Survivable Optical Mesh Networks

Transcription

1 Resilient Mesh Networks (Protection and Restoration) Biswanath Mukherjee Professor of Computer Science, UC Davis Acknowledgement: UC Davis Graduate Students and Alums: Dr. S. Ramu Ramamurthy, Dr. Laxman H. Sahasrabuddhe, Dr. Hui Zang, Dr. Wushao Wen, Dr. Jian Wang, Dr. Keyao Zhu, Dr. Canhui (Sam) Ou, Jing Zhang, Yurong (Grace) Huang, et al. Research Sponsors: NSF, Alcatel, Bellsouth, Cisco, Sprint, UC MICRO Page Survivable Optical Mesh Networks (Protection and Restoration) References:. J. Zhang and B. Mukherjee, Review of Fault Management in WDM Mesh Networks: Basic Concepts and Research Challenges, IEEE Network, Special Issue on "Protection, Restoration, and Disaster Recovery", vol. 8, no., pp. -8, March 00 (for general review of the subject).. S. Ramamurthy, L. Sahasrabuddhe, and B. Mukherjee, Survivable optical WDM networks, IEEE/OSA Journal of Lightwave Technology, vol., no., pp , April 003 (for a network-design problem; static traffic). 3. J. Wang, L. Sahasrabuddhe, and B. Mukherjee, Path vs. subpath vs. link restoration for fault management in IP-over-WDM networks: performance comparisons using GMPLS control signaling, IEEE Communications Magazine, vol. 0, no., pp , Nov. 00 (for a traffic-engg. problem; dynamic traffic). Page

2 NSFNet: A Wide-Area Backbone Network WA MI NY CA UT CO PA NJ NE IL MD CA TX GA Page 3 Why Survivable Provisioning? IP Router ADM A Fiber cut OXC Port S Z Duct D FIT (Failure In Time): # of failures in 0 9 hours Fiber:,000 FIT per 0 km of fiber; Mux, coupler: FIT OXC: control & switch fabric + protected (so node failure is rare) Single-cable cut: 8 fibers/cable * 0λ/fiber * OC9/λ.38 Pbps data-loss rate Page

3 Call Setup Conn ID IN port,,0 Local, R OUT port,, λλ? RED STATE UP Reserved TYPE Primary 3 Conn ID IN port OUT port STATE TYPE,,0,, λλ? RED Local, R Reserved UP Primary LC Req: LC Resp: ACK LC LC Req Req Mapping: Mapping: λ RED RED LC Req Mapping: λ RED LC LC Req Req RED Mapping: Mapping: λ RED LC Req LC Resp: : ACK Conn ID IN port OUT port STATE TYPE,,0,, λ RED?,, λλ? RED Reserved UP Primary Conn ID IN port OUT port STATE TYPE,,0,, λλ? RED,, λλ? RED Reserved UP Primary Page Call Setup (with Protection) Conn,,0 IN L, R,,0 L, R OUT, λ RED STATE UP, λ BLUE UP TYPE Primary L, R Backup Lightpath (OXC to OXC) Service path (ED to ED) Backup Lightpath Backup Service path 3 Control Channel Page 3

4 Fault Detection & Recovery 3 Fiber Cut Occurs!... Fault Detected by! 3 Alarm: Link Broken! Switch to Backup!! 3 Switch to Backup!! Connection Rerouted/ Restored! Page 7 Illustrative Example - Path Protection,00 9,000,00 90,00,000,900,00,00 900,300 00, ,000,000 80,000,0 7 9,000, ,000,000, , ,00, , Overall Fiber Distance =,300 Km Page 8

5 Path Protection Failure Recovery,00 90,000,000,900,00,00 900,300,00 00, ,00 3,000,000 0 alarm 7 80,000,0,00 8 alarm ,000 alarm, ,000 alarm 900,000, ,00 900, ms 0 ms Failure occurrence Failure detection Failure notification Page 9 Path Protection Failure Recovery (contd.),00 9,000,00 90,00,000,900,00,00 900,300 00, ,000,000 0 alarm 7 80,000,0 alarm 9,000 alarm, ,000 alarm, , ,00, , Failure occurrence 0 ms Failure detection 0 ms Failure notification 3. ms Failure recovery 7. ms > 0 ms Page 0

