Optimal capacity placement for path restoration in STM or ATM mesh-survivable networks
IEEE/ACM Transactions on Networking (TON)
Evaluating the Overheads of Source-Directed Quality-of-Service Routing
ICNP '98 Proceedings of the Sixth International Conference on Network Protocols
Survivability in optical networks
IEEE Network: The Magazine of Global Internetworking
Routing demands with time-varying bandwidth profiles on a MPLS network
Computer Networks: The International Journal of Computer and Telecommunications Networking
Routing demands with time-varying bandwidth profiles on a MPLS network
Computer Networks: The International Journal of Computer and Telecommunications Networking
QoS-IP'05 Proceedings of the Third international conference on Quality of Service in Multiservice IP Networks
Hi-index | 0.00 |
MPLS can be used to provide network robustness to faults through path protection techniques. In this paper we present a dynamic model supporting different classes of end-to-end protection, including protection against Single Fault and Dual Fault, with and without sharing of backup bandwidth. Beyond link and node failures we also consider protection against Shared Risk Link Group (SLRG) failure. An interesting feature of the proposed scheme is the ability to offer service differentiation with respect to the recovery probability, by coupling the differentiation on the number of backup paths with bandwidth assignment policy. In this paper we describe the underlying algorithms for route selection and backup bandwidth sharing. The route selection is based on explicit load-dependent routing of service and backup paths. We show by simulation that the proposed route selection algorithm is effective in improving the network utilization. We discuss two alternative implementations of our model: distributed and partially centralized. The primary concern with the distributed approach is the message overhead implied by link-load dissemination, e.g. by flooding. However we show by simulation that message overhead can be taken under control by adopting a well-tuned adaptive overhead reduction algorithm. Our conclusion is that both distributed and partially-centralized implementation are feasible.