Consolidation Server consolidation p Server virtualisation p Storage consolidation p

Consolidation Server consolidation ................................. p.40 Server virtualisation .................................. p.44 Storage consolidation .............................. p.48 Storage virtualisation ............................... p.52 39 Consolidation Server consolidation Low cost and easy administration with fewer servers The IT infrastructure of a company or institution is characterised by large numbers of servers which have been procured over a period of time and which usually serve a single purpose. The Active Directory service, for example, is usually resident only the corresponding dedicated server. If other essential server services are required or in widespread use, they too tend to be resident on such dedicated servers. Examples include mail servers, file servers, terminal services and database servers. To achieve the highest possible stability of those servers, only one service per server is usually configured, so as to facilitate troubleshooting by administrators. But often key servers also need to be paired to create a High Availability Cluster. In this, the critical service runs on one server actively as the active-passive variant while the second server functions only as the hot-standby server, ready to intervene and execute the services of the first server if it should fail. This increases the number of low-workload servers even more. Consequently, mediumsized enterprises are often running as many as 10 servers which have been procured over time to provide the services cited. The average CPU workload of such servers is often just a few percent of their capacity. The same is true in terms of memory capacity utilisation. On the other hand, high-availability concepts pose a quite different problem. As such concepts often entail a wish to incorporate an existing server at a later date, difficult to overcome obstacles are suddenly encountered, because identical server hardware can often no longer be sourced. siderations are also relevant, in terms of the targeted reduction of CO2 emissions. The management of complete computer centres featuring a wide variety of different server technologies is no easy task for IT administrators: stocking the wide range of spares costs money and takes up space, and the know-how required of the IT department often comprising just a single person - is highly unlikely to be available to cope with a heterogeneous server environment. There is unfortunately no single, unified management interface available for infrastructures which have grown up in this way. More recent servers may feature [IPMI] functionality for example (Yellow Pages, page 78), while older machines are unable to provide the same hardware status monitoring and remote management facilities. All these factors represent good reasons to change server infrastructure quickly, in order to make optimum use of existing servers working at low capacity. The newly acquired servers should, however, deliver generally lower energy costs. This will also result in lower running costs for the air conditioning system in the server room. The capacity utilisation of the CPU and the memory must be configured higher, so as to make full use of available resources. A unified interface is advisable, to replace the heterogeneous administration environment and, as far as possible, provide centralised monitoring of all vital server functions. This will significantly cut administration costs. Particular attention should be paid to failsafety (see “High availability based on clustering“). If multiple business-critical services are assigned to one server, it must also be provided with redundant components so as to deliver the maximum availability. Lengthy downtimes of such key servers not only hinder business processes, they also incur costs and result in loss of earnings. transtec SuperBlade Top performance in a tiny space Features in focus  Low space take-up and reduced cabling  10 nodes per chassis, up to 80 processor cores in 7 HU  Easy to manage and optionally expandable  Energy-efficient power supply units with up to 93% peak. GO WEB GOWEB Detailed information under GOweb SuperBlade Why server consolidation? Server accumulation of this kind entails many disadvantages. First of all, in a time of rapidly increasing electricity prices, the energy costs are a notable factor. Climate protection con- 40 Examples of possible means of server consolidation [Blade systems:] Blade systems represent an alternative to the familiar server designs in differing heights for 19” racks. The servers are executed as so-called blades (server units) which are slotted into a shared chassis. A typical blade chassis may house between 10 and 14 blades. Such systems are of interest wherever 10 servers or more are being consolidated. The blade system can then be fully loaded to maximise cost-effectiveness. In this case the higher cost of the blade chassis is rapidly amortised, as it also brings savings in energy costs compared to single systems. Other design features are the central contacts on a single backplane, by which all blades have simultaneous contact with a network switch. Each blade module can also be fitted with I/O interface cards. This enables high-performance I/O interfacing using InfiniBand adapters for example, which is also suitable for HPC clustering. And last but not least, the blade chassis’s compact design saves space in the rack and permits simpler wiring than single systems, which again saves money as well as optimising cooling air flow. It is also easier for administrators to administer the individual blades in a system than it is to administer separate single systems. The identical feature specification of the blade hardware means a unified interface can be used to read the vital data of all modules. The facilities for remote installation of operating systems by way of the local drives of a management computer make operating system deployment much simpler. It is justifiable to purchase a blade system whenever full configuration of the blade chassis is assured within the shortest possible time. If not enough blades are configured, then single servers are the much more cost-effective option. Modular server systems: The blade concept also provides the foundation for a new development which will be groundbreaking in terms of server consolidation. For many small and medium-sized enterprises, five or six servers are adequate. However, this would not fully utilise the capacity of a standard blade chassis, which holds ten or more blades, and so a purchase would not be cost-effective. Moreover, it is usually necessary to procure a fibre-channel or iSCSI-SAN infrastructure to provide the servers with sufficient storage space for user data, and that is always costly. The blades themselves usually only have space for two hard disks, which are used as RAID-1 for the operating system. This means it is not possible to provide sufficient storage space for applications and data. The new transtec Modular Servers offer groundbreaking