services; • values for the access Routing Table, based on the cost drivers for each network element and how they map onto network services as shown below: Figure 24: Access Network Routing Table Network Network Elements Services Trench Copper FWA MDF etc PSTN Access Routing Table: ISDN2 Weightings to use to allocate costs from Network Elements to Network Raw Cu Services Fibre etc • validation outputs, to enable the consolidation model to check that the access model and the core model have used consistent data in areas where they share it, e.g. the cost of fibre. 4.4 Modelling the access network: equipment at the scorched node and links to island sites Some costs in the access network are calculated in the core model. Their costs are transferred to access network services in the consolidation model. These costs comprise: • line cards at RSS, RSM, and FAM sites; Post- och telestyrelsen 74 • those line cards at Demux sites that serve links from FAM (not RSM) to Demux; • MDF, DSLAM, and splitter equipment; • some accommodation costs, including common site costs, shared pro rata to use with the core network equipment; and • fibre, FAM, and microwave point-to-point links, used to connect island sites to their scorched nodes. 4.5 Shared costs in the access network Most of the cost categories modelled in the access network are shared between different access services. The costs of all these elements are derived and apportioned to the services that share the infrastructure through the access routing table in the consolidation model (see relevant section). Ducts and trenching costs are shared between the core network and other utilities. These figures are expressed in terms of percentages in the core model, where the shared route kms are calculated for access, and access and other utilities combined. In addition to this physical sharing the access model includes a cost sharing parameter. This input allows the user to adjust the cost amount shared. These are allowed to be different for trench and for duct. Finally, in consolidation, access services are allocated a share of common business costs, using an equal mark-up approach. 4.6 Network elements for the access network Having derived a breakdown of the costs of building and operating the access network, the final stage is to convert them into costs for each access network service. This is done in the consolidation model using demand for each network service and an access routing/allocation table that defines the weightings to apply to each network service to reflect its usage of each of the main elements in the access network. Access network services will reflect the mix of line connections – PSTN, ISDN2, xDSL and fibre, plus raw copper, dark fibre and shared copper lines (i.e. those where both voice and data services are provided over one copper pair). The calculations may be summarised as follows: • Allocate costs from network elements to network services using the demand for each network service, based on the dimensions of the network and the weightings in the allocation table. This yields the total cost per service. • Calculate the cost per unit of each network service by dividing the total cost of the service by demand (the number of connections). The model is able to distinguish between costs that are recovered annually and those that are one-off. Please refer to the section on consolidation for more details on this. Post- och telestyrelsen 75 Prior to version 4.1 of the model, TeliaSonera changed the conditions for access costing, related to end-users. Following the new rules, PTS has expensed (both investment and installation cost) the following elements: • NTP; • The mini-duct from NTP to the EFSD • The copper cable costs from NTP to the SDP; and • Acc.Copper on poles. It should also be noted that trenching (and ducting) costs from the SDP to EFSD are still assumed to be recovered by the SMP operator as an annualised cost. For raw copper costs, the costs recovered are those from the core/access boundary up to (and including) the MDF. Note that since some part of the access network (even if it is “raw copper” in the real world) has been modelled using fibre, where this is a cheaper alternative (e.g. access to islands or when combined with fibre cables from other services) costs have been proportionally allocated between the relevant services. The cost of the raw copper service therefore also utilises network elements ‘Acc. Fibre from Core’, ‘Acc. SDH other’, ‘Acc. FAM’ and ‘Acc. DeMux’. 4.7 Bitstream access The main difference between shared access and Bitstream access as modelled in the hybrid model is the provisioning of the DSLAM and necessary Ethernet switching/IP routing equipment. In the case of shared access a separate DSLAM is provided and operated by the other operator, whereas in the case of Bitstream access, the DSLAM is operated by the SMP operator (and is the same DSLAM as is used for retail broadband provision). The only aspect of the Bitstream products modelled within the access model is the relevant share of the local loop. The remainder is modelled within the core model. The consolidation model then combines the two cost sources into the final product costs. Post- och telestyrelsen 76 5 Co-location model The co-location model is the simplest of the three models. In this section, a description is provided of the model methodology used to enable the co-location service costs to be derived. 