Design of Flexible Protection Plans in Survivable WDM Networks: An Application to PWCE Samir Sebbah ECE, Concordia University Montreal, Qc, H3G 1M8, Canada Email: s sebba@ece.concordia.ca Brigitte Jaumard CIISE, Concordia University Montreal, Qc, H3G 1M8, Canada Email: bjaumard@ciise.concordia.ca Abstract—We propose a new flexible design approach of protection plans in survivable WDM networks by using protection structures with no predefined shapes in order to maximize the protected working capacity in an end-to- end basis. Previous design approaches of survivable WDM based on Protected Working Capacity Envelope (PWCE) have looked at the optimization problem of maximizing the protected capacity on a link basis, independently of the source and destination nodes of the potential traffic. Moreover, those approaches have only investigated the design problem with pre-configured protection cycles (p- cycles). Our design approach proposed in this paper differs from those previously proposed in two main points: (ı) We use pre-configured protection structures (p-structures) with no predefined shapes. By using protection structures with unrestricted shapes, we want to identify the most flexible ones, i.e., those that can provide the highest protected capacity even within constrained spare capacity budget or low network connectivity. (ıı) We maximize the availability of the protected capacity on an end-to-end basis rather than on a link basis. This allow us to more efficiently track fluctuation of the traffic in the networks and among nodes. In order to deal with the large solution space, we develop an ILP optimization model, and use an efficient large scale optimization tool called the Column Generation tool (CG). Results show that a design based on unrestricted p-structure patterns is ∼ 10% less capacity redundant, ∼ 15% more reliable, and allow recovery along shorter backup paths compared to the p-cycle based scheme. Index Terms—Survivable WDM networks, p-structures, end-to-end protected working capacity, column generation. I. I NTRODUCTION The increasing demands for high bandwidth services have pushed telecommunication operators to extensive deployment of WDM networks in access and backbone networks [1]. In WDM, a single fiber link can transport up to hundreds of wavelength channels, each operating at a speed of several Gbps. Thus, any failure in an optical fiber equipment, even for a short time duration, may result in tremendous data losses. Therefore, recovery in case of a network failure is an important issue [2]. In this paper, we develop recovery mechanisms against link failures (the most often encountered failure scenario). Protection methods have evolved from dedicated 1:1 and 1+1 to shared link and path protection methods. Ded- icated protection schemes was used to create fault tol- erant logical topologies by configuring a primary and a backup path for each connection in a way that guarantees less reconfiguration in case of a network failure. Indeed, if the dedicated 1+1 is based on signal monitoring and selection, the 1:1 is based on dedicated backup protection paths that are end-to-end pre-configured. In case of a link failure, only the two end-nodes of the affected link need to switch over the backup paths, and intermediate nodes only relay the received traffic to the next hops in the protection paths. Thus, it results in a high restoration speed and less signaling overhead in case of a failure. However, as a backup protection path is necessary for each protected link channel, the protection redundancy is at least 100%. In order to reduce the capacity redundancy of dedi- cated protection plans, different sharing methods based on different protection structures have been proposed [2], [3]. In protection plans using Shared Backup Link Protection (SBLP) [4] resources are reserved and shared by link disjoint working requests, but configured for recovery (end-to-end) only in case of a failure. However, sharing has its price in terms of signaling overhead. Indeed, in case of a link failure, end-to-end signaling is required in order to isolate (setup) the backup protection paths of the affected link channels. Figure 1 illustrates two protection plans: a dedicated backup link protection in (a), and shared backup link protection in (b). In the dedicated plan, working paths W 1 and W 2 are each assigned an end-to-end backup path P 1 and P 2 , respectively. In the shared plan, the two working paths share the protection capacity (two channels) on link {3, 4}. However, in case of a link failure on one of the working paths, an end-to-end signaling will be required to setup the backup path (local