Characterization of pre-cross-connected trails for optical mesh network protection

Journal of Optical Networking

Volumn 5, Issue 6, 2006, Pages 493-508

  Grue, Aden ^a   Grover, Wayne D ^a  

^a NONE

Author keywords

[No Author keywords available]

Indexed keywords

COMPUTATIONAL COMPLEXITY; FAILURE ANALYSIS; FUNCTION EVALUATION; HEURISTIC METHODS; INFORMATION TECHNOLOGY; OPTICAL COMMUNICATION; STATISTICAL MECHANICS;

HEURISTICALLY OBTAINED DESIGNS; OPTICAL MESH NETWORK PROTECTION; OPTICAL NETWORKS; PRE-CROSS-CONNECTED TRAILS (PXT);

NETWORK PROTOCOLS;

EID: 33745251647     PISSN: 15365379     EISSN: 15365379     Source Type: Journal    
DOI: 10.1364/JON.5.000493     Document Type: Article

References (20)

1. K. Poulsen, "The backhoe: a real cyberthreat," Wired News, January 9, 2006, http://wired.com/news/technology/0, 70040-0.html?tw=wn_tophead_1
2. If nodal cross-connection actions were assumed to be relatively slow, however, pre-cross-connection may still be desirable even in an opaque network, sheerly from a speed-of-restoration standpoint. This in fact was the original motivation for studying pre-cross-connection in Ref. [3].
3. M. H. MacGregor, W. D. Grover, and K. Ryhorchuk, "Optimal spare capacity preconfiguration for faster restoration of mesh networks," J. Netw. Syst. Manage. 5, 159-171 (1997).
4. R. L. Freeman, Fiber-Optic Systems for Telecommunications (Wiley, 2002), Chaps. 6 and 10.
5. In addition to Ref. [4], we hear from colleagues in network operating companies interested in adopting efficient shared mesh optical protection schemes that this seems to be the main practical obstacle to deploying such schemes with optically transparent optical cross connects. Interestingly, most academic literature and IETF-based deliberations have overlooked this in considering the "backup multiplexing" or shared-backup path protection principle which always involves on-the-fly assembly of backup paths from preplanned, but not preconnected, spare channels on backup routes.
6. W. D. Grover and M. H. MacGregor, "On the potential for spare capacity preconnection to reduce crossconnection workloads in mesh-restorable networks," Electron. Lett. 30, 194-195 (1994).
7. W. D. Grover and M. H. MacGregor, "Method for preconfiguring a network to withstand anticipated failures," U.S. Patent 5850505 (December 15, 1998).
8. T. Y. Chow, F. Chudak, and A. M. French, "Fast optical layer mesh protection using pre-cross-connected trails," IEEE/ACM Trans. Netw. 12, 539-547 (2004).
9. While it is true that 1+1 automatic protection switching (APS) and demandwise shared protection [10] schemes also employ preconnected linear segments to protect against failures, we refer here to architectures that achieve high-capacity efficiencies by sharing the spare capacity on these segments over multiple nonsimultaneous failure scenarios that each simultaneously affect demands exchanged between multiple node pairs. In other words, the protection structures here are not dedicated to only a single node pair.
10. C. Gruber, A. Koster, A. Zymolka, R. Wessäly, and S. Orlowski, "A computational study for demand-wise shared protection," presented at Fifth International Workshop On Design Of Reliable Communication Networks (DRCN 2005), 16-19 October 2005.
11. Ref. [8] says specifically in regards to p-cycles, "... in many situations, achieving high bandwidth efficiency requires the deployment of large p-cycles." Not only is this an implied claim that PXTs somehow avoid concerns of excess length (which our results contradict), but the statement itself perpetuates a misunderstanding that is only true for the unnecessary but intriguing case of employing a single Hamiltonian p-cycle to protect an entire network or domain. In all other cases typical network designs can achieve near-lower limits on bandwidth efficiency using a set of p-cycles of controlled maximum length, typically of no more than 5 to 7 hops. See Chap. 10 in Ref. [12] or Refs. [13, 14] for further discussion supporting these points.
12. W. Grover, Mesh-Based Survivable Networks: Options and Strategies for Optical, MPLS, SONET and ATM Networking (Prentice-Hall, 2003), pp. 210-211.
13. A. Sack and W. D. Grover, "Hamiltonian pcycles for fiber-level protection in homogeneous and semi-homogeneous optical networks," Special Issue on Protection, Restoration, and Disaster Recovery, IEEE Netw. 18, 49-56 (2004).
14. A. Kodian, A. Sack, and W. D. Grover, "The threshold hop-limit effect in pcycles: comparing hop- and circumference-limited design," Opt. Switch. Netw. 7, 72-85 (2005).
15. "Greedy" refers technically to a type of optimization algorithm where a series of local subproblems are optimized as a surrogate for direct optimization of the overall global problem.
16. R. Bhandari, Survivable Networks: Algorithms for Diverse Routing (Kluwer Academic, 1999).
17. Despite these careful considerations, the trap situation did not actually arise in any of the experimental trials that follow. It is, however, an important consideration in general to ensure solution feasibility. Without it an algorithm such as that of Ref. [8] can actually produce a design that is not fully restorable.
18. The single remaining test case showed a discrepancy of almost 70% and remains unexplained despite our efforts. The discrepancy occurred with the K6,6 network topology combined with the "uniform" demand pattern. The algorithm's dependence on demand order is a possible source of this discrepancy and makes the agreement between the other 11 test cases all the more striking. This outlier in our ability to reproduce the results from Ref. [8] may ultimately be due to a publication error or perhaps something attributable to the very regular nature and high connectivity of the K6,6 graph which makes reproducing the results significantly more dependent on using the same ordering of demands.
19. W. D. Grover and D. Stamatelakis, "Cycle-oriented distributed pre-configuration: ring-like speed with mesh-like capacity for self-planning network restoration," Proceedings of the IEEE International Conference on Communication (ICC '98) (IEEE, 1998), pp. 537-543.
20. Note the important distinction here between the length of a PXT itself and the length of any individual part of the PXT used as a protection path for a given failed demand. (This is similar to the difference between the circumference of a p-cycle and the maximum length of a protection path obtained from the cycle, which may be controlled if desired by design methods in Ref. [14]). We thank one of the reviewers for suggesting that we also inspect this characteristic measure for PXT designs.