Protection Mechanisms for Optical WDM Networks based on Wavelength Converter Multiplexing and Backup Path Relocation Techniques

1
Protection Mechanisms for Optical WDM
Networksbased on Wavelength Converter
Multiplexing andBackup Path Relocation
Techniques
Sunil Gowda, Expedia.com, Seattle, WAKrishna M.
Sivalingam, Univ. of Maryland, Baltimore County,
MDIEEE INFOCOM 2003

• Presented by Brian V. Jarvis
• Wright State University CEG790-01
• Winter 2004, 2 February 2004
• bvjarvis_at_woh.rr.com

2

• Introduction
• Background
• Proposed Mechanisms
• Performance Analysis
• Conclusions

3

• Introduction
• Background
• Proposed Mechanisms
• Performance Analysis
• Conclusions

4

• Introduction
• Exponential growth in bandwidth requirements
• This paper considers routing algorithm designs
  for optical WDM networks
• Fault-tolerance (survivability) is an important
  consideration
• Survivability is introduced into an optical
  network
• Three different mechanisms to help improve
  network performance
• Conversion Free Primary Routing (CFPR)
• Converter Multiplexing Technique (CMT)
• Backup Path Relocation Scheme (BPRS)
• Results indicate that the proposed algorithms
• Reduced the required number of wavelength
  converters at each node by 75.
• Reduced the overall blocking probability from 60
  to nearly 100.
• Reduced the number of primary paths undergoing
  conversion.
• Reduced overall network blocking probability by
  up to an order of magnitude.
• Provided higher improvements for low to moderate
  loads.
• All with negligible additional overhead.

5

• Introduction
• Background
• Proposed Mechanisms
• Performance Analysis
• Conclusions

• Present relevant background material on
  wavelength division multiplexing (WDM) networks,
  routing and wavelength assignment (RWA)
  algorithms, protection techniques, and wavelength
  router architectures that incorporate conversion.

6

• Background
• Routing and Wavelength Assignment (RWA) Problem
• A dynamic network model where connection requests
  arrive randomly.
• Determining the end-to-end route and the specific
  wavelength
• Static or dynamic routing scheme approaches
• Static routing scheme
• Dynamic routing scheme (this technique is used in
  this paper)

7

• Background
• Wavelength Router Architecture
• Without wavelength conversion, connections are
  often blocked
• With wavelength conversion, the lightpath can use
  different wavelengths
• High cost, lack of availability, signal
  degradation
• Various optical wavelength conversion techniques
• Minimize the usage of wavelength conversion
• Reduce the number of converters required
• Location of wavelength converters
• Three architectures for a wavelength convertible
  switch
• Dedicated
• Share-per-link
• Share-per-node

8
Dedicated Wavelength Convertible Switch
Architecture

• A wavelength converter is available at each
  output port.
• The incoming optical signal is de-multiplexed
  into separate wavelengths.
• The output signal may have its wavelength
  changed.
• The various wavelengths are multiplexed.

9
Share-per-node Wavelength Convertible Switch
Architecture

• A wavelength converter bank (WCB) is available at
  the optical switch.
• The incoming optical signal is de-multiplexed
  into separate wavelengths.
• Only the wavelengths that require conversion are
  directed to the WCB.
• Converted wavelengths are switched to the
  appropriate outbound fiber link.
• The various wavelengths are multiplexed.

10
Share-per-link Wavelength Convertible Switch
Architecture

• A wavelength converter bank (WCB) is available at
  each outgoing fiber link.
• The incoming optical signal is de-multiplexed
  into separate wavelengths.
• The optical switch is configured to direct
  wavelengths toward a particular link.
• Only the wavelengths that require conversion are
  directed to the WCB.
• The various wavelengths are multiplexed.