6 Shared-Path Protection Shared Page Typical Assumptions Single fiber cut Predominant form of failures in telecom networks Failure fixed before next failure occurs Advanced approaches: accommodate multiple failures Primary and backup paths should be link disjoint Node disjointedness not necessary Since OXC s switch fabric and control unit are + (Master/Slave) protected Page

7 Algorithms for Finding Link-disjoint Paths Two-step approach Step I: compute the shortest path (P) in the given topology Step II: remove the wavelength links on P; compute the shortest path in the remaining graph Sub-optimal because of no backtracking One-step approach Joint computation of two link-disjoint paths Optimal solution and no more computation time consumption Referred to as Suurballe s algorithm or Bhandari s algorithm in the literature Page 3 Example Where The Two-Step Approach Fails Step I: compute the shortest path (P) in the given topology Unidirectional links - Cost of each link is given - One wavelength per link - Connection request (0, ) Page 7

8 Example Where The Two-Step Approach Fails Step II: remove the links on P; compute the shortest path in the remaining graph Unidirectional links - Cost of each link is given - One wavelength per link - Connection request (0, ) Page Example Where The Two-Step Approach Fails Unidirectional links - Cost of each link is given - One wavelength per link - Connection request (0, ) Page 8

9 One-Step Algorithm Step I: compute the shortest path (P) in the given topology Unidirectional links - Cost of each link is given - One wavelength per link - Connection request (0, ) Page 7 One-Step Algorithm Step II: reverse the links along P and negate their costs Unidirectional links - Cost of each link is given - One wavelength per link - Connection request (0, ) Page 8 9

10 One-Step Algorithm Step III: compute the shortest path (P ) in this graph Unidirectional links - Cost of each link is given - One wavelength per link - Connection request (0, ) Page 9 One-Step Algorithm Step IV: remove the links used by P and P Unidirectional links - Cost of each link is given - One wavelength per link - Connection request (0, ) Page 0 0

11 One-Step Algorithm Step V: reverse the links shown by green lines and make their cost positive Note, we have two paths now: (0,,, ) and (0, 3,, ) Unidirectional links - Cost of each link is given - One wavelength per link - Connection request (0, ) Page Two-Step vs. One-Step Route Computation Example Please find a pair of shortest paths between source node A and destination node C By applying the two algorithms Which is better? Page

12 Protection vs. Restoration Protection Reserve backup resources in advance while setting up primary resources Guaranteed recovery from designed failures, e.g., single fiber cuts (dominant failure in optical networks) Fast recovery but resource utilization may be low Restoration Dynamic discovery of backup resources after fault occurs Very good resource utilization (no wastage under normal conditions) No guarantee on recovery Significantly slower recovery than protection Page 3 Multiple Dimensions for Survivability Protection vs. restoration Off-line (protection) vs. on-line algorithms Protection: dedicated vs. shared Both: path vs. sub-path vs. link Page

13 Performance Metrics Control (GMPLS) signaling volume Speed of recovery (or Restoration Time) Computational complexity Blocking performance (for on-line traffic) Page Fault Management Tutorial FIT (Failure in Time): # of failures in 0 9 hrs. Failure frequencies: Fiber:,000 FIT per 0 km of fiber (~ once every.8 yr) Mux, coupler: FIT SONET equipment: 0 0 FIT The 0-ms-recovery-time myth: Persistence of hearing human ear can tolerate 00-ms delay For voice traffic, ok to have 0 ms recovery time For data traffic, minimize recovery time as much as possible Need for fast recovery for data traffic: Fiber cut bundle or conduit cut need fiber-risk-group awareness Example: 8 fiber strands/bundle, 0 bundles/conduit, 0 lambdas/fiber, OC9/lambda Total (maximum) data rate (per conduit) = 3.8 Petabits/sec Up to 3,8 gigabits of data lost per ms of down time Need fast recovery to minimize loss of data (and revenue!) Page 3

