Capacity Planning of Survivable MPLS Networks Supporting DiffServ

It is essential for ISPs to offer both performance and survivability guarantees at the IP/MPLS layer. Capacity planning is needed to ensure there will be sufficient resource availability and quality of service under normal circumstances, and during network failures. In this paper we study the issue of capacity planning for survivable MPLS networks providing DiffServ EF and BE traffic classes. Our goal is to minimize the total link cost, subject to the performance and survivability constraints of both EF and BE classes. The problem is formulated as an optimization problem, where we jointly select the routes for edge to edge EF and BE user demand pairs, and assign a discrete capacity value for each link. Capacity for the EF class is fully restorable under the single link failure model, while restorability of the BE class can be adjusted by the network operator. We propose an efficient solution approach based on the Lagrangian Relaxation and subgradient methods. Computational results show that the solution quality is within a few percent of optimal, while the running time remains reasonable for networks with 1000 nodes, 2500 links, and 40,000 demand pairs. This represents the first work on capacity planning for survivable multiple-class-of-service IP/MPLS networks with non-linear performance constraints.