GMPLS control plane CTTC by liaoqinmei

VIEWS: 3 PAGES: 35

									GMPLS-based provisioning of CO-
Ethernet connections over WSON
with Quality of Service

 Anica Bukva
 PhD student
 Centre Tecnològic de Telecomunicacions de Catalunya (CTTC)
 Optical Networking Area
Outline

• Introduction
    • Evolution towards multi-layer networks: CO Ethernet over
      WSON
• Unified GMPLS control plane for multi-layer networks
    • GMPLS control plane concept: Forwarding adjacency
    • RSVP-TE signaling in multi-layer networks
    • Path computation in multi-layer networks
• Proposed path computation algorithm
• Evaluation of the algorithm and numerical results
    • Part I: Comparing the proposed algorithm with the current
      solution
    • Part II: Three approaches for VNT configuration
• Conclusion




 Weekly seminar, Anica Bukva, 1st June 2011                       2
Outline

• Introduction
    • Evolution towards multi-layer networks: CO Ethernet over
      WSON
• Unified GMPLS control plane for multi-layer networks
    • GMPLS control plane concept: Forwarding adjacency
    • RSVP-TE signaling in multi-layer networks
    • Path computation in multi-layer networks
• Proposed path computation algorithm
• Evaluation of the algorithm and numerical results
    • Part I: Comparing the proposed algorithm with the current
      solution
    • Part II: Three approaches for VNT configuration
• Conclusion




 Weekly seminar, Anica Bukva, 1st June 2011                       3
Introduction: Existing problems in
backbone transport networks
• Need to satisfy the explosive growth of the data
  traffic
    • High definition IPTV, video-conferences, etc
    • WDM technology
• High-capacity, flexible access
    • Packet transport networks – Connection-Oriented
      Ethernet
• Automatic control mechanisms and intelligence
  to deal with the dynamic service provisioning
    • Quality of service
    • High level of reliability



 Weekly seminar, Anica Bukva, 1st June 2011             4
Wavelength Division Multiplexing

• Technique of placing multiple optical signal
  within the same fiber
    • Coarse WDM
    • Dense WDM
• WDM switches:
    • OEO (opto-electronic)
    • OOO (all optical switch)
    • Tranlusent switches
• Wavelength Switched Optical Networks –
  transport networks capable of switching optical
  signals within the optical domain

 Weekly seminar, Anica Bukva, 1st June 2011      5
Connection-oriented Packet Transport
Technologies for Ethernet Service
Delivery
• Need for interoperability with access and
  aggregation
• Ethernet – the most deployed technology in LAN
• Standard Ethernet is not adapted for MAN and
  WAN environments
• Major standard organizations are involved in the
  expansion of Ethernet
• The two main Connection-oriented Packet
  Transport Technologies:
    •     PBT
    •     MPLS-TP
Weekly seminar, Anica Bukva, 1st June 2011     6
Outline

• Introduction
    • Evolution towards multi-layer networks: CO Ethernet over
      WSON
• Unified GMPLS control plane for multi-layer networks
    • GMPLS control plane concept: Forwarding adjacency
    • RSVP-TE signaling in multi-layer networks
    • Path computation in multi-layer networks
• Proposed path computation algorithm
• Evaluation of the algorithm and numerical results
    • Part I: Comparing the proposed algorithm with the current
      solution
    • Part II: Three approaches for VNT configuration
• Conclusion




 Weekly seminar, Anica Bukva, 1st June 2011                       7
     GMPLS control plane: Origins

     • Idea: To have dynamic and advanced control for transport
       networks
     • Origins of GMPLS->MPLS
         •      In a Packet switched networks MPLS uses a label for each data
                packet so it can be forwarded through the network
         •      Switching Capability
         •      Lambda Switched Path
         •      The same concept was extended/generalized for other networks (in
                a WDM network wavelengths are switched, in TDM-timeslots are
                switched, in , etc…)->Generalized MPLS which can handle multiple
                switching capabilities (PSC, LSC, TDM…)



IP        IP   15     IP   10   IP




      Weekly seminar, Anica Bukva, 1st June 2011                             8
GMPLS control plane: Definition

• Network (a node) can be devided into set of
  functional components-planes:
    • Data plane
    • Control plane
    • Management plane
• Control plane
    • Infrastructure and distributed intelligence which
      includes all operations, mechanisms and protocols
      needed to establish, manage and release
      connection between the different network elements




 Weekly seminar, Anica Bukva, 1st June 2011          9
GMPLS control plane: Protocols

• GMPLS control plane:
    • Link Management Protocol – LMP
    • Routing protocol (e.g., OSPF-TE)
    • Signaling protocol: RSVP-TE
• LMP responsible for discovery of link capabilities and
  its identifiers
• OSPF-TE: Distribution of the information that will be
  used for path computation
    • TE links (link type, ID, local interface address, TE metric,
      ISCD, etc…)
    • TE database (information needed for path computation)




 Weekly seminar, Anica Bukva, 1st June 2011                          10
GMPLS control plane: Protocols

• Path computation
    • Source based (node that receives the request)
    • Centralized (e.g., PCE)
    • The most used path computation algorithm – Shortest
      path algorithm
• RSVP-TE: Responsible for establishing of
  connection (LSP)
    • Exchanging signaling messages composed of objects
      and sub-objects
    • Path, Resv, PathTear, PathErr, ResvTear, etc…
    • A path inserted into an ERO object



 Weekly seminar, Anica Bukva, 1st June 2011            11
GMPLS control plane: Architectural
models

• Architectural models:
    • Peer (unified) model
    • Overlay model
    • Hybrid model




                                              Overlay Model
             Peer Model



 Weekly seminar, Anica Bukva, 1st June 2011                   12
Unified GMPLS control plane:
Advantages
• Having a multi-layer network, a single control
  plane instance may be used to control all the
  switching layers
    •   Lower control complexity
    •   Better scalability
    •   Single addressing
    •   Efficient use of the network resources
• A single Traffic Engineering Database
  consolidates network resource information of all
  the involved layers




 Weekly seminar, Anica Bukva, 1st June 2011        13
GMPLS control plane concept:
Forwarding Adjacency
• In WSON network a label = a wavelength,
  therefore, a whole wavelength is reserved even
  if it was required a fraction
• Problem resolved by Forwarding Adjacency
• Control plane concept
  where the lower layer
  LSP is used for forwarding
  the upper-layer LSP
• Forming Virtual Network Topology



 Weekly seminar, Anica Bukva, 1st June 2011   14
Virtual Network Topology

• FA TE links
    • The recources are reserved
• Virtual links
    • The lower-layer LSPs are computed but not signalled
      in the network
    • The resources will be not cross-connected when there
      is a request from the upper-layer


• Improving scalability since multiple upper-layer
  can be nested over a single FA-LSP (grooming)



 Weekly seminar, Anica Bukva, 1st June 2011            15
RSVP-TE signaling for multi-layer
networks
•     RSVP-TE supports establishment of FA TE links and the signaling in
      a multi-layer networks

                                      L2SC                                               L2SC TE Link                                        L2SC
                                                                                                                                                               L2SC L2SC
                  L2SC L2SC                                                                                                                      L2SC
                                     L2SC
                                             LSC                                                                                           LSC
       10.1.0.1           10.1.0.2                            LSC                  LSC                LSC                 LSC                           10.1.0.3           10.1.0.4
             Path
             Session:                              10.2.0.1                                10.2.0.2                             10.2.0.3
             10.1.0.4           Path                            Path                                        Path                      Path
             SwCap: L2SC        Session: 10.1.0.3               Session:                                    Session:                  Session:
                                SwCap: LSC                      10.1.0.3
                                LTII:10.1.0.2/ a                                                            10.1.0.3                  10.1.0.3
                                                                SwCap: LSC                                  SwCap: LSC                SwCap: LSC
                                                                LTII:10.1.0.2/ a                            LTII:10.1.0.2/ a          LTII:10.1.0.2/a


                                                                                                                                      Resv
                                                                Resv                                        Resv                      LTII:10.1.0.3/b
                                Resv                                                                        LTII:10.1.0.3/ b
                                LTII:10.1.0.3/ b                LTII:10.1.0.3/ b

                                                                       LSC FA-LSP = L2SC TE LINK

                                                                Path                                                                                          Path
                                                                Session: 10.1.0.4                                                                             Session:
                                                                                                                                                              10.1.0.4
                                                                SwCap: L2SC
                                                                                                                                                              SwCap: L2SC


                                                                Resv                                                                                          Resv
            Resv


                                                                                          L2SC LSP




    Weekly seminar, Anica Bukva, 1st June 2011                                                                                                                                        16
Virtual Network Configuration

• Two alternatives to establish LSPs that induce
  FA creation
    • Static: pre-computed without respect to the traffic demands
    • Dynamic: computed on-line according to the traffic demands
• Reconfiguration of Virtual Network Topology
  may be triggered by:
    •   Traffic demand changes
    •   Topology configuration changes
    •   Signaling request from the upper layer
    •   Network failures




 Weekly seminar, Anica Bukva, 1st June 2011                   17
Some of the open issues in a MLN

• Path computation in a MLN
    • Take advantage FA TE links and virtual links
    • Pick links that provide desired Quality of Service
    • Shortest path algorithm - mostly used for path
      computation: It takes into account only cost/TE metric
      of a link
    • Other contrains should be taken into account
      (avalability of one link, WCC, WC usage, etc…)
• Finding an optimal solution in choosing between
  pre-provisioned and triggered signaling,
  regarding network performance and quality of
  service

 Weekly seminar, Anica Bukva, 1st June 2011               18
Outline

• Introduction
    • Evolution towards multi-layer networks: CO Ethernet over
      WSON
• Unified GMPLS control plane for multi-layer networks
    • GMPLS control plane concept: Forwarding adjacency
    • RSVP-TE signaling in multi-layer networks
    • Path computation in multi-layer networks
• Proposed path computation algorithm
• Evaluation of the algorithm and numerical results
    • Part I: Comparing the proposed algorithm with the current
      solution
    • Part II: Three approaches for VNT configuration
• Conclusion




 Weekly seminar, Anica Bukva, 1st June 2011                       19
A dynamic path computation
algorithm for MLN
• Shortest path - based algorithm
• Source - based
    • It firstly checks if there are direct FA TE link with
      enough bandwidth that connects the source and the
      destination
    • If not, it checks if there is a direct computed but not
      established virtual link between the source and the
      destination
    • If there are neither direct FA TE link nor virtual links, a
      path computation algorithm will be computed
                A                                                               D
                                          FA TE link = 3

                    L2SC   1
                                     b    Virtual link = 4 c         1       L2SC
                               LSC       LSC        LSC        LSC
                       1                       1                         1

                                          New LSC LSP = 5




 Weekly seminar, Anica Bukva, 1st June 2011                                         20
A dynamic path computation
algorithm for MLN
•     Total cost:

           C( )   aai   ( bbij  ccij  dd ij  eeij  ff ij )
                     i        ij

       •      The a, b, c, d, e and f coefficients are defined as zero or any positive
              real number
       •      a: Ratio between the number of the used WCs and the number of the
              total WCs equipped at a WSON node
       •      b: Current traffic load in the optical link
       •      c: Parameter which is set to zero if both ends of a TE link support the
              same switching capability (e.g., LSC)
       •      d: Traffic load in a FA TE link
       •      e: Number of underlying optical links of a FA TE link. If the link is a virtual
              link that is still not established, than the cost is increased with “+1”
       •      f: The parameter which is set to “1” and it is added to the cost of each
              optical link


    Weekly seminar, Anica Bukva, 1st June 2011                                          21
Pseudo-code
  INPUT
           G, V, E, FA, VL, E’, s, d, req_band
  OUTPUT
          P
  Shortest path based algorithm
          for i | i  V
               dist[i] ← INFINITY
               pred[i] ← NULL
          dist[s] ← 0
          Q ← s
          While Q is not empty
               u ← extract min dist (Q)
               for e,v | e=(u, v)  E’
                  if unreserved_band(e)<req_band
                      continue
                  if e  E
                      if WCC is not satisfied and WC(u)==0
                           continue
                  new_dist ← dist[u] + C(uv)
                  if new_dist < dist(v)
                      dist[v] ← new_dist
                      pred[v] ← u
                      Q ← v
          end while

           u ← d
           while pred[u] is not NULL
                insert u at the beginning of P
                u ← pred[u]
           end while
           return P




Weekly seminar, Anica Bukva, 1st June 2011                   22
Outline

• Introduction
    • Evolution towards multi-layer networks: CO Ethernet over
      WSON
• Unified GMPLS control plane for multi-layer networks
    • GMPLS control plane concept: Forwarding adjacency
    • RSVP-TE signaling in multi-layer networks
    • Path computation in multi-layer networks
• Proposed path computation algorithm
• Evaluation of the algorithm and numerical results
    • Part I: Comparing the proposed algorithm with other shortest
      path-based algorithm
    • Part II: Three approaches for VNT configuration
• Conclusion




 Weekly seminar, Anica Bukva, 1st June 2011                      23
Evaluation of the algorithm

•     Part I:
       •   The proposed algorithm is compared with a basic shortest path
           algorithm in which are taken into account only:
             • Number of hops
             • The cost of every FA TE link is equal to number of underlying optical
               links. That is case in which:
                a = b = d = 0 and c = e = f = 1
•     Part II
       •   Algorithm is used to compare three different approaches for VNT
           configuration:
             • Semidynamic
             • Virtual
             • Dynamic




    Weekly seminar, Anica Bukva, 1st June 2011                                         24
Network topology

• Framework for the simulation of RSVP-TE
  protocol in a dual-layer network infrastructure
  is made in OPNET Modeler




 Weekly seminar, Anica Bukva, 1st June 2011     25
Model assumptions & Evaluation of
the network performance
• Unified GMPLS control plane: TED (network
  topology and resources) is known
• Four RSVP-TE signaling messages have been
  implemented:
    • Path, Resv, PathErr and PathTear
• Used simplified RSVP-TE objects
• No RSVP-TE protocol refreshes
• The only source of delay is the link propagation
  delay
• The signaling message transmission and
  processing delays are neglected


 Weekly seminar, Anica Bukva, 1st June 2011     26
Model assumptions & Evaluation of
the network performance
• Every node in the WSON area is equipped with two
  wavelength converters
• There are ten wavelength per link in the optical area
• L2SC LSP requests are generated randomly
  following a Poisson distribution with the average
  inter-arrival time of 1s
• The LSP holding time is exponentially distributed
  and varies in different simulation sets to evaluate the
  network behavior under various traffic loads
• Control links have the propagation delay of 3.5ms
• Each data point is obtained simulating 7.5x105 L2SC
  LSP requests

 Weekly seminar, Anica Bukva, 1st June 2011           27
Numerical results: Part I

• Performance metrics: Blocking probability

                              1
                                   700   900     1100               1300   1500

                             0.1
  Blocking probability




                            0.01


                           0.001


                          0.0001
                                                                            Policy 1
                                                                            Policy 2
                         0.00001
                                               Traffic load (Erl)




 Weekly seminar, Anica Bukva, 1st June 2011                                            28
Numerical results: Part I
 • Performance metrics: Usage of the WC


               1.2
                  700   800   900   1000   1100        1200     1300   1400   1500    1600
                1


               0.8
    WC usage




               0.6


               0.4


               0.2                                                             Policy 1
                                                                               Policy 2
                0
                                           Traffic load (Erl)




Weekly seminar, Anica Bukva, 1st June 2011                                                   29
Numerical results: Part II

• Performance metrics: Blocking probablity
                                      600   800   1000           1200   1400           1600
                                 1



                                0.1
     Blocking probability




                               0.01



                              0.001



                             0.0001                                     Semi-dynamic
                                                                        Virtual
                                                                        Dynamic
                            0.00001
                                                  Traffic load (Erl)



 Weekly seminar, Anica Bukva, 1st June 2011                                                   30
Numerical results: Part II

• Performance metrics: Setup delay
                             600   800   1000          1200     1400        1600
                        35


                        34


                        33
     Setup delay (ms)




                        32


                        31


                        30

                                                                 Semi-dynamic
                        29
                                                                 Virtual
                                                                 Dynamic
                        28

                                           Traffic load (Erl)




 Weekly seminar, Anica Bukva, 1st June 2011                                        31
Numerical results: Conclusion

• Part I:
    • Proposed path computation gives better performance
      in the sense of connection blocking probability
    • Better resource usage
• Part II:
    • The dynamic scenario is able to better utilize the
      resources by setting up the FA when required, but at
      the price of increasing the setup delay
    • The setup delay is the lowest with the fully static
      scenario, since there are no new established LSPs,
      but we are always using the existing ones if there are
      available resources


 Weekly seminar, Anica Bukva, 1st June 2011               32
Outline

• Introduction
    • Evolution towards multi-layer networks: CO Ethernet over
      WSON
• Unified GMPLS control plane for multi-layer networks
    • GMPLS control plane concept: Forwarding adjacency
    • RSVP-TE signaling in multi-layer networks
    • Path computation in multi-layer networks
• Proposed path computation algorithm
• Evaluation of the algorithm and numerical results
    • Part I: Comparing the proposed algorithm with the current
      solution
    • Part II: Three approaches for VNT configuration
• Conclusion




 Weekly seminar, Anica Bukva, 1st June 2011                       33
Conclusion

• Overview of the backbone transport networks
    • Focusing on Ethernet over WSON
• Unified GMPLS control plane
    • Concepts of Forwarding adjacency and Virtual
      Network Topology configuration
• Constrained path computation algorithm in a
  MLN
    • Part I: Compared with the standard shortest path
      algorithm
    • Part II: Three approached for VNT configuration




 Weekly seminar, Anica Bukva, 1st June 2011              34
Thanks for your kind attention!

• Questions?




Anica Bukva
PhD student
anica.bukva@cttc.es




                                  35

								
To top