solutions to this problem. They combine up to six blades and a pool of 14 2.5” hard disks in one blade chassis. Their special feature is that the [Blades] themselves have no hard disks of their own (meaning they are diskless systems). The hard disks can never- transtec Modular Server Your 6-HU computer centre The Modular Server, based on Intel hardware, offers SMEs a server system featuring uninterrupted installation, seamless migration and scaleable growth. Features in focus  Up to six server compute modules in six HU  Up to two GbE switches and storage control modules  Easy administration with the server administration module  Shared storage in the form of up to 14 2.5“ SAS hard disks GO WEB GOWEB Detailed information under GOweb Modular Modular Server chassis & components 41 Consolidation theless be assigned to the individual blades by way of a controller. This facility provides for flexible assignment, which even integrates some SAN functionality within a blade chassis. So if a blade fails it can be exchanged very easily for a replacement unit, as there is no need to move the local hard disks. The system is administered solely by way of a web browser. This means remote management is very simple, enabling hardware vendors to carry out maintenance easily when provided with dedicated system access – a plus-point in which the Modular Server system even outdoes the functionality of commonly used blade systems. Server consolidation Conception  High power supply efficiency. Efficiencies of up to 90% are possible nowadays.  The CPUs must feature power management mechanisms to reduce power consumption under low CPU workload.  The partitions for the operating system must be set up on a RAID hard disk array. To enable exchange of a defective hard disk while the server is running, the servers must be housed in hot-swap frames.  Remote administration facility based on [IPMI]-functionality  Protection by adequately dimensioned UPS systems is also essential. Do remember, though, that when virtualising server services multiple virtual servers rely on the availability of a small number of physical systems. Server consolidation based on virtualisation entails even greater necessity for a fail-safe solution with an optimally reduced MTBF (Mean Time Between Failures), as any failure would affect much more than just one application. Consequently, it is vital that the critical components such as power supply units should be of redundant design and be mounted on a tried and proven platform. It is advisable to achieve a balance between hardware and software. Our transtec experts will be glad to help you find the ideal balance for your specific virtualisation. Modular Server Management interface Server consolidation should not be reduced merely to deployment of the right hardware. Virtualisation is also an aspect which should not be neglected. To minimise the number of consolidated servers in hardware terms and yet still meet a customer’s wish for separate GO services, server virtualisation is a useful add-on. GOWEB Software vendors such as VMware (GOweb VMware) and ( GOweb Xen) offer the key server consolidation tools. The operating system then no longer runs directly on the physical server, but instead the virtualisation software, such as VMware ESX Server, is installed on the server. The core of this is the Hypervisor, which represents an abstraction layer between the physical hardware and the so-called virtual machines. These virtual machines represent a software map of the server hardware. The desired operating systems are then installed into these virtual machines. For a detailed description of the latest virtualisation technologies see “Server virtualisation” on page 44. WEB GO WEB GOWEB 42 emulex.com Enabling the Next Generation Data Center Superior quality and reliability* Maximum SAN performance Most efficient installation and management Fastest diagnosis and recovery Broadest enterprise deployment * Field Mean Time Between Failures (MTBF) of 3.5 million hours Emulex Host Bus Adapters (HBAs) deliver what IT managers need. That’s why 90% of the world’s largest companies trust Emulex to provide high–performance connectivity solutions for their business critical data. Whether you’re deploying an enterprise SAN, a blade server solution, a virtualized server environment or an entry–level storage solution, Emulex can connect you and protect you—with HBAs that are Generations Ahead of the competition. Find out more about Emulex next generation Fibre Channel HBAs. Visit www.emulex.com. NYSE:ELX | +44 (0) 118 977 2929 This document refers to various companies and products by their tradenames. In most, if not all cases, their respective companies claim these designations as trademarks or registered trademarks. This information is provided for reference only. Although this information is believed to be accurate and reliable at the time of publication, Emulex assumes no responsibility for errors or omissions. Emulex reserves the right to make changes or corrections without notice. 08-048 43 Consolidation Server virtualisation Cost savings by server reductions Since the early 1990s, a trend towards decentralisation has been spreading through the computer centre business, guided by the well-justified aim of enhancing application and system safety, as well as fail-safety, by horizontal spreading of resources and avoiding the creation of any single point of failure. This trend was driven forward by the continually reducing cost of hardware. The downside to these advantages, though, is inadequate capacity utilisation of what are, in effect, utterly over-dimensioned hardware systems. There are also higher infrastructure costs in terms of power, network connections and air conditioning. Another factor is the increased operating and maintenance cost, because of course 20 distributed server systems are not easier to administer than a single mainframe computer. Virtualisation of server systems and services is one of the key means of countering those problems. Virtualisation provides the link between the centralised server operations on the one hand and the decentralisation of systems on the other: the physical layer is centralised to consolidate the infrastructure and the computer landscape and so cut costs. At the same time, the provision of individual dedicated virtual servers for special services represents a decentralisation on the logical layer. The main virtualisation methods at present are:  full virtualisation  paravirtualisation  native or hardware-assisted virtualisation. What all the methods have in common is their role as a so-called Hypervisors, assigned the highest privileges and controlling the virtual machines and their access to resources such as memory, the CPU and so forth. Server virtualisation with Citrix XEN™ transtec CALLEO 300  Citrix Xen™ certified Server Virtual data center with high availability Citrix XenServer™ Virtualisation software transtec offers turnkey Citrix XenServer™ virtualisation solutions based on CALLEO300 servers. Customised and installed on-site. Including  Basic installation of Citrix XenServer™ on the server  Upgrade to a latest patch levell  Mirrored hard disks in the nodes  Creation of data stores for VMs  Integration into existing 19“ environment  Interface to corporate network  Function test and administrator induction Virtualisation technologies Virtualisation means breaking away from one of the central paradigms of computer architecture and introducing an abstraction layer between the basic hardware and the operating system and applications running on it. The idea is not new: as far back as the early 1960s, concepts were devised to split the mainframe computers used in the research environments of the time into so-called partitions, so as to isolate different applications and user environments from each other. The new feature is that virtualisation has penetrated into the everyday world of computer centres and server environments, both in the research field and in industry, based on considerations of cost-effectiveness and security. Full virtualisation: A complete hardware environment is simulated. Neither the guest operating system nor the installed applications notice that they are running in a virtual machine. This method of virtualisation offers the most comprehensive support to guest operating systems, and is deployed for example in VMware products such as VMware Infrastructure 3. Paravirtualisation: The original virtualisation method employed by the XenServer family. This does not simulate a complete hardware environment. Rather, the guest system is allowed access to a number of central hardware-based structures of the host system such as page tables or interrupt tables. The guest system is implemented at a lower privilege level than the Hypervisor. The virtualisation overhead is correspondingly lower, so paravirtualisation permits very high guest system performance. The guest operating system notices that it is running in a paravirtual environment, and must be adapted accordingly – so virtualisation of Windows systems is not possible. Following the launch of the latest GO WEB GOWEB More details on the XEN servers GOweb XenServer 44 processor families from Intel and AMD, pure paravirtualisation is becoming less and less important however. Native or hardware-assisted virtualisation: This is, so to speak, a hybrid solution between full virtualisation and paravirtualisation, combined with I/O speed-up functions. This method of virtualisation requires systems with latest-generation processors featuring appropriate add-ons, known as Intel-VT for Intel processors and AMD-V for AMD processors. Its advantage is that the guest system does not have to be adapted, and also notices nothing of its virtual environment. This makes virtualisation of Windows systems also possible. Another key advantage is the added security based on the partitioning (mutual insulation) of server services: processes are not able to break out of their virtual environment. This means the effects of memory leaks, compromise due to software bugs or security gaps are restricted to the specific virtual machine and do not spread to other services. And there is also added availability to be gained from fail-safety: a HA cluster of multiple host systems can operate one or more virtual machines together. They can be migrated live to another system in the event of a failure of one of the host systems. The VMware VMotion component of VMware Infrastructure 3 offers this functionality, as does XenServer, provided a shared storage area is set up. A virtualisation solution can be tailored to your needs and adapted to your environment. Transtec experts will be glad to assist in evaluating which virtualisation concept will be of most benefit to you. Server virtualisation with VMware® Intel® Managed Server or transtec Modular Server What are the advantages of server virtualisation? Server virtualisation enables a number of objectives to be achieved, the most important of which is consolidation: the resources of a usually high-spec server are assigned to the individual virtual machines according to their specific needs. This avoids overcapacities and so directly cuts hardware procurement and maintenance costs. It also eliminates infrastructure costs, as virtual machines do not need dedicated network or power connections. Another advantage is that virtual machines do not age, and so do not have to be repeatedly reinstalled. Moreover, using virtual machines means you are not under continual pressure to update: you can be sure that a specific legacy application which only runs on a now outdated operating system will still run on a virtual machine in 10 years’ time. VMware® Infrastructure 3 ESX Virtualisation software Intel® managed and transtec Modular Servers are the transtec recommendation for VMware® virtualisation. In this field transtec offers an extensive portfolio of solutions tailored to your needs. Includes integration and on-site installation. Including  Upgrade to a latest patch level  Interface to a directly connected storage system and creation of data stores  Solution setup on-site  Interface to corporate network  Function test and administrator induction in application of the solution GO WEB GOWEB More details on the Intel®-Server GOweb Intel-Managed VMotion technology 45 Professional Wireless LAN from D-Link: No wires, no problems. DES-6500 Server Farm PROFESSIONAL WIRELESS LAN DWS-3024 Wireless Switch DWS-3024 Wireless Switch Our intelligent wireless LAN technology provides you with an entirely new world of flexibility, delivering innovative new possibilities and professional protection for your network. Administer your network centrally and wirelessly by way of wireless switches, offering the highest security levels and featuring optimized data transfer. Intelligent and secure as wired – flexible as never before: professional WLAN solutions from D-Link. http://wireless-switching.dlink.de 46 Reference: Josef Manner & Comp. AG The English slogan above is used in advertising for Manner AG. Manner has been around for more than 100 years and with approximately 800 different products is considered to be the flagship for the Austrian confectionery market. The traditional Viennese company relies today on transtec, the European IT system manufacturer, to ensure the security of its Windows Server solution. Manner relies on Windows Server solutions from transtec Manner decided on an Intel-based server landscape from transtec to ensure company stability, make it possible to run e-mail and database applications and also guarantee that sensitive data would be correctly backed up. All in all, Manner now uses 15 Pentium® III dual processor servers and two Xeon™ dual processor servers, which work as domain or backup domain servers, SQL servers with OLAP applications, and also as exchange or file servers. In addition to all of that, a sales information service, plus backup and virus scan solutions run on the transtec servers. Company security is guaranteed with a transtec SAN solution that is made up of two fibrechannel RAID systems with a capacity of 400 GB and two corresponding Gadzoox Capellix fibre hubs. Manfred E. Schwarzbier, Country Manager of transtec Austria, sees this successful cooperation with the Manner AG as confirmation of transtec’s strategy to concentrate on individual cluster, server and storage solutions. “transtec is already one of the top 10 server providers in Europe. Specialising in our core competence benefits, most of all, customers who are looking for a tailored hardware solution, such as the Manner AG. DJosef Manner founded the company in Vienna in 1890. In the beginning, he sold chocolate and coffee from a small shop in the Stephansplatz, but shortly thereafter went into chocolate production himself. He established the “Josef Manner Chocolate Factory” which quickly became known for its high-quality sweets. Today, the Manner AG is the largest confectionery manufacturer in Austria. The best-known Manner product - Manner Wafers - is exported to 50 countries throughout the world. The secret to Manner’s success is without a doubt a highly committed management team and approximately 700 equally motivated employees, all of whom contributed to a turnover of 102.4 million euros last year. But another important factor to positive business developments at Manner is a perfect interplay between all the communication technologies used in the company, plus tailored server and network infrastructures that increase the company’s efficiency even more. Customer Evaluation “The quality and, above all, excellent priceperformance ratio of the hardware solution plus competent on-site support by the transtec team were all decisive factors,” explains Gerhard Fischer, IT manager at the Manner AG. “Other than that, transtec also has years of experience in the server, storage and network sectors, which is a fantastic advantage that really accelerated integration of our new hardware.” 47 Consolidation Storage consolidation Strategies for the growing complexity and rising demands made of storage systems The issue of storage consolidation remains as vital as it has always been for many businesses. It conceals a number of very real demands in terms of the everyday operations of a growing IT infrastructure which require more than just formulaic answers. Based on growth over a number of years, computer rooms frequently house a widely varying mix of equipment from different vendors featuring internal storage, directly connected RAID storage systems, SAN islands and file servers. The maintenance of such legacy structures is not only time-consuming and costly; they also pose problems in terms of establishing and safeguarding uninterrupted operation. Consolidating and simplifying storage infrastructure brings many systemic advantages, cuts IT operating cost and delivers the foundation for high-availability and business continuity solutions. A successful consolidation project should incorporate a perspective beyond current day-to-day operations, formulating responses to a continuous or sudden increase in capacity and performance requirements, so as to avoid having to implement the next consolidation project a few years later. The overall concept should also incorporate backup and restoration of systems and data. Most of the requirements encountered in storage consolidation projects can be derived from the essential tasks of storage systems:  Capacity: providing the application servers with capacity in the form of logical drives  Performance: handling access to data at fast speed  Security: protecting the stored data optimally and keeping it continuously available for access Existing storage systems often no longer meet one or more of the above requirements after a time and must be replaced or supplemented by additional storage systems, which in turn necessitates further consolidation later. Can such problems be avoided altogether, or is it possible only to slow the growth in complexity? 3,5” Desktop Capacity per Platter 1.000 Areal Density (Gb/in²) 100% CAGR 40% CAGR 480 320 240 160 960 640 100 120 PATTERNED MEDIA, HAMR? 40 PERPENDICULAR 10 10 LONG. SAF LONGITUDINAL 1999 2006 2012 Areal Density and Capacity vs. Time Responses to continuously rising capacity demand Market research organisations report that data volumes have risen at an annual rate of around 60 percent in recent years. No slowing of the trend is in sight. Fortunately, hard disk capacities have been doubling every 18 to 24 months, while prices have remained largely stable. Hard disk manufacturers such as Seagate are already researching and testing new technologies such as Heat Assisted Magnetic Recording (HAMR), High Anisotropy Bit Patterned Media and Self Organised Magnetic Arrays (SOMA). They are expected to deliver further annual growth in hard disk capacities of 40 percent through to 2012 and beyond, so existing growth rates and proposed solutions remain applicable in terms of long-term capacity planning. Storage consolidation demands scaleable storage systems which will adapt to the increasing data volumes of the business over a number of years while also protecting the initial investment in storage media. Capacity-scaleable storage systems were previously limited mainly to fibre-channel arrays, but are now also available in lower-cost SAS/SATA technology as well as the transtec PROVIGO 550 and PROVIGO 600 RAID series. Potential follow-up costs for a later capacity upgrade should be analysed at an early stage. A theoretically possible expansion to a hundred or more hard disks is practically useless if, when an upgrade and extension of the service contract SAN & NAS solutions from transtec, including  Project design & conception  Installation & configuration  Integration into your IT environment  Training of your users 48 is due, the storage system turns out to be an expensive money trap which can only be escaped by procuring a completely new system. Storage virtualisation is an essential element of any consolidation planning and long-term IT strategy extending beyond the service lives of individual storage systems. A virtualisation layer presents the application servers with logical drives irrespective of the underlying storage systems. Whatever maintenance or other modification is required on the hardware side, it is done in live operation and unnoticed by the applications, thanks to the protective virtualisation layer. This delivers increased flexibility in meeting rising demands and at the same time enhancing the availability of storage services. As a result, data can be migrated from the old storage system to the new without interrupting IT operations. Online upgrading of the storage pool with new storage systems is also a simple procedure. For more information refer to the “Storage virtualisation” section. are also provided by RAID system architecture. Traditionally, RAID storage systems scale by cascading hard disk storage enclosures (JBODs) on two to four cables, with the disks in two (opposite-running) loops. This results in increasing latency and, in conjunction with the extra administrative commitment for the RAID arrays, can even produce a net loss of performance following an upgrade. The more recent SBOD (Switched JBOD) architecture should