11

• Protection in Mesh-Topology WDM Networks
• Link Failures
• Node Failures
• Recovery Mechanisms
• Protection (proactive)
• Backup lightpaths are identified and resources
  are reserved along the backup lightpaths at the
  time of establishing the primary lightpath
  itself.
• Faster recovery
• 100 percent restoration guarantee
• Restoration (reactive)
• When an existing lightpath fails, a search is
  initiated to find a new lightpath which does not
  use the failed components. (After the failure
  happens)
• Longer restoration time
• It can not guarantee successful recovery,

12
(No Transcript)
13

• Lightpath Migration
• Migration of lightpath onto new paths to
  accommodate other connections.
• A virtual topology reconfiguration scheme to
  adapt to the changing traffic patterns
• modeled as an integrated linear programming
  (ILP) formulation
• Lightpaths are torn down and re-established,
  providing better paths for the primary.
• During reconfiguration, transmission on that path
  is terminated.
• Other migration schemes studied. (Not presented
  here.)

14

• Introduction
• Background
• Proposed Mechanisms
• Performance Analysis
• Conclusions

• Presents the network architecture studied and the
  details of three proposed mechanisms.

15

• Network Architecture
• Dynamic routing shortest path is computed
  between nodes based on current situation.
• Path level protection with dedicated and shared
  protection schemes for backup paths.
• Wavelength router architecture is based on
  share-per-node wavelength converter
  configuration.
• Offers the best cost to performance ratio.
• Connections are blocked only if free wavelength
  or wavelength converters are unavailable.
• The design goals of the proposed mechanisms are
  to
• Improve performance
• Increase cost effectiveness
• Proposed mechanisms
• Conversion-Free Primary Routing (CFPR)
• Converter Multiplexing Technique (CMT)
• Backup-Path Relocation (BPR)

16

• Conversion Free Primary Routing (CFPR)
• The goal is to avoid wavelength conversions while
  routing primary connections.
• Multi-layered graph is used, the layers
  representing individual wavelength planes.
• CFPR algorithm models such a graph.
• For each wavelength plane, the nodes are the
  physical nodes.

2
4
Wavelength 0
1
3
5
2
4
Backup path b
1
Wavelength 1
3
5
2
4
1
Wavelength 2
3
5
Blocked route
Unoccupied wavelengths// Wavelengths occupied
by Backup paths
Wavelengths occupied by Primary paths
17

• Conversion Free Primary Routing (CFPR)
• An overlapping segment is defined as the part of
  the backup path occupying the wavelength assigned
  to the requested connection.
• Relocation schemes may be used to relocate these
  overlapping segments.

• Advantages of CFPR
• Reduces conversion delays and degradation due to
  converters
• Reduces costs
• Lower computational complexity
• Computes path on each wavelength separately
  alternate paths are available if shorted paths
  are blocked

Potential primary path with some component
links overlapping existing backup paths
18

• Converter Multiplexing
• Based on backup path multiplexing.
• Allows wavelength converters to be shared among
  multiple backup paths
• The converters are reserved during the
  establishment of the backup paths.
• Objective is to
• Reduce the number of connections blocked
• Reduce the numbers of converters in use.

19

• Converter Multiplexing
• The node returns a CONV-RESV-ACK message if the
  request is accepted.
• The node returns a CONV-RESV-NACK message if the
  request is not accepted.
• A wavelength conversion status table (WCST) is
  maintained.
• When a network fails, a CONV-SETUP message is
  sent to the node.

Converter Multiplexing Between Paths b1 and b2
20
Wavelength Converter Status Table

• Converter Multiplexing
• Node checks the WCST to select a wavelength
  converter.
• The appropriate entry is inserted into the WCST
• Wavelength converter identifier
• Connection (session) identifier
• Incoming port associated wavelength
• Outgoing port associated wavelength
• Path status (free, primary, reserved)

21

• Backup Path Relocation (BPR)
• Used when it becomes necessary for primary paths
  to accommodate certain routes which are occupied
  by the backup path.
• Objective is to
• Help in providing primary paths with fewer hops.
• Reduce blocking.
• Improve network utilization.
• Two relocation schemes are used to migrate an
  overlapping backup segment.
• Wavelength Relocation (WR).
• Segment Relocation (SR).
• Backup paths are relocated only if all
  overlapping segments can be relocated.
• If relocation fails, none of the overlapping
  segments are relocated.

22
(Network states before relocation)
(Network states after relocation)
Examples of backup path relocation mechanisms.
23

• Introduction
• Background
• Proposed Mechanisms
• Performance Analysis
• Conclusions

• Presents the performance analysis of the proposed
  techniques, based on a discrete-event simulation
  model.

