Interior point-based optimization for joint admission control and routing in IP networks

Abstract

For MPLS-like IP networks, we propose a novel and practically implementable optimal joint admission control and routing (JACR) mechanism for incoming trunk traffic classes with distinct QoS requirements. The proposed mechanism is based on interior point-based optimization techniques to solve the admission control/routing problem in large service provider networks. The significant contributions of our proposed mechanism are a) an efficient resource allocation mechanism that i) simