Dynamic Traffic Grooming under a Differentiated Resilience Scheme for WDM Mesh Networks

Introduction

The introduction of optical fibres as a transmission medium began a revolution in telecommunications. Optical fibre offers a number of advantages and capabilities that can meet the requirements of modern telecommunications networks (Shepherd, 2004). The huge bandwidth is considered the main advantage of optical fibre; however the challenge is to develop the necessary technologies to exploit the huge bandwidth promised by optical fibre to keep network capacity at pace with the enormous bandwidth demand of current and future applications. WDM has emerged as the technology of choice to harness the huge bandwidth available in an optical fibre. Currently WDM is widely deployed in networking infrastructures and is expected to play a significant role in supporting the requirements of the next generation networks in terms of capacity, latency and reliability.

In wavelength-routed WDM networks, the provisioning of network resources to satisfy connection requests is known as the routing and wavelength assignment (RWA) problem (Chlamtac, Faragó, Zhang, 1996) where lightpaths are created to span multiple fiber links. In all-optical networks, a lightpath remains in the optical domain by optically bypassing intermediate nodes. Most of the previously studied RWA algorithms assume that connections require a full wavelength. However, while the wavelength transmission rate has reached OC-192 (10Gbps) and is expected to reach OC-768 (40Gbps) and beyond in the future, networks are still required to support traffic connections at rates lower than the full wavelength capacity (as low as OC-3 (155Mbps)). In addition, for networks of practical size, the number of available wavelengths is still a few orders of magnitude lower than the number of source-destination connections. The bandwidth gap between the low-rate connections and the high-rate wavelengths is addressed by allowing sub-wavelength granularity connections to be groomed onto a single lightpath which results in efficient utilization of network resources.

Fault-tolerance for WDM networks is a major architectural and design issue as a single link failure can cause loss of an enormous amount of information. Network resiliency is defined as the network ability to reconfigure and re-establish communication incase of failure. Different resilience approaches for WDM networks have been extensively studied in the literature (Aneja, Jaekel, & Bandyopadhyay, 2007), (Autenrieth, & Kirstädter,2002), (Bouillet, 2002), (Van Caenegem, 1998) . Resilience procedures can be classified depending on various criteria. The backup path establishment method was considered for many years as the main classification criterion as it implies the general quality parameters. Based on this criterion recovery methods can be classified to: protection and restoration methods. Under protection the protection path or the set of possible protection paths are calculated while the working path is established. Resources can be allocated before a failure occurs or they can be allocated after the failure. Under the second option backup resources can be effectively shared among different working paths. In this case a recovery time is considerably longer as signaling is required to allocate the backup recourses. On the other hand, under restoration procedures the backup path is computed on demand to re-route the traffic affected by the failure. Restoration disadvantages include long switching time, temporary instabilities, and the risk of loop creation. However, it is more efficient in terms of bandwidth utilization compared to protection. In practice, protection can be combined with restoration where fast but not efficient protection techniques are used to recover the traffic, and then using restoration techniques a better route is computed and the traffic is switched. According to the allocation of backup resources, protection methods can be classified into two major classes: protection with dedicated or shared backup resources. In the case of dedicated backup resources, backup resources can be used exclusively to establish a backup path related to a particular working path. On the other hand, backup resources are shared among several disjoint working paths. Obviously, the dedication of backup resources is considered to enable fast recovery compared to sharing backup resources where signaling is required to associate backup paths with a traffic coming from faulty working paths. However, the dedication of backup resources is very expensive due to exclusiveness usage of backup resources (Vasseur, 2004).