Traffic Grooming for Survivable WDM Networks – Shared Protection

Traffic Grooming for Survivable WDM Networks – Shared Protection Kevin Su University of Texas at San Antonio Kevin Su (xsu@cs.utsa.edu)

Outline • Introduction • Motivation • System Model • Grooming Node Architecture • Network Model • Proposed Schemes • Protection-at-lightpath (PAL) level • Mixed protection-at-connection (MPAC) level • Separated protection-at-connection (SPAC) level • Heuristic Algorithms • Performance Evaluation Kevin Su (xsu@cs.utsa.edu)

Introduction • WDM: stands for wavelength division multiplexing, it is a technology that divides the bandwidth of an optical fiber into many non-overlapping wavelengths, so that multiple communication channels can operate simultaneously on different wavelengths. • Increases the transmission capacity of optical fibers. • Allows simultaneously transmission of multiple wavelengths within a single fiber. • Up to 320 wavelengths per fiber; per wavelength, 10Gb/s, STS-192 (OC-192), today; expected to grow to 40Gb/s, STS-768(OC-768), soon. Kevin Su (xsu@cs.utsa.edu)

Introduction • Traffic Grooming: refers to problem of efficiently packing low-speed connections onto high-capacity lightpaths to better utilize network resourses. • The bandwidth requirement of a typical connection request is between STS-1 (51.84 Mb/s) and STS-192(full wavelength) • Protection: is a proactive procedure in which spare capacity is reserved during connection setup. • Working path: a path that carries traffic during normal operation • Backup path: a path over which the connection is rerouted when a working path fails • Single Failure (single-fiber failure, single-node failure) ---0--- ---0------0--- • Dedicated Protection • Shared Protection Kevin Su (xsu@cs.utsa.edu)

Introduction Kevin Su (xsu@cs.utsa.edu)

Motivation • Survivable Traffic Grooming • Efficiently utilize the network resources (traffic grooming) • A failure of a network element can cause the failure of several lightpaths, thereby leading to large data and revenue loss (protection) • Static Case • Dynamic Case Kevin Su (xsu@cs.utsa.edu)

Grooming Node Architecture Kevin Su (xsu@cs.utsa.edu)

Network Model • A network is represented as a weighted, directed graph G=(V, E, C, λ, P) • V: set of nodes • E: set of unidirectional fibers (referred to links) • C: the cost function for each link • λ: the number of wavelengths on each link • P: number of grooming ports at each node • Connection request is represented as a quadruple <s, d, B, > • s: source node • d: destination node • B: bandwidth requirement • : holding time Kevin Su (xsu@cs.utsa.edu)

Proposed Schemes - PAL • Protection-at-lightpath (PAL) level provides end-to-end protection w.r.t. lightpath. Under PAL, a connection is routed through a sequence of protected lightpath, or p-lightpath. • A p-lightpath has a lightpath as working path and link-disjoint path as backup path • Working path consumes a grooming-add port at the source node and a grooming-drop port at the destination node • Backup path doesn’t consume any grooming port and wavelengths along a backup path are only reserved • When working path fails, backup path is set up as a lightpath by utilizing the grooming ports previously used by the working path • Two p-lightpaths can share wavelengths along common backup links if their working paths are link-disjoint. Kevin Su (xsu@cs.utsa.edu)

Proposed Schemes - PAL • Initial network configuration • Edge represents bidirectional fiber, each fiber has 2 wavelengths • Wavelength capacity STS-192, every node has 3 grooming ports • c1<0,2,STS-12, t1>; c2 < 0,3,STS-3, t2>; c3 <4,3,STS-48, t3> Kevin Su (xsu@cs.utsa.edu)

Proposed Schemes - PAL • After provisioning c1 • c1<0,2,STS-12, t1>; c2 < 0,3,STS-3, t2>; c3 <4,3,STS-48, t3> Kevin Su (xsu@cs.utsa.edu)

Proposed Schemes – MPAC • Mixed Protection-at-Connection (MPAC) level provides end-to-end protection w.r.t. connection. Under MPAC, a connection is routed via link-disjoint working path and backup path, each of which traverses a sequence of lightpaths. • A lightpath traversed by a working path utilizes a portion of its capacity to carry traffic during normal operation • A lightpath traversed by a backup path reserves part of its capacity for that backup path • “Mixed” means that capacity of one wavelength can be utilized by both working paths and backup paths. Kevin Su (xsu@cs.utsa.edu)