5.1 Definitions and assumptions The primary purpose of the co-location model is to cost existing co-location services used for access to the unbundled loop. Therefore the model considers costs that would be borne by the SMP operator in the event of offering co- location services at appropriate sites in the network. Co-location is relevant in relation to switched interconnection, access to unbundled local loop, and for other potential purposes. The only co-location services that need to be priced on the basis of LRIC are related to access to the unbundled local loop. However, the model has much of the functionality required to model additional services, such as installation of retail access products, WLR or Bitstream access, and some of these are now also addressed. Unlike services in the core and access network, co-location services consist of relatively few cost categories. These can be divided into costs that are specific to a particular service and into costs that are shared with other services. Services in the core and access network are costed by combining costs from a pool of cost categories using a routing style allocation table technique. This is not the case with co-location products. These are mostly standalone “sub-products” that may be combined by the operator who demands co-location. Therefore, although the co-location model is simpler in structure compared to both the core and access models, special care must be taken with the co-location model in order to model and capture costs at a sufficiently granular level. The main co-location cost is the cost of space. It is assumed that space in buildings is, in the long run, an incremental cost – hence the building size is variable in the long run. Without this assumption the building costs would be fixed. Other services such as power and cabling can be considered to be ancillary services. For power the SMP operator supplies different alternatives. Similarly for cabling, different options can be chosen depending on the operator’s needs. 5.2 Structure of co-location model The co-location model includes a map of all major information flows, as shown in the figure below. Post- och telestyrelsen 77 Figure 25: Co-location model structure and navigation map The arrows show the flow of information between the groups of calculations. Solid lines are used for main flows, dotted lines for minor flows. Note that the co- location model does not show any results at the service level. Results can only be viewed in the consolidation model. 5.3 Modelling co-location services Co-location services offer an operator access to a location at the SMP operator’s site. This may either be an empty space where the operator may place racks or it may be space on a rack provided by the SMP operator. Other operators may share this SMP-provided rack. In addition, co-location services also include interconnecting the SMP operator’s equipment and the co-locating operator’s equipment. The service also requires access to the site’s infrastructure (in terms of power supply, cooling, ventilation and security). The list of co-location services modelled includes: • Location of equipment: Costs related to installing equipment dedicated to the co-locating operator. The equipment is at the SMP operator’s node. Service excludes any cable-related costs. • Station wiring: Installation of copper cable connecting the HDF (hand-over distribution frame) to the MDF (main distribution frame). This type of cable is also sometimes referred to as an ‘internal tie cable’. • Placing: Annualised costs recovered on a quarterly basis relating to equipment placed at the SMP operator’s node. For the calculation of cost Post- och telestyrelsen 78 for rack space an average utilisation rate of 75% (for shelves in a rack) is assumed. • Power, cooling and ventilation: Power consumption costs (not cabling). Each of these services is costed separately in the model. The co-location costs that are not modelled are the costs of fee for tenders, demonstration of co- location node and some services related to installation and mounting of equipment. However, much of the functionality to model these costs has been added so they can be more easily added at a later stage if required. When modelling co-location services, it is useful to separate out and categorise the different cost elements making up the services. The different cost types for co- location may be characterised as follows: • Direct costs: The costs that may be allocated directly to the operator wishing to co-locate at the SMP operator’s site. • Shared costs: The costs that are shared between co-locating operators. These costs are affected by the demand for co-location services at