Introducing a Relative Priority for the Shared-Protection Schemes Nizar Bouabdallah and Bruno Sericola Abstract—One of the major challenges of optical network operators is ensuring the stringent levels of availability required by their highest class clients. To achieve this, we introduce relative priorities among the different primary connections contending for access to the shared-protection paths. In this paper, we provide an analytical model for the proposed priority-enabled scheme. As a key distinguishing feature from existing literature, we derive explicit analytic expressions for the average availability and service disruption rate for the different priority classes. Index Terms—Optical networks, protection, performance analysis. Ç 1 INTRODUCTION O VER the last decade, networks have been witnessing a perpetual growth in data traffic. Due to the new incumbent challenges, the operators are progressively migrating toward optical core networks taking advantage of the tremendous transmission capacity offered by the optical technology, thanks to the revolutionary Wavelength- Division Multiplexing (WDM) technology. Meanwhile, the relentless need for more capacity may have been satisfied. However, in such an environment, the cut of a fiber link can lead to a tremendous traffic loss. In this regard, network survivability becomes a critical concern for operators. To alleviate this, backup resources are used to restore failed connections. These resources are usually shared among several primary connections to improve the network utilization. Generally, the primary connections are considered as equally important when contending for the use of the backup resources [1], [2], [3]. However, this solution is unsuitable from the service perspective. Indeed, the quality of service (QoS) required by different clients can be very different because of their diverse services’ characteristics. For instance, banking services will require stringent reliability, whereas Internet Protocol (IP) best effort packet delivery services may be satisfied without a special constraint on reliability. One possible solution to provide different levels of reliability is to use a priority mechanism for the shared-protection scheme. Recently, some service differentiation schemes have been proposed in literature [4], [5]. The impact of these schemes on the system reliability is conducted based on simulations. Note that service reliability can be mea- sured by means of two basic parameters: service availability and service disruption rate. These parameters were assessed analytically in [6] for the single backup path (BP) shared-protection scheme (that is, the 1: N case). To achieve this, the authors adopted simplifying assumptions such as all the paths (primary and backup) are considered equally available (that is, having the same failure and repair rates). This assumption is not realistic. Specifically, the BP is usually chosen to be the longest link-disjoint path among all the precomputed ones. In so doing, the BP has the lowest availability compared to its associated primary paths. Likewise, the primary path availabilities of different connections are not the same and depend on their classes of service. Moreover, the authors in [6] limited their study to only two classes of service, whereas the majority of the standards deal with multiple (that is, more than two) classes of service. Generally, four classes of service can be identified (pre- mium, gold, silver, and bronze), and according to the customer profiles’ and the provided service, a finer classification can be achieved [7]. As our first main contribution, we therefore derive the reliability parameters associated to the 1: N protection scheme for multiple classes of service with different path availabilities. Furthermore, [6] only handled the 1: N scheme. In view of this, as our second main contribution, in this paper, we give explicit formula of reliability parameters related to the multiple BPs protection scheme (that is, the M : N scheme, with M  1). In this case, analytical results are derived only for two classes of service. Finally, previous works [4], [5], [6] suggested the use of strict preemptive priority to differentiate among different classes of connections. This kind of policy may not be desired, since it is extremely penalizing for low-priority classes. To cope with such limitation and as our third main contribution, we suggest a new policy that allows relative priority among different classes of service. We develop a new analytical model to analyze our proposal. Accordingly, all underlying policies are compared. We demonstrate that our new scheme provides operators much more flexibility to manage their networks in order to satisfy various client requirements. Specifically, it enables cost savings compared to existing solutions. IEEE TRANSACTIONS ON DEPENDABLE AND SECURE COMPUTING, VOL. 4, NO. 3, JULY-SEPTEMBER 2007 205 . The authors are with the INRIA-IRISA, Campus Universitaire de Beaulieu F-35042, Rennes CEDEX, France. E-mail: {nizar.bouabdallah, bruno.sericola}@inria.fr. Manuscript received 23 June 2006; revised 23 Jan. 2007; accepted 3 May 2007; published online 29 May 2007. For information on obtaining reprints of this article, please send e-mail to: tdsc@computer.org, and reference IEEECS Log Number TDSC-0086-0606. Digital Object Identifier no. 10.1109/TDSC.2007.1008. 1545-5971/07/$25.00 ß 2007 IEEE Published by the IEEE Computer Society Authorized licensed use limited to: UR Rennes. Downloaded on March 23, 2009 at 07:51 from IEEE Xplore. Restrictions apply.