Network protection design for MPLS networks

Abstract

In this paper, we consider the problem of designing partially and fully survivable explicit label switched paths for MPLS networks. We model the problems as Mixed Integer Linear Programs. Furthermore, we present a survivability assessment of the models. We also evaluate the performance of the models presented in the paper, in terms of total used capacity in the network.

Policy-constrained bio-inspired processes for autonomic route management

“…The common approach towards supporting network survivability is through link-disjoint paths (e.g. primary and alternative paths) [45,27,28]. Using this technique, in the event of any failures, routes will be immediately diverted to backup secondary paths (in what is referred to as a protection switching process), which are pre-computed paths using spare capacity of neighbouring nodes.…”

On OR-based routing approaches for the Internet

“…Four major routing principles that use mutipath routing may be identified: traffic splitting in which the traffic demand offered from origin to destination is divided by several paths, satisfying the total bandwidth requirements of the corresponding service class with obvious advantages in terms of allowing better network load distribution and improved reliability/survivability under failures; alternative routing in which one or more alternative path(s) are attempted to be used when a first choice path is unavailable for the considered node‐to‐node traffic flow due to current bandwidth occupations in the links; protection routing when a number of paths (usually node‐disjoint) have to be calculated in order to guarantee total protection of the origin‐destination traffic in certain failure scenario(s) (or at least protection of the traffic for some sub‐paths of the active path in which the traffic is originally carried) (see e.g. Kodialam and Lakshman, 2003; Agrawal et al, 2005; Rosenbaum et al, 2005); multicast routing involving the calculation of a set of paths from an originating node to multiple destination nodes (e.g. for distributive services in the Internet) which involves in OR terms, the computation of Steiner trees (see examples in Cerulli et al, 2006; Donoso et al, 2004), seeking to optimize some metric(s) such as load cost or average delay; broadcast routing involving the calculation of a set of paths from an originating node to all remaining nodes of the specified network (e.g.…”