14 Fault Management Tutorial Fault-Management Schemes Protection Backup resources (routes and wavelengths) are precomputed and reserved in advance Guaranteed recovery Shorter recovery time Backup resources wasted (unless alloted to preemptable traffic) Suitable for optical layer (with Lambda Routing) Restoration Backup resources are dynamically discovered after failure occurs No guarantee on recovery (backup resources may not be found) Longer recovery time Suitable for Layer 3 (IP packet switching) Ring Protection APS (Automatic( Protection S/w) SHR (Self-Healing Rings) Mesh Protection Page 7 Fault Management Tutorial Mesh Protection Link Protection Path Protection 3 Before Fault Fault Occurs Recovery After Fault 3 Before Fault Fault Occurs Recovery After Fault Resources to back up link link (,3) failure (backup for other links not shown here) Backup (end-to-end) lightpath Inefficient use of backup resources Fast protection-switching time Lot more efficient than L. P. Long routes may require somewhat longer switching time Dedicated L.P. Shared L.P. Dedicated P.P. (+, :) Shared P.P. (M:N) Page 8

15 Fault Management Tutorial + Protection : Protection 3 primary 3 primary backup Both primary and backup are carrying live traffic M:N Protection 7 3 primary share the backup λ s A on link (,) 8 9 primary Multiplexed protection more efficient than : backup Backup activated after failure detected normally, can carry other low-priority preemptable traffic Different categories of recovery more + expensive : X X ms guaranteed recovery time M:N 0: Not preemptable less 0: Preemptable expensive Page 9 Soft Optics Feature Set Network topology discovery (OSPF based) Lightpath provisioning Intelligent routing (MPLS/CR-LDP based) User-selectible routes Conduit identification (for shared risk groups) SONET/SDH configurable interfaces Lightpath monitoring Fault detection and recovery (rerouting around failed links) Differentiated fault-recovery schemes Guaranteed recovery-time service Hitless upgrade of network segments: lightpath rolling Efficient grooming of sub-rate circuits Optical OSPF areas Optical VPNs Network Planning Tool Page 30

16 Lightpath Create (OIF UNI ) Lightpath Bandwidth 0x00: Reserved 0x0: OC-8 0x0: OC-9 0x03: OC-78 0x0: STS- Lightpath Priority (lowest) 8 (Highest) Protection Mode TLV Mode 0x00: Unprotected & not preemptable 0x0: Unprotected & preemptable 0x0: +, end-to to-end + protection 0x03: Shared-path protection Reversion 0x0: Revert to primary path after repair 0x0: Do not revert Retention Mode TLV 0x0: Retain unrestorable primary path 0x0: Delete unrestorable primary path Page 3 TE vs. NE vs. NP Traffic Engineering (TE) Put the traffic where the bandwidth is Network Engineering (NE) Put the bandwidth where the traffic is Network Planning (NP) Put the bandwidth where the traffic is forecasted to be TE online, dynamic, provisioning problem, ms time scale NE intermediate problem, months time scale NP offline, static, dimensioning problem, -yr time scale Page 3

17 Example Network Planning Problem Path Protection Problem Statement (Network Planning) Given: Network topology Static traffic demands Need/Requirements/Constraints: Set up a primary path and a backup path for each demand The two paths must be link disjoint (node disjoint too?) Dedicated or shared backup (based on problem specs) Guaranteed to recover from a single fiber cut Goal: minimize cost, e.g., # of wavelength channels See [Ramu-jlt03] for solution method (ILP) see also [Laxman-jsac0] for WDM protection vs. IP restoration Page 33 Example TE Problem Restoration Problem Statement (Traffic Engineering) Given: Network topology (including # of wavelengths per fiber) Dynamic traffic demands (or connections) Need/Requirements/Constraints: Set up only one path (primary path) for each connection Try to quickly restore the connection when a fault occurs Control signaling (GMPLS?) for connection setup + restoration Note: This method can handle multiple network failures Three restoration methods: path, sub-path, link Performance tradeoffs: availability, restoration time, restoration success rate, See following slides (and [Jian-commag0] for details) Page 3 7