the SMP operator’s site. • Common costs: The costs that are shared between different increments. A share of these common costs should be allocated to co-location services. This categorisation is shown in the figure below, where normal (orange) text marks direct costs, italic (green) text marks shared costs and bold (blue) text common costs. The figure also provides an overview of the modelled cost items. Figure 26: Overview of co-location items to cost General site Installation of Power overheads: power supply, Power Cooling consumption security systems, equipment supply unit Cooling consumption costs costs storage, toilets etc. fitting etc. unit rack space Power Cooling / supply ventilation Switch DSLAM HDF Rack Rack Operator A MDF HDF Operator B Exchange Co- Opto Copper cable (copper) location cable cable space Co-location site fit-out costs (floors, lighting, fire surveillance (smoke detectors), cable chutes, cable lead-in etc Post- och telestyrelsen 79 5.4 Direct costs in the co-location model The direct cost categories are exclusively related to a specific service. Typically, they consist of a few different cost types. Being direct costs they are not shared by operators and are therefore independent of operator demand. Direct costs include: • installation costs related to location of equipment; • cable installation and annual costs (exchange cables and operator owned cables); and • power installation and consumption costs. Each of these categories are discussed in turn below. 5.4.1 Installation costs related to location of equipment Installation costs related to location of equipment consist of material and manpower costs of setting up a rack. Furthermore the model includes the cost of a standard ETSI rack. In the case where the co-locating operator uses the SMP operator’s rack, the SMP operator should be compensated for this. 5.4.2 Installation costs and annual costs relating to cable products Cable costs may be divided into cable costs related to exchange cables and those related to operator owned cables. Exchange cables are all 100 pair copper cables. The cost of these is split into different length categories (25, 50, 75 and 100 metre). No length category is specified for an operator’s own cables. These are, however, divided into copper cables at a node, optical (fibre) cables at a node, and copper cables at a cabinet. Operator owned cables are by definition owned by the operator seeking co- location. The equipment cost of these cables is therefore not included in the model. Apart from the unit cost of different exchange cables the following inputs are used to calculate the cable service costs: • materials costs used for installation (it is assumed that some fitting will occur when installing the cables); • number of hours used for installation by different staff, e.g. administrative staff used for order handling and a technician for the actual installation. With regard to station wiring the model assumes that the resource requirements may be divided into: - Planning: The costs in connection with updating and redrawing of the cabling diagrams for the internal cabling systems in the exchange; - Cable run: The costs for the cable layout, inspection of the cable guide and/or preparing existing or new cable ducts; - Installation: Costs incurred during the actual cable installation; Post- och telestyrelsen 80 • material costs used on an annual basis to maintain the installed equipment; and • number of hours used by different staff on a yearly basis for fault repair, maintenance of equipment etc. Given the hourly wage cost of different staff and the material costs incurred at installation, the model calculates the total installation costs of exchange cable types. For simplicity, installation costs are assumed independent of the length of cable to be installed. The costs of exchange cable are, however, assumed length dependent. Likewise the annual costs of different cable types are a combination of material costs and staff costs. When calculating the average staff costs an efficiency adjustment is used. The rationale for such an adjustment is that some of a technician’s time is spent on non-billable activities which include failed appointments, waiting for customers and general administrative tasks. The rest of technician time is spent on activities that are directly billable to a customer-specific operation. Customer-specific operations include installations, travel time, scheduling, registering and statistical recording. 5.4.3 Power consumption costs The following inputs are used to calculate annual power service costs: • room build costs related to power measured in SEK per kW; • operating costs calculated as a mark-up on capital costs; and • annual consumption in kWh. The consumption based costs are calculated using input on the cost per kWh. For simplicity it is assumed that the cost per kWh (this value should be the same as used in the core model) is the same for AC and DC-power. Given the cost per kWh power consumption, and assuming an average utilisation rate of 70% of the maximum power consumption allowed, the model calculates the annual cost of power consumption. 