always be preferred in this respect. It does not organise the disks in a loop, but instead addresses them directly by an embedded switch in the backplane of each enclosure in a switched fabric. As the number of disks increases, the response time for the applications is improved by as much as 50 percent. Storage consolidation with transtec PROVIGO 550 RAIDs transtec PROVIGO 550 Business  Scaled up to 60 SAS/SATA HDDss  Single/dual controller  RAID-6  Up to 192 snapshots wit hVSS supportt  4Gbit FC host interface More performance for IT applications Efficient business processes demand an efficient IT function which supports staff by means of fast information and transaction systems and does not unnecessarily irritate customers due to lengthy response times to queries. Consequently, the performance capability of the storage solution is the second essential element of any consolidation strategy alongside the capacity reserves. The performance capability of classic RAID storage systems is much more limited than their capacity, and depends on the types and numbers of storage media used as well as, essentially, on the storage controller - see also the transtec Technology kompass, “Storage concepts“. Storage consolidation solutions should therefore ensure adequate performance reserves are available or, even better, select a storage system which grows in a largely linear manner in terms of performance as well as capacity. Looped JBOD vs. Switched SBOD transtec PROVIGO 550 SUMO SATA  48 HDDs in just 4 HU  ½ PetaByte per 19” rack  Redundant controllers  RAID-6  Up to 192 snapshots wit hVSSSupportt  4Gbit FC host interface Storage clusters: The best method would be a clusterable storage system in which - as in HPC clusters - performance capability also increases with each storage node based on additional controllers, caches and host interfaces. In a storage system of this kind buyers can be certain that they will not come up against the limits of their solution, either in the short or long term, and need to procure additional storage systems which would increase the complexity of the overall solution. The best example of a clusterable storage system of this kind is the EqualLogic Peer Storage (PS) series. A PS module features redundant RISC controllers, 16 SAS or SATA hard disks and three Gigabit-Ethernet (iSCSI) host connections. Up to 16 PS modules can be combined to form a The PROVIGO 550 series is a flexible, high-availability storage platform for the consolidation of your applications and services. Based on the mix of SAS and SATA drives, you can tailor the performance, capacity and cost optimally to your needs. The SUMO SATA RAID provides a fail-safe near-line and archive system with high storage density. GO WEB Performance scaling: The 100,000 per second or more cached hit rates quoted by many manufacturers are largely worthless. Alongside realistic benchmarks with genuine disk hits, reliable indications GOWEB Detailed information under GOweb PROVIGO550 49 Consolidation transtec PROVIGO 632 Premium RAID With the compact PROVIGO 632 Premium RAID, IT administrators can run business-critical applications on 2.5” high-performance SAS drives. At the same time, lowcost SATA hard disks can also be used for snapshot storage or nearline data. Storage Cluster data, demand their own dedicated performance optimisation methods. In OLTP systems the key factor is I/O performance. Databases should be resident in the cache itself or at least on fastspinning hard disks. Also, the cache should be optimised for write operations and each category of I/O activity (database, transaction log files, tempdb, etc) should run on separate hard disks, RAID sets and - ideally also - separate storage controllers. Applications with read-intensive as opposed to write-intensive access should not be resident on the same storage system – or at least on separate controllers and cache partitions. The same applies to applications with primarily sequential as opposed to random access. Areal Density (Gb/in²) Features in focus  12x 2.5” SAS hard disk (SFF)  Scaleable up to 60 disks  Also supports connection of 3.5” SATA JBODs  Single/dual controller  High I/O performance of up to 39000/143000 IOPS (w/r)  RAID-6 support  VSS snapshot support  4GB FC host interface cluster. Load balancing then not only takes place on the host side across all ports and controllers, but the data is also distributed dynamically across all available disk spindles in the cluster. As a logical consequence, even a three-node iSCSI cluster from EqualLogic outperforms comparable Tier-1 FC storage systems. And administering an EqualLogic PS cluster is unbelievably easy too. Many traditional storage administration tasks have been automated, and require no lengthy IT staff training courses or accustomisation times. A live demo on-site or remotely by webcast demonstrates exactly how user-friendly the solution is. A schedule of demos is offered on a regular basis by transtec and EqualLogic, or they can be arranged specifically by appointment. Which storage medium should be used in consolidation? An increasing number of businesses - not just at the top end of the enterprise sector - additionally face the challenge that the I/O transaction rates achievable by hard disks are no longer adequate to handle specific business transactions or peak loads. Storage consolidation concepts must take account of this peak demand too. Much faster hard disks than the current 15,000 rpm generation are not yet in sight, so alternative storage media need to be used. Small form factor storage: The 3.5“ fibrechannel hard disks currently still dominating the market in external storage systems will be gradually replaced over the next five years by 2.5“ Capacity (Gbytes) Performance (IOPS) Rack Height Power (Watts) 1.000 100 GO WEB External storage controllers (virtualisation): A similar method is employed by means of storage virtualisation. In this, multiple virtualisation appliances (SAN storage servers) serve as external storage controllers, splitting data access between them and protecting each other as a (N+1) cluster. This means storage systems and virtualisation appliances can be added step-by-step in line with capacity and performance demands. This method is detailed in the “Storage virtualisation“ section. It does, however, entail greater complexity than an EqualLogic PS cluster, because the storage systems and virtualisation appliances are administered separately. Limits of consolidation: It does not always make sense in technical terms to host applications identically in the same storage system. Different applications, such as online transaction systems (OLTP), data warehousing and sequential file 10 1999 10K. 3.5” 4.5TB 126 HDDs x 36 GB 15120 126 HDDs x 120 IOPS 9 3U arrays 27U 2016 (W) Savvio 2.5” 4.5TB 15120 10U 1/3rd space 1125 (W) 45% less Advantages of 2.5“ HDDs in the computing centre GOWEB Detailed information under GOweb PROVIGO632 SAS disks. The smaller magnetic disk size not only shortens travel when accessing data, but also reduces mechanical strain and power consumption. The result: At 70 percent lower volume in the computer centre, fail-safety is improved by up to 15 percent and power consumption is cut by as much as 50 percent. This not only cuts power 50 costs at the computer centre, it also reduces thermal pollution. At the same time, I/O performance is boosted significantly due to the lesser radius and the use of more than twice as many spindles in the same space. Solid-state disks / Flash memory: Another increasingly practicable approach is to supplement the hard disk by alternative storage media such as solid-state disks [SSD], flash drives or [hybrid drives]. It is a tried and tested practice to load as many parts of a database under high load into the server RAM so as to speed up access. As non-volatile storage for such hotspots, enterprise customers have at their disposal SSD appliances such as RamSan from Texas Instruments. There are, however, increasingly also SSD, flash and hybrid drives in 2.5“ and 3.5“ format for installation in conventional storage systems, such as the STEC Zeus drives. These can in future assume the role of performance-boosters for selected applications. with the applications - whether a snapshot with Oracle-Agent or a Virtual Shadow Copy Service (VSS) snapshot for Microsoft Exchange/SQL, as available in the transtec PROVIGO 550 and PROVIGO 600 storage systems. Software is now also available for continuous, timed or transactionspecific backup of data in the form of a “Snapmovie“ (Continuous Data Protection – [CDP]) to downstream near-line hard disks. Whether as an optional component of conventional backup software or as third-party software permitting to-the-second data rollback. Snapshot and CDP solutions in turn require asynchronous replication to transfer the saved data to the off-site storage and transfer it back in the event of loss or damage to the primary data set - see also “Asynchronous replication“. The compatibility of storage systems procured as part of storage consolidation with such backup mechanisms is an important secondary consideration in terms of the long-term benefit of the investment. FASTER HARDWARE SPEED SLOWER 1970 2000 CPU Solid-state disk closes the gap! Real disk speed Improvement SSD closes CPU/HDD performance gap Security and availability of data The backup and restoration of storage systems and data also needs to be considered as part of consolidation planning. The concentration of applications on a small number of centralised storage systems also increases the demands in terms of availability and the impact of any downtime on business operations. Traditional tape backup has now become obsolete as the first level of storage: it is too slow and to complex to restore - see also „Efficient backup concepts“. Aids to backup are the snapshot mechanisms, which are becoming increasingly tightly meshed Turnkey storage networks from transtec, including  Project design & conception  Installation & configuration  Integration into your IT environment  Training of your users 51 Consolidation Storage virtualisation Real benefits in IT operations based on virtual storage space Although storage virtualisation is marketed as a new accomplishment, the technology has long been known and successfully used in many companies under a different name. Storage virtualisation primarily involves abstracting real, physical storage resources by means of a logical, vendor-neutral intermediate layer which provides the applications and their servers with virtual storage pools. That sounds groundbreaking, but merely describes a function of any standard commercially available RAID controller, whether as software, a PCI card or in an external hard disk subsystem. A RAID controller consolidates multiple physical hard disks - including from different vendors - into one or more logical volumes and makes them available to the applications as LUNs. The LUN (Logical Unit Number = virtual disk) may be larger or smaller than a physical hard disk. This virtualisation is one of the basic prerequisites for the subsequent protection against failure of individual storage media. And anyone who has used the operating system Volume Manager to combine two logical drives of different RAIDs to create one large drive has also already practised the so highly-valued vendor-neutral virtualisation of external storage. So if storage virtualisation is such a well-known principle, what is this trend all about? or Linux as its operating system. But should you really interpose a server between your application servers and storage systems? Is that not an avoidable performance bottleneck in the SAN? Is it not better for the storage services to reside in the storage disk array, as in device-based virtualisation, and be administered by the RAID controller itself? We dare to suggest that it ultimately makes no difference. A look at the specifications of a mid-range disk array such as the market-leading EMC CLARiiON series shows that the controllers feature Intel Xeon processors. The controller is de facto a proprietary server board featuring the classic Intel x86 architecture, RAM modules and operating system. In earlier years it was even possible to see Windows NT booting in the background when starting up a CLARiiON. Like in all RAID systems, the controllers are of course positioned in the data path between the servers and the storage media (hard disks). So ultimately external virtualisation appliances and mid-range disk arrays are technically equivalent both represent a form of [in-band virtualisation]. Virtualisation not only has a lot in common with the functions of a RAID controller. A virtualisation appliance can ultimately be seen as an external storage controller, which is merely not in the same housing as the hard disks. What remains of the debate is ultimately the question as to the performance capability of a specific controller and its firmware or software. Because any I/O handling requires a processor, a bus architecture, I/O ports and software to control the system. The software is usually written in languages such as C/C++, Assembler and Java/HTML (GUI), and either loaded as embedded firmware into a nonvolatile flash memory chip or running as software on an operating system. Whether that construct is located together with the hard disks in the same housing or not is ultimately irrelevant. But it was not for nothing that EMC selected x86 processors in this context. The latest multi-core server processors from Intel, AMD and SUN Microsystems (SPARC) provide significantly higher performance than any available embedded CPU. Owing to their shorter development lead times and lifecycles, servers are usually several generations ahead of the storage systems. And although other factors such as cache management, the number of available PCI channels for I/O traffic in the controller Higher performance in the SAN - Storage virtualisation or mid-range disk arrays? The literature on storage virtualisation tends to confuse the interested reader by a large number of contradictory statements as to which method offers the best performance and security. The debate is focused on ”fabric-based“ (appliances) and “device-based“ (mid-range disk arrays) virtualisation – see also “Fully redundant storage“. In fabric-based virtualisation an item of hardware is usually inserted in the I/O data path between the server and the storage systems. This hardware might be an intelligent network switch, or a virtualisation appliance with x86 server architecture and Windows Virtualisation solutions from transtec, including  Project design & conception  Installation & configuration  Integration into your IT environment  Training of your users 52 and the efficiency of the software routines play an important role (see Technology kompass, “Storage concepts“), an external storage controller or virtualisation appliance will in almost all cases offer much higher performance than any single entry-level or mid-range storage system. It of course makes little sense to connect a dozen high-performance mid-range storage systems to just two redundant virtualisation appliances – a performance bottleneck would then be no surprise. The same applies to the large enterprise disk arrays with crossbar architecture which internally often feature as many as 128 processors and I/O buses. By contrast, it makes a lot of sense to connect virtualisation appliances to entry-level fibre-channel, SAS and SATA storage systems in the price class below 25,000 Euros, as the storage services of the more expensive mid-range storage systems are no longer required. needed. Owing to the short product lifecycles in the IT industry, extending a support contract and upgrading an existing mid-range disk array is usually a costly option - if available at all. Purchasing a new system is sometimes the more cost-effective option. In a virtualised storage landscape in which the essential storage services such as mirroring, snapshots and replication are not resident in the disk array itself, the storage pool can be upgraded by adding a more-or-less random hard disk subsystem without interrupting productive operations. Users are not obliged to choose a specific vendor or product series. Simple performance upgrades: Only rarely are the hard disks the bottleneck. Usually the limiting factors are controllers or the host interfaces of storage subsystems. Unfortunately, the controllers of external RAID storage systems are almost impossible to upgrade in practice. Virtualisation software solutions such as DataCore SANmelody or SANsymphony open up a while new range of possibilities, because they are viable on practically any x86 server. The virtualisation server can simply be upgraded with faster fibre-channel HBAs, more cache or updated boards or processors. Even buying a new server and uploading the existing virtualisation software is less costly than upgrading or b uying new mid-range disk arrays. Anyone relying on network switches, embedded appliances or closed systems for virtualisation would also usually lose a large portion of their investment if they chose to make a new purchase for the purposes of performance upgrading, because the software functions are inseparably linked to the hardware. Storage clusters for even more performance: For even more performance in the SAN, multiple Sample solution for storage virtualisation Heterogeneous application servers (Windows®, Linux, Solaris™, HP-UX, AIX, VMware®, ...) QLogic™ SANbox® 5600Q Stackable FC-Switch mit 8-64 Ports Advantages for the long-term IT strategy and storage administration As stated at the beginning, although virtualisation appliances are in principle a well-known external storage controller technique, they do offer many advantages both for the long-term storage strategy of a company and for the IT administrator in everyday working. This ultimately also explains the steadily rising demand for storage virtualisation solutions. Simple capacity upgrades: The storage system purchased on the last consolidation comes up against its limits after just a few years. More capacity - and usually also more performance - is transtec SAN Storage Server 8000 (Storage Domain Server, SDS) transtec PROVIGO 600 Premium Powerful, scaleable SATA, SAS and FC storage systems High-performance SAN solution for FC and IP networks  Advanced Caching for faster I/O access  Thin Provisioning  (N+1) redundancy for HA and linear scaling SAN performance  Remote Mirroring with automatic fail-over / fail-back GO  Application-specific snapshots  Asynchronous IP replication GOWEB WEB Detailed information under GOweb SAN Virtualisation with DataCore™ 53 Consolidation storage controllers (appliances) can be interconnected to form an n-fold cluster in in-band storage virtualisation. They are therefore not restricted to the traditional controller pair of a mid-range storage subsystem. The EqualLogic Peer Storage (PS) arrays also work according to a related principle. They are able to combine up to 16 PS modules to create a loadbalanced cluster, of which the storage is made available to the application servers across all the modules by way of one or more virtual LUNs. With each PS array, as well as a number of hard disks you also get another controller pair complete with cache and I/O ports, so that the performance of the overall system is scaled almost 90 percent in linear mode. Consequently, despite the iSCSI host interface, EqualLogic PS arrays regularly outperform traditional mid-range FC disk arrays in benchmark tests, which is why EqualLogic has within a few years grown into a market leader in virtualised IP-SAN solution. Avoidance of SAN islands: When using a virtualisation appliance, it makes sense to transfer the storage services from the hard disk arrays to that level. Whether Thin Provisioning, snapshots, synchronous or asynchronous replication or archive functions, such services are almost always compatible only within one vendor‘s products, and even then only within certain product families. Anyone looking to upgrade a high-availability SAN must stay with the initially chosen vendor or create individual vendor islands which will increase complexity and indeed contradict the basic point of a SAN. 0 be allowed in order to avoid the risk of continually having to enlarge LUNs or of disk-full downtimes. Consequently, the maximum useful memory capacity – depending on the dynamics – is around 60 percent. Thin Provisioning offers a remedy to this, and additionally frees the IT department from unproductive routine tasks. In Thin Provisioning, the logical presentation of the volume is completely separate from the capacity of the underlying drives. The applications see a logical drive of maximum capacity (depending on the limits of the host operating system) while the virtualisation software merely assigns the application the currently required physical memory space. In the specific example, Oracle always sees a 2 terabyte database LUN, while only the current 390GB memory requirement has automatically been reserved from the hard disk pool. If the demand increases, the memory assignment is automatically and dynamically adapted. When certain limits are reached for the capacity of the global memory pool the administrator is