18 Three Restoration Techniques Path restoration Sub-path restoration Link restoration Note: These approaches can easily handle multiple (link/node) failures Pros vs. cons of restoration Page 3 NSF Network Seattle 0 Palo Alto Salt Lake City 3 Boulder Lincoln Ithaca Cambridge 3 Ann Arbor Chicago 9 8 Princeton 7 Pittsburgh Champaign College PK San Diego 0 Atlanta Houston Number of wavelengths per fiber = (one control channel and data channels) Page 3 8

19 Availability vs. Load (in Erlang) Availability = (total uptime for an optical_lsp) / (total duration of the optical_lsp) A network load of Erlangs corresponds to an average link load of about 3.%, while a network load of 00 Erlangs corresponds to an average link load of about 70%. f = 0 = (average failure interarrival time) / (average call holding time) Path Sub-path Link Note: In these experiments, time is normalized to the average call duration, which is assumed to be unity (and equal to 0 seconds); average fault inter-arrival time is denoted by f normalized units; and the average duration of a fiber fault (or fault-repair time) is assumed to be normalized units. The failure-detection time is assumed to be ms. Page 37 Restoration Success Rate vs. Load Restoration success rate = (total number of successfully restored connections) / (total number of disrupted connections) f = Path Sub-path Link Page 38 9

20 Restoration Time vs. Load Restoration time (ms) = from the time a connection is disturbed to the time the connection is restored (unsuccessful restoration effort is not counted) f = Path 0.0 Sub-path 0.0 Link Page 39 Blocking Probability vs. Load Blocking probability = (number of successful connection requests) / (total number of connection request) f = Path Sub-path Link Page 0 0

21 Availability (Path Restoration with Retrials) Availability = (total uptime for an optical_lsp) / (total duration of the optical_lsp ) f = No Retrial One Retrial Tw o Retrials In this experiment, each retrial takes place ms after the alarm message reaches the source node. Recall that path-restoration failure is mainly caused by the contention among restoration requests. We can improve the performance of the path-restoration technique significantly by giving retrial opportunities. Page Restoration Success Rate (Path Restoration with Retrials) Restoration success rate = (total number of successfully restored connections) / (total number of disrupted connections) f = No Retrial 0. One Retrial 0. Tw o Retrials Page

22 Restoration Time (Path Restoration with Retrials) 0. Restoration time (ms) = from the time a connection is disturbed to the time the connection is restored (unsuccessful restoration effort is not counted) f = No Retrial One Retrial 0.0 Tw o Retrials Page 3 Blocking Probability (Path Restoration w/ Retrials) Blocking probability = (number of successful connection requests) / (total number of connection request) f = No Retrial One Retrial Tw o Retrials Page

23 Additional Problems and Current Trends in Survivability Research Availability-aware provisioning [Jing-ofc03] Differentiated (reliability) services Take into account component failure rates and user SLA Detailed link model [Grace-jlt0] New parameter: Link and resource availability (LRA) Backup reprovisioning after fault occurs [Jing-icc0] Sub-path protection [Sam-jsac0] Differentiated Quality of Protection (QoP) [Sam-net0] Guaranteed recovery time Survivable traffic grooming [Sam-jsac03] Note: connection bandwidth may be OC-, OC-3, OC-, OC-8, Survivable Virtual Concatenation (VC) [Sam-ofc0] Ethernet/Data over SONET/SDH over WDM Page Availability-Aware Connection Setup Treated as a Network Planning Problem Next several slides See [Jing-ofc03] for details Page 3

24 Motivation SLA and Provisioning Customer s concerns: Bandwidth Availability Operator s concerns: Resource Protection Fee etc. Penalty Profit = Revenue - Cost Traffic engineering decision is very important Page 7 Differentiated Services: Availability Service Availability Service Type Availability Down Time / Year Basic 99% 87. hours Premium 99.0% 3.8 hours Silver 99.9% 8.7 hours Gold 99.99%. mins Platinum %. mins Leads to differentiated services in mesh networks Customers willing to pay more can get better quality of service Page 8