24

• Simulation Model
• A dynamic network traffic model
• Connection requests arrive at a node
• Each connection request is assignment a
  wavelength
• Traffic load is defined in Erlangs.
• Share-per-node architecture is used all nodes
  allocate an equal number of converters.
• Both the dedicated and shared protection schemes
  are studied.
• Simulations are performed for two networks
• A 24-node ARPANET-like network with 16 and 32
  wavelength on each link. The results for this
  network is discussed.
• Also, a random 50-node network with 32 wavelength
  per link.

25

• Comparison of mechanisms
• Basic hop-count (HC) based shortest path routing
  algorithm,
• CFPR routing algorithm with wavelength
  relocation, and
• CFPR routing algorithm with segment relocation.
• The notation X-Y-Z is used to specify an
  algorithm, where
• X HC,CFPR denotes the routing algorithm,
• Y NR,WR,SR denotes no relocation,
  wavelength relocation and segment relocation
  respectively, and
• Z DP,SP denotes dedicated and shared
  protection, respectively.
• The performance metrics presented are the
• blocking probability (Pb),
• link and converter utilization,
• average hop count, and
• backup path relocation statistics

26

• Blocking Probability
• Network blocking probability attempt to minimize
  this metric
• Architectures employing converter multiplexing
  and backup path relocation schemes perform
  significantly better than the basic scheme.
• CFPR with WR/SR has lower Pb than that of the
  basic scheme.

27
Reduction in Number of Converters.

• Graph demonstrates the reduction in the number of
  converters required per node.
• For an offered load of 3.0 Erlangs, the basic
  architecture needs a minimum of 16 converters to
  offer a blocking probability of less than10-1.
• The new techniques offer the same performance for
  as few as 4 converters.
• We can also observe that having more than 8
  converters, does not lower the blocking
  probability any further.

28

• Reasons for blocked connections
• Converter unavailability accounts for
• 2 - 3 for the CFPR-based techniques
  w/dedicated protection
• 20 for the CFPR-based techniques w/shared
  protection
• 99 for the basic scheme w/dedicated protection
• 100 for the basic scheme w/shared protection

Percentages of Connections Blocked Due
to Wavelength Unavailability and Converter
Unavailability.
29

• Average Hop Count
• Graph shows the average hop count of the primary
  paths for accepted connections.
• Basic scheme exhausts all the converters as the
  network load increases.
• In comparison, the converter multiplexing based
  algorithms have a steady hop count.

hop the number of links between a pair of
nodes.
30

• Revenue Metric
• Metric based on the number of hops routed.
• Defined as the shortesthop count based on the
  static topology.
• For the basic scheme the revenue drops when load
  increases
• The proposed algorithms, show only a marginal
  drop in revenue.

31

• Conversion Statistics
• One objective was to provide wavelength-conversio
  n free paths for the primary paths.
• Around 30 of the basic routing scheme
  connections need at least one wavelength
  converter.
• The proposed algorithm eliminates the need for
  wavelength conversion.

32

• Relocation Statistics
• Performance of systems with and without backup
  path relocation schemes.
• When relocation was deployed, the blocking
  probability was lower.
• More significant benefits were seen in the
  context of wavelength conversion.
• Results indicate that the backup relocation
  overhead is manageable, and only a reasonable
  number of relocations are necessary.

33

• Introduction
• Background
• Proposed Mechanisms
• Performance Analysis
• Conclusions

• Wraps up and concludes the paper.

34

• Conclusion
• Three different mechanisms were proposed and
  analyzed.
• The proposed converter multiplexing scheme
  reduces the number of connections blocked.
• The CFPR routing algorithm significantly reduced
  the number of primary connections undergoing
  wavelength conversion.
• Two different backup path relocation mechanisms
  were presented
• Analysis showed that the combination of the
  mechanisms results in substantial reduction in
  blocking probability.
• Also, lower number of converters were required
  per node to achieve a target blocking
  probability.
• The additional overhead of using segment
  relocation compared to the wavelength relocation
  scheme did not result in much improvement.