Proposed Schemes - MPAC • After provisioning c1 • c1<0,2,STS-12, t1>; c2 < 0,3,STS-3, t2>; c3 <4,3,STS-48, t3> Kevin Su (xsu@cs.utsa.edu)

Proposed Schemes – SPAC • Separated Protection-at-Connection (MPAC) level provides end-to-end protection w.r.t. connection. Under SPAL, a connection is routed via link-disjoint working path and backup path. • A working path traverses a sequence of lightpath. • A backup path traverses a sequence of links, each of which has judiciously reserved a number of wavelengths as backup resourses • “Separated” means that the capacity of a wavelength can be utilized by either working paths or backup paths, but not both. Kevin Su (xsu@cs.utsa.edu)

Proposed Schemes - SPAC • After provisioning c1 • c1<0,2,STS-12, t1>; c2 < 0,3,STS-3, t2>; c3 <4,3,STS-48, t3> Kevin Su (xsu@cs.utsa.edu)

Proposed Schemes Kevin Su (xsu@cs.utsa.edu)

Heuristic Algorithm • It is NP-complete to provision a connection request with shared protection. • The Author proposed heuristic for MPAC, SPAC, PAL • MPAC • Backup-sharing measuring Every lightpath is associated with a conflict set to identify the sharing potential between paths. conflict set for lightpath can be represented as an integer set { } where represents the amount of traffic that will be rerouted on lightpath when link e fails. The amount of backup capacity reserved on lightpath is thus • Route computation • Enumerates K candidate working paths • For each candidate working path, computes a disjoint minimal-cost path as backup path based on some cost function • Selects the path pair of minimal cost Kevin Su (xsu@cs.utsa.edu)

Heuristic Algorithm • SPAC (the same as MPAC except) • Different backup-sharing measurement • Different cost function in route computation • PAL • Different backup-sharing measurement • Route computation • Extend a stand shortest-path algorithm such that every hop along the resultant shortest path corresponds to a p-lightpath, which can be either an exisiting p-lightpath or a new p-lightpath consisting of fresh wavelength links and free grooming ports Kevin Su (xsu@cs.utsa.edu)

Performance Evaluation • Connection-arrival process is Poisson process • Connection-holding time follows a negative exponential distribution • Capacity of wavelength is STS-192 • # of connection requests follows the distribution STS-1: STS-3: STS-12: STS-48: STS-192 = 300: 20: 6: 4 :1 • Load (in Erlang) is defined as connection-arrival rate times average holding time times a connection’s average bandwidth normalized in the unit of STS-192 • Number of grooming ports is set as # of wavelengths times its nodal degree times a scalar ( implies that any incoming wavelength to the W-Fabric can be dropped to the G-Fabric) • The number of alternate paths K = 2 • Measurement metrics • Bandwidth-blocking Ratio • Resource-Efficiency Ratio Kevin Su (xsu@cs.utsa.edu)

Network Topology 24-node example network topology Kevin Su (xsu@cs.utsa.edu)

Performance Evaluation Kevin Su (xsu@cs.utsa.edu)

Performance Evaluation BBR versus network offered load with k =1,2 and 3 Kevin Su (xsu@cs.utsa.edu)

Conclusion and Future Work • Investigate the survivable traffic-grooming problem in dynamic case • PAL, MPAC, SPAC • Findings: • It is beneficial to groom working paths and backup paths separately as in PAL and SPAC • Separately protecting each individual connection yields the best performance when the number of ports is suffcient • Protecting each specific lightpath achieves the best performance when the # of grooming ports is moderate or small • Future work • Considering the residual connection holding time Kevin Su (xsu@cs.utsa.edu)

References • C. Ou, K. Zhu, H. Zang, L. H. Sahasrabuddhe, and B. Mukherjee. “Traffic Grooming for Survivable WDM Networks – Shared Protection”. Accepted to IEEE Journal of Selected Area in Communication 2004. • H. Zhu, H.Zang, K. Zhu, and B. Mukherjee, “A novel, generic graph model for traffic grooming in heterogeneous WDM mesh networks” IEEE/ACM Trans. Neworking, vol.11 pp.285-299, Apr. 2003 Kevin Su (xsu@cs.utsa.edu)

Traffic Grooming for Survivable WDM Networks – Shared Protection