25 Availability Analysis Rationale and Consideration Rationale Facilitate availability-aware connection setup Provide availability guarantee to customers Enable differentiated services in mesh networks Handle multiple network component failures Factors affecting connection availability Routing Backup resource allocation Backup resource sharing Page 9 Availability of a Network Component Can be calculated based on the component s failure statistics --- MTTF/(MTTF+MTTR) MTTR Failure MTTF Failure 0 t Start of failure End of failure Lifetime of a component MTTF (Mean Time To Failure): average On time. MTTR (Mean Time To Repair): average repair time. Page 0

26 Availability of Network Components Can be calculated based on the component s failure statistics --- MTTF/(MTTF+MTTR) Failure Type Typical Value Equipment MTTR hours Cable Cut MTTR hours Cable Cut Rate.39/year/000 miles Tx failure rate in FIT 087 Rx failure rate in FIT 3 MTTF (Mean Time To Failure): average On time. MTTR (Mean Time To Repair): average repair time. FIT (Failure in Time): # of failures in 0 9 hrs (app., yrs). Page Availability of a Connection Unprotected connection Available when all the network components used are available Dedicated-path protected connection Available when either the primary path or backup path is available Shared-path protected connection Next slide Page

27 Availability for Shared-Path Protection Connection t will be available if: ) Primary is available; or ) Primary is unavailable, backup is available, and t can get the backup resources when the other paths in the contending set have also failed. l l t t t S N p i i δ t : Contending set : Size of S : Prob. that i paths in S unavailable : Prob. for succ. resource contention A t = A p + ( A ) A δ p b N i = 0 i t p i Page 3 Conditional Probability (for Resource Contention) s t t. t K d State-transition diagram for K= while assuming backup available λ = ; µ = MTTF MTTR b π λ π Only determined by the repair rate, not the arrival rate! π λ µ U,D, t µ µ D,D, t U,U,0 δ t π = π + π = µ µ + µ µ λ D,U,t π 3 µ λ µ D,D,t π Page 7

28 General Problem Statement Formulated here as a Network Planning Problem [Jing-ofc03] Given Physical topology Network nodes and links Availability of each component (link) Free wavelengths on each link Static traffic demands (T) Source, destination, and availability requirement of each demand Objective Provision each connection Decide route and protection scheme Minimize overall network cost Wavelength links or wavelength mileage Page Provisioning Strategies Integer Linear Program (ILP) Scheme I: no protection for any connection Scheme II: dedicated protection for all connections Scheme III: dedicated protection for selected connections Can also invent other schemes (e.g., shared protection) Heuristics: fixed-alternate-routing based (K routes K single routes; K link-disjoint route pairs) Minimal-cost Most-reliable Just-above-threshold Iteratively-select (randomly select with backtracking to minimize cost) Page 8

29 Link Model Optical OXC port Node (A) MUX Conduit DMUX λ Amplifier λ Fiber λ n λ n λ λ n Fiber m Optical link Optical channel λ λ n Node (B) LRA port MUX/DMUX fiber amplifier port C λ λ λ,λ C C λ 7 λ C,λ,λ 0 C C λ λ C 3 C 9 C 3 λ 3 λ 3,λ λ 3,λ conduit λ 3,λ λ 3 C C C 8 C C λ λ C Node A Physical-link graph (PLG) model Node B LRA defines the probability that at least one wavelength channel is free and alive in a link. Compute LRA by a recursive procedure Decompose a physical-link graph into simple graphs where LRA can be computed. A simple graph consists of M failure-independent parallel component sets while each set has K failure-independent serial components. See [Grace-jlt0] for details as a TE problem. Page 7 Sub-Path Protection See [Sam-jsac0] for details treated as a Network Planning Problem Page 8 9

30 QoP: Differentiated Protection-Switching Time Problem Statement (TE problem) [Sam-net0] QoP = Quality of protection Could be different for different connections Given a network state A network Primary & backup of existing lightpaths A new lightpath request Objective To compute a primary & backup for the request with customer-desired protection-switching time while minimizing additional resources used (using shared backup path protection) Page 9 Survivable Connection Grooming (and Provisioning) Problem statement (TE Problem) [Sam-jsac03] Given A WDM network Current network state (existing lightpaths/connections) A new connection arrival (of b/w OC-n, n=,3,,8, ) Provision (groom) the connection w/ shared-mesh protection to utilize resources efficiently Approaches Protection-at-lightpath (PAL) level Mixed protection-at-connection (MPAC) level Separate protection-at-connection (SPAC) level Page 0 30