warned. This “just-in-time“ storage relieves the administrator of routine LUN/memory management tasks and enables more efficient use of available memory resources. On changing over to Thin Provisioning, between a fifth and a third of available capacity is often released for use all at once. Fail-safety: Remote Mirroring is a bonus feature of mid-range storage systems - see also “Fully redundant storage“ and “Asynchronous replication“. Unfortunately, it does not incorporate automatic fail-over (transfer of services) in the event of failure of the primary storage system or the primary LUN. The applications must first be switched either manually or by scripting to the mirrored LUNs of the secondary system before work continues, which is irritating and unnecessary. There are already solutions on the market, such as the transtec SAN Storage Server based on DataCore virtualisation software, which feature genuine automatic fail-over/fail-back in their Remote Mirroring functionality. If a SAN Storage Server experiences a downtime, the applications are transparently provided with access to the secondary storage pool without interruption. Storage pooling: With a virtualisation appliance you can create a virtual LUN comprising all existing USERS 10000 8000 6000 64 DISKS 64 DISKS 76 DISKS 80 DISKS 146GBx15Krpm 146GBx15Krpm 146GBx15Krpm 73GBx15Krpm HP EMC²® NetApp® EVA 8000 Fibre CX3-20 Fibre FAS 980 Fibre EQUALLOGIC® PS3800XV iSCSI EqualLogic® IP-SAN: Supports > 25% more Microsoft® Exchange users with comparable number of HDDs than comparable FC solutions Microsoft Corp: http://www.microsoft.com/technet/ prodtechnol/exchange/2003/esrp.mspx If the storage services are separated from the physical storage device however, users can deploy any desired storage subsystem compatible with the virtualisation appliance. The services run as usual without interruption, regardless of which memory arrays are currently connected to the bottom. Since the storage systems then need to have less “intelligence“, much more cost-effective models than the traditional mid-range disk arrays can be used. In this respect the high-performance RAID systems from the transtec PROVIGO 600 Premium series are particularly to be recommended. Automation based on Thin Provisioning: In the traditional assignment of LUNs to the individual applications, a certain capacity reserve must always 54 physical storage subsystems. This is only desirable in very rare cases however. The Storage Pooling function offered by transtec SAN Storage Server therefore creates abstract storage classes of which the virtual LUNs save their data only to specific physical storage systems or memory areas. They thereby ensure that, even in a virtualized environment, the data of business-critical applications is stored only on the hard disks of the securest and fastest storage subsystems. Non-critical files are saved to available low-cost storage. This ultimately represents a tieredstorage implementation covering all physical storage systems at minimal commitment. Convenient switching to storage networks: The switch from a direct-attached configuration to a storage network (SAN) is often not easy. Several hundred gigabytes or terabytes of data must be relocated to the new storage solution and the server services tested with new solution. For most IT administrators this entails weekend and evening shifts, because of the downtimes. Storage virtualisation thus offers a major simplification. A component of the transtec SAN Storage Server is the DataCore SANmotion software. This enables you to migrate data in live operation from the local hard disks to the SAN storage, test it there at your leisure and then disable the local storage; and all with no maintenance downtimes interrupting operations. You could even continue to use the capacity of the internal server hard disks or DAS storage systems permanently or temporarily for your virtual storage pools, though this is an option which is rarely taken up in practice. In the field of virtualisation with storage clusters, the EqualLogic Peer Storage solutions we offer feature an extraordinarily high degree of automation, making them highly user-friendly. The induction time in this regard is just a few hours, aided by the deployment of familiar IP technology as the interconnect. Virtual solutions based on intelligent IP switches feature less performance and automation than the above transtec/EqualLogic storage solutions, but they are lower in price and likewise quite easy for experienced users to administer. At the upper end of the spectrum are the transtec SAN Storage Servers for FC and IP SANs based on the virtualisation software from DataCore, offering the widest range of features but also involving the most induction time. transtec will assist you throughout the process of introducing a storage virtualisation solution, from initial conception, through implementation, to expert after-sales-support from an experienced team allied to a comprehensive package of ancillary services. EqualLogic® Peer Storage Cluster Construction and operation of a SAN does not have to be complex and costly The EqualLogic® PS series uses standard IP networks, automates many day-to-day administrative tasks and provides your application servers with virtual storage pools. Your high-availability SAN will be installed and ready to run within an hour. This makes EqualLogic® the ideal storage solution for a VMware® installation or server consolidation. Features in focus  Redundant controllers with dual-core 64 Bit RISC CPU  16 SAS or SATA HDDs  Clusterable with up to 16 PS modules  Dynamic load balancing across all modules  3x1 Gbit Ethernet (iSCSI) ports with MPIO support  Thin Provisioning  Volume cloning, snapshots and replication  VMware® certified and recommended GO WEB Summary Storage virtualisation offers real benefits to IT operations based on increased automation, higher availability and improved growth strategies. This is backed by the positive experience of existing users across all sectors – large concerns, SMEs and public-sector organisations. It does, however, entail a certain induction and accustomisation phase for the virtualisation level. Consequently, transtec offers a series of virtualisation solutions for fibre-channel and IP storage networks in order to find the optimum mix of features, price and complexity for the individual customer based on a consultation session. GOWEB Detailed information under GOweb EqualLogic 55 Just imagine: a stunningly eye-catching design. If you have an eye for design, the latest SyncMaster monitors from Samsung will be your new passion. As eye-catching works of art, these high-performance monitors feature an innovative pedestal design concept and impress with their elegant look. Let yourself be inspired by their charms. With Samsung monitors anything is possible. imagine www.samsung.de SyncMaster 2032BW SyncMaster 2232BW · Response time: 2 ms (G/G · Dynamic contrast value: 3000:1 · Innovative pedestal design concept · iF Product Design Award Winner · iF Material Award Winner · Windows Vista ™ Premium certified 56