658 IEEE COMMUNICATIONS LETTERS, VOL. 9, NO. 7, JULY 2005 Analysis of Capacity Re-provisioning in Optical Mesh Networks Chadi Assi, Member, IEEE, Wei Huo, Abdallah Shami, Member, IEEE, and Nasir Ghani, Senior Member, IEEE Abstract— We consider optical mesh networks and we study the problem of improving service availability under dual near- simultaneous failures. We propose a re-provisioning algorithm that improves the robustness of optical mesh networks and we compare its performance with another scheme under various degrees of resource sharing. The effectiveness of our proposed approach in improving restorability is demonstrated through simulation analysis. Index Terms— Optical networks, protection, capacity re- provisioning, robustness. I. I NTRODUCTION A S THE size and the complexity of optical networks continue to grow, dual failures become increasingly probable. Hence designing recovery algorithms to protect against such failures is a paramount concern. Normally, the network is protected against single failures using one of the protection schemes (e.g., dedicated/shared, path/link/segment [1,2,6]). Whenever a failure occurs, all affected connections are re-routed on their corresponding protection paths [1,2]. However, since protection resources may also be shared with other unaffected connections, these may become unprotected and vulnerable to subsequent failures [4,5]. Generally, unpro- tected connections can be classified into three types: 1) Indirectly Affected Connections: Upon failure, shared protection resources are activated by the failed connections which may cause some connections (whose backup lightpaths share these protection resources) to become unprotected. 2) Directly Affected Working Connections: A failed demand that is re-routed to its backup is still vulnerable to another failure that may affect its protection path. 3) Directly Affected Backup Connections: Demands whose protection connections have failed due to the first failure. Clearly, larger numbers of unprotected connections in the network increases its vulnerability to subsequent failures. To improve the overall service availability, re-provisioning [3,4,5] exploits the available capacity in the network to re-establish new backup paths for unprotected connections right after the recovery from the first fault without a priori knowledge of the Manuscript received December 8, 2004. The associate editor coordinating the review of this letter and approving it for publication was Prof. Maode Ma. Chadi Assi and Wei Huo are with the Concordia Institute for Information System Engineering, Concordia University, Montreal, Quebec, Canada (e- mail: assi@ciise.concordia.ca). Abdallah Shami is with the Department of Electrical and Computer Engineering, the University of Western Ontario, London, Ontario, Canada. Nasir Ghani is with the Department of Electrical and Computer Engineer- ing, Tennessee Tech. University, USA. Digital Object Identifier 10.1109/LCOMM.2005.07004. location of the second failure. Now, as more connections are allowed to share their protection resources (e.g., to achieve better network utilization), more demands will be packed together. Hence a recovery from a failure will leave a larger number of demands in the network unprotected. Alternatively, limiting the level of sharability of resources will reduce the number of unprotected demands; however this evidently comes at the expense of reduced network performance since bandwidth will not be used efficiently. Additionally, since re-provisioning makes use of available resources in the network to provision new protection capac- ity, limited level of resource sharability will yield a lower flexibility in finding and assigning resources. Therefore, it is clear that there are two conflicting design constraints: on one hand limited sharability may reduce the number of unprotected connections but at the expense of less flexibility in allocating protection capacity for unprotected connections. On the other hand, higher sharability may result in larger number of unprotected connections after the first failure with higher degree of flexibility in provisioning protection capacity. One objective of this work is to provide a study on the performance of capacity re-provisioning under different levels of sharability for protection resources. We also propose a new re-provisioning algorithm and contrast its performance with a conventional scheme. Here, the objective of the algorithm is to reduce the total number of connections that have to be re-provisioned. The motivations are twofold: (1) reduce management overheads in simultane- ously provisioning a large number of connections, and (2) to lower reservation contention between multiple unprotected connections trying to simultaneously establish backup capac- ity. II. NETWORK RE- PROVISIONING A re-provisioning algorithm typically takes several inputs including network topology/usage information and a list of unprotected demands (as classified earlier) that require re- provisioning. The algorithm then tries to establish backup lightpaths [5] for unprotected demands using the available capacity in the network (this algorithm is referred to as Scheme I). Clearly, when the level of sharability of protection resources is high, more connections are packed together and as a result a failure in the network will result in a larger number of unprotected connections. These connections, in turn, will require protection capacity re-provisioning in order to improve the network restorability. Note, the larger the number of unprotected connections in the network, the higher is the management overhead and may result in excessive contentions 1089-7798/05$20.00 c  2005 IEEE