31 A Qualitative Comparison PAL MPAC SPAC End End Protection W & B Traffic Lightpath Connection Connection Separate Mixed Separate Backup Sharing Fixed routing Fixed routing &λ Fixed routing Complexity Low High High Page Modeling Disasters Define SRGs (shared-risk groups) for (potentially) correlated failing components e.g., links + nodes in close geographical proximity SRLG (shared-risk link group) Special case of SRG Quite common since fibers are laid in ducts, and duct cuts are more frequent than we wish Extend link models (or develop new models) to apply to SRGs Page 3

32 Multi-Path Routing p sd p sd Working s i d b p sd Multi-path routing How many paths? State-dependent path capacities Virtual concatenation (VC) ( traffic grooming ) 3 p sd Backup Survivable Routing Primary + backup path(s) Precomputed backup (?) Shared backup (?) Need reliable path (five-9 s) Component/link availability Need fast recovery Page 3 EoS: Service Resilience EoS = Ethernet over SONET (or Ethernet/data over SONET/SDH over WDM) Protect a GbE connection with 00 Mbps paths e.g., peak rate = Gbps, minimum guaranteed rate = 00 Mbps See [Sam-ofc0] for details 00 Mbps Source 00 Mbps 00 Mbps? Utilize route splitting? Take advantage of backup sharing Destination Page 3

33 Which Method? s p sd b p sd i 3 p sd p sd Working d Backup s, p b i p w Backup, p b, p b p w Working d PREV: Provisioning fast REstorable VCG One backup path per node pair PIVM: Protecting Individual VCG Member One working VCG per connection One backup VCG per working VCG member Page Other Related Research: A Sampling Application to MPLS restorable tunnels [Kodialam03] Comparable to our studies with fine (continuous) bandwidth granularity and full wavelength conversion at each node Note: The survivability research studied in this presentation has wide applicability e.g., to IP / MPLS restorable tunnels with appropriate adjustments on bandwidth granularity and wavelength conversion Comprehensive review of shared-path-protection methods [Sam-jlt0] p-cycles (or protection cycles) [Grover00] Virtual ring embeddings on mesh for ring-type protection in mesh nets A very good treatment of relevant issues [Doshi99] Very good treatments of practical issues [Ori00a, Ori00b] Shared-Leap Shortest Path [Pinhan0] An alternate sub-path method which also provides node protection Path protection with duct-layer constraints [Hui-ton03] Page 33

34 Review of Shared-Path-Protection Methods [Sam-jlt0] Page 7 Bibliography. [Jing-ieeenet0] J. Zhang and B. Mukherjee, Review of Fault Management in WDM Mesh Networks: Basic Concepts and Research Challenges, IEEE Network, Special Issue on "Protection, Restoration, and Disaster Recovery", vol. 8, no., pp. -8, March 00 (for general review of the subject).. [Ramu-jlt03] S. Ramamurthy, L. Sahasrabuddhe, and B. Mukherjee, Survivable optical WDM networks, IEEE/OSA Journal of Lightwave Technology, vol., no., pp , April 003 (for a network-design problem; static traffic). 3. [Jian-commag0] J. Wang, L. Sahasrabuddhe, and B. Mukherjee, Path vs. subpath vs. link restoration for fault management in IP-over-WDM networks: performance comparisons using GMPLS control signaling, IEEE Communications Magazine, vol. 0, no., pp , Nov. 00 (for a traffic-engg. problem; dynamic traffic).. [Laxman-jsac0] L. Sahasrabuddhe, S. Ramamurthy, and B. Mukherjee, ``Fault Tolerance in IP- Over-WDM Networking: WDM protection vs. IP restoration," IEEE Journal on Selected Areas in Communications, vol. 0, no., pp. -33, Jan [Jing-ofc03] J. Zhang, K. Zhu, H. Zang, and B. Mukherjee, ``Service Provisioning to Provide Per- Connection-Based Availability Guarantee in WDM Mesh Networks," Proc., OFC 03, Atlanta, GA, pp. -, March 003 (expanded version provided).. [Grace-jlt0] Y. Huang, W. Wen, J. P. Heritage and B. Mukherjee, ``A Generalized Protection Framework using a New Link-State Availability Model for Reliable Optical Networks," IEEE/OSA Journal of Lightwave Technology, vol., Dec. 00, to appear (see also ICC-0). 7. [Jing-icc0] J. Zhang, K. Zhu, and B. Mukherjee, "A Comprehensive Study on Backup Reprovisioning to Remedy the Effect of Multiple-Link Failures in WDM Mesh Networks," Proc., IEEE ICC-0, Paris, June 00. Page 8 3

35 Bibliography (contd.) 8. [Sam-jsac0] C. (Sam) Ou, H. Zang, N. K. Singhal, K. Zhu, L. H. Sahasrabuddhe, R. A. MacDonald, and B. Mukherjee, Sub-Path Protection for Scalability and Fast Recovery in Optical WDM Mesh Networks, IEEE Journal on Selected Areas in Communications, 00, accepted for publication (see also OFC-0). 9. [Sam-net0] C. (Sam) Ou and B. Mukherjee, Differentiated Quality-of-Protection Provisioning in Optical/MPLS Networks, Proc., Networking '0, Athens, May [Sam-jsac03] C. (Sam) Ou, K. Zhu, H. Zang, L. H. Sahasrabuddhe, and B. Mukherjee, Traffic Grooming for Survivable WDM Networks-Shared Protection, IEEE Journal on Selected Areas in Communications, vol., pp , Nov [Sam-ofc0] C. (Sam) Ou, K. Zhu, N. Singhal, and B. Mukherjee, Survivable virtual concatenation for data-over-sonet/sdh networks, Proc., OFC-0, March 00.. [Sam-jlt0] C. (Sam) Ou, J. Zhang, H. Zang, L. H. Sahasrabuddhe, and B. Mukherjee, New and Improved Approaches for Shared-Path Protection in WDM Mesh Networks, IEEE Journal of Lightwave Technology, vol., pp. 3-3, May [Hui-ton03] H. Zang, C. Ou, and B. Mukherjee, Path-Protection Routing and Wavelength- Assignment (RWA) in WDM Mesh Networks under Duct-Layer Constraints, IEEE/ACM Transactions on Networking, vol., no., pp. 8-8, April [Hui-onm0] H. Zang and B. Mukherjee, Connection Management for Survivable Wavelength- Routed WDM Mesh Networks, SPIE Optical Networks Magazine, Special Issue on Protection and Survivability in Optical Networks, vol., no., pp. 7-8, July 00. Page 9 Bibliography (contd by Others). [Kodialam03] M. Kodialam and T.V. Lakshman, Dynamic routing of restorable bandwidth-guaranteed tunnels using aggregated network resource usage information, IEEE/ACM Transactions on Networking, vol., pp , June 03.. [Grover00] D. Stamatelakis and W. D. Grover, Theoretical underpinnings for the efficiency of restorable networks using preconfigured cycles ( p-cycles ), IEEE Transactions on Communications, vol. 8, no. 8, pp. -, Aug [Doshi99] B. T. Doshi, S. Dravida, P. Harshavardhana, O. Hauser, and Y. Wang, Optical network design and restoration, Bell Labs Technical Journal, vol., pp. 8-8, Jan.-Mar [Ori00a] O. Gerstel and R. Ramaswami, Optical layer survivability a services perspective, IEEE Communications Magazine, vol. 38, no. 3, pp. 08-3, March [Ori00b] O. Gerstel and R. Ramaswami, Optical layer survivability--an implementation perspective, IEEE J. Selected Areas in Communications, vol. 8, no. 0, pp , Oct [Pinhan0] P.-H. Ho and H. Mouftah, A framework for service-guaranteed shared protection in WDM mesh networks, IEEE Communications Magazine, vol. 0, no., pp , Feb. 00. Page 70 3