5.5 Shared costs in the co-location model The costs shared between the co-locating operators are co-location specific site costs. These costs are the additional costs needed to prepare a space at the SMP operator site for co-location. The model does not distinguish between costs that are dependent on site size and those that are independent of site size16. In order to share these costs between the co-locating operators, the model uses demand data including the total demand in m2 for co-location and number of co- locating operators per site. The co-location demand data specifies three inputs: 16The available data did not allow for a distinction: that the cost of some elements will depend on the size of the co-location space. Post- och telestyrelsen 81 • total demand for co-location space and services by site type; • the average number of operators in each site by site type; and • number of sites where co-location is required by site type area required. The demand sheet specifies the demand for different years. Any year (and demand level) can be selected. As stated in the previous section, cable costs are also allocated some shared costs. These are the room fit-out costs that are independent of size. These are cable wiring looms, cable chutes, cable lead in, etc. To estimate the costs per operator- owned cable, the fixed cost per site is divided by the weighted number of operators per site. 5.6 Common costs in the co-location model The area of space in sites that is used by co-location services defines how much of the buildings’ common costs should be apportioned to the co-location services. This calculation is carried out in the core model. To do this the core model must have the total co-location space required at different types of sites. The outputs of these values from the co-location model are manually entered into the core model, where a square metre cost is calculated specifically for co-location, including co- location’s proportion of these common costs. Costs that are common to co-location services and services in the core and access networks are: • power supply units (including back-up systems); • air conditioning units (including power consumption); • security systems; and • site preparation and maintenance. 5.7 Other service costs included in the co-location model The cost inputs for a number of services that are specific to core and access services have been modelled in the co-location model. The costs are the one-off costs related to the installation of e.g. lines. These costs are also designated as “Change costs”, being costs that are related to wholesale services where work is required by TeliaSonera when setting up a line for the provision of a wholesale service (in this case raw copper and shared raw copper). Because these services do not use core and access directly and in many cases consist of costs that are specific to work process, PTS has decided to model these in the co-location model. The services for which the co-location model contains cost inputs are: • installation of POTS/ISDN2 (new to version 5.0.5) • installation of broadband (new to version 5.0.5) • WLR – migration from TeliaSonera (new to version 5.0.5) Post- och telestyrelsen 82 • installation and operation of Bitstream Operator Access (new to version 6.0) • installation of raw copper; • installation of shared raw copper; • installation and operation of two regional Points of Interconnect (POI); • installation and operation of a local POI; • installation and operation of (2 Mbit/s) interconnection capacity; • information requests for unbundled network data, cabinet data or product data; • DACS/line conditioner removal; • Change of operator for full and shared access, or change from shared access to full access; • Suspension/Suspension Removal of shared access. The starting point for the calculations is the activities or working tasks involved in providing these services. For the first three bullets above, they are at this stage only modelled in terms of including a fixed fee for their provision. 5.7.1 Bitstream Operator Access Bitstream Operator Access involves the provision of a VPLS connection to an operator. It consists of two elements: an initial one off charge and an ongoing monthly charge. The one off charge covers the administrative and technical activities involved with the provision, together with the cost of fibre line card that is plugged into the relevant layer 2 Ethernet switch and into which the operator themselves connects. The ongoing charge relates to a portion of the common costs of the layer 2 Ethernet switch (which within the co-location model is based on a Metro switch). When installing the Bitstream Operator Access, the cost driving activities considered are: • Order processing – all ordering costs before and after physical coupling in exchange, i.e. reception, confirmation and key-in of orders; and • Coupling in exchange – ODF coupling to trunk, transportation etc. • VPLS configuration – configuration of the VPLS feed to the operator 5.7.2 Raw copper When installing raw copper the cost driving activities considered are: • Order processing – all ordering costs before and after physical coupling in exchange, i.e. reception, confirmation and key-in of orders; and Post- och telestyrelsen 83 • Coupling in exchange and in access network – dismounting of old wire, coupling to trunk, transportation etc. Coupling in access network is only needed for non-active lines. • Coupling in fastighetsnät – inspection of connection, changes made to the connection at end-user and from exchange to end-user etc. • Work in sparsely populated areas – additional transportation costs. The cost is based on actual connections made in these areas. The model calculates the cost for three different installation services: • Installation raw copper, no visit at end-customer • Installation raw copper, including visit at end-customer • Coupling in fastighetsnät Note that costs regarding examination and reservation of possible cable routes are included in the annual charges for raw copper. Costs are calculated on the basis of cost information from TeliaSonera, PTS estimates and on unit prices for activities offered by subcontractors and how frequent different activities are needed for an efficient operator. The cost for future disconnection is added to the installation charge. The NPV is calculated using the WACC applied in the model and on the assumption that a line will be disconnected every 6.7 years. 5.7.3 Shared raw copper and Bitstream access When installing shared raw copper and Bitstream access the cost driving activities considered are: • Order processing – all ordering costs before and after physical coupling in exchange, i.e. reception, confirmation and key-in of orders; and • Coupling in exchange – coupling in MDF and to DSLAM, transportation etc. • Work in sparsely populated areas – additional transportation costs. The cost is based on actual connections made in these areas. Note that the cost of installation only includes work at the exchange site. Work at the customer premises is not included. Costs are calculated on the basis of cost information from TeliaSonera, PTS estimates and on unit prices for activities offered by subcontractors and how frequent different activities are needed for an efficient operator. The cost for future disconnection is added to the installation charge. The NPV is calculated using the WACC applied in the model and on the assumption that a line will be disconnected every 6.7 years. 5.7.4 Regional and local POI In the core model, the assumption is made that any switch that is to become a point of interconnect (POI) will require upgrades to the signalling and call routing Post- och telestyrelsen 84 tables. As the cost of this upgrade depends on the switch type, the upgrade costs will depend on whether it is an LE or a TS. The POI upgrade costs are treated as if it were a switch asset. Hence the costs are annualised. The costs have been allocated to the POI_Traffic element for uses only by interconnect calls. In addition to these costs are the specific costs related to the services: regional and local POI. These are the costs incurred when setting up a POI for other operators and the costs of operating and maintaining these. For both regional and local POI, the model calculates the installation cost based on estimates of tasks involved and time spent on these tasks. The relevant tasks are: • Order processing – administrative staff tasks such as reception, confirmation and key-in of orders but also academic staff tasks such as contract negotiations and communication of information; and • Network changes – elaboration of signalling parameters, supervision functionality, traffic data etc. The annual operating costs of a POI is assumed to include the following: • Administration tasks – billing, secretarial duties, case handling, etc; and • Network management – day-to day operation and maintenance activities, e.g. individual repairs. In order to take account of the software costs related to the operating of a POI, an additional mark-up has been added to the technician’s salary. 5.7.5 Interconnection capacity Interconnection capacity refers to a 2 Mbit/s port on the interconnection exchange. The installation costs of interconnection capacity are driven by: • Order processing – administrative staff tasks such as reception, confirmation, key-in of orders and ordering of hardware; and • Mounting of hardware – including coupling. The annual costs for interconnection capacity consist of: • An annualised cost of a port unit; and • Maintenance costs calculated as a mark-up capital costs. 5.7.6 Information requests When an operator is considering either unbundling a specific line or offering service in a different exchange area, it is often necessary for that operator to make information requests on the SMP operator for certain data. This data could relate to unbundled network data, cabinet data or product data. Post- och telestyrelsen 85 Although there are no “installation” costs for information requests, there are still one-off costs associated with the activities: • Manual investigation – time spent by support staff when fulfilling the information request; and • Use of systems – relating to the accessing of data held on relevant systems. There are no annual costs related to information requests as they are a one-off activity. 5.7.7 DACS/line conditioner removal From time to time, an operator might wish to provide service to a prospective customer over a local loop that currently has incompatible equipment on it, such as DACS equipment or line conditioners. The one-off costs of DACS/line conditioner removals are driven by: • Order processing – administrative staff tasks such as order handling, line status analysis; and • Removal of hardware – technical staff tasks associated with the physical removal of the relevant equipment. There are no annual costs related to DACS/line conditioner removals as they are a one-off activity. 5.7.8 Change of operator or shared to full access Should there be a change of operator on a currently unbundled line, or a request to migrate from a shared unbundled line to a full access one, then there will be certain tasks necessary to fulfil such requests. The one-off costs of product change requests are driven by: • Order processing – administrative staff tasks such as reception, confirmation, key-in of orders; and • Technical reconfiguration – including coupling. There are no annual costs related to product change requests as they are a one-off activity. 5.7.9 Suspension/Suspension Removal of shared access Occasionally the SMP operator might experience problems with the operation of a telephony service on a shared access line due, for example, to interference caused by the other operator’s equipment. In such instances, the SMP operator might be forced to disconnect the other operator’s equipment until such time as the problem is resolved. The one-off costs of such suspension/removal of suspension are driven by: • Technical reconfiguration – including coupling. There are no annual costs related to these suspension/suspension removal costs as they are a one-off activity. Post- och telestyrelsen 86 Appendix 1 List of abbreviations used in this report ABC Activity Based Costing AC Alternating Current ADM Add Drop Multiplexer ADSL Asymmetrical Digital Subscriber Line ATM Asynchronous Transfer Mode BHCA Busy Hour Call Attempt BHE Busy Hour Erlang BHG Busy Hour Gigabits per second BHMP Busy Hour Mega packets per second BU Bottom-Up BUWG Bottom-Up Working Group CAPEX Capital Expenditure CAPM Capital Asset Pricing Model CCA Current Cost Accounting CoC Cost of Capital Codec A device capable of coding and/or decoding a signal CVR Cost Volume Relationship DACS Digital Access Carrier System DC Direct Current DP Distribution Point DSLAM Digital Subscriber Line access Multiplexer DWDM Dense Wavelength Division Multiplex EAR Engine Access Ramp EFSD Exit From Street Duct ETSI European Telecommunications Standards Institute FAM Fibre Access Multiplexer FCM Financial Capital Maintenance FWA Fixed Wireless Access Gbps Gigabits per second GBV Gross Book Value GIS Geographic Information System GoS Grade of Service Post- och telestyrelsen 87 GRC Gross Replacement Cost HCC Homogenous Cost Category HDF Hand-over Distribution Frame HSS Host Subscriber Switch IN Intelligent Network IP Internet Protocol IPTV Internet Protocol (based) Television ISDN Integrated Services Digital Network IX Interconnect Exchange Kbps Kilobits per second KW Kilowatt KWh Kilowatt Hour LE Local Exchange LIC Line Card module LRAIC Long-Run Average Incremental Cost LRIC Long-Run Incremental Cost LX Local Exchange Mbps Megabits per second (also Mbit/s) MDF Main Distribution Frame MEA Modern Equivalent Asset MRP Model Reference Paper Mpps Mega packets per second MSG Multi-Server Gateway NBV Net Book Value NGN Next Generation Network NPV Net Present Value NRC Net Replacement Cost NTP Network Termination Point OCM Operating Capital Maintenance OPEX Operating Expenditure PDH Plesiochronous Digital Hierarchy PDP Primary Distribution Point POI Point of Interconnect POP Point of Presence Post- och telestyrelsen 88 POTS Plain Old Telephony Service PSTN Public Switched Telephony Network PSU Power Supply Unit PTS Post och Telestyrelsen QoS Quality of Service RSM Remote Subscriber Multiplexer RSS Remote Subscriber Switch SDH Synchronous Digital Hierarchy SDP Secondary Distribution Point SEK Swedish Kroner SMP Significant Market Power SS Subscriber Switch SSP Service Switching Point STM Synchronous Transport Module (as in STM-1 etc) TD Top-Down TS Transit Switch ULL Unbundled Local Loop VLAN Virtual Local Area Network VoIP Voice over IP VPC Virtual Private Circuit VPLS Virtual Private LAN Service (a way to provide Ethernet based multipoint to multipoint communication over IP/MPLS networks) WACC Weighted Average Cost of Capital WDM Wavelength Division Multiplexing WLR Wholesale Line Rental xDSL x Digital Subscriber Line (includes such technologies as ADSL) Post- och telestyrelsen 89
"Hybrid Model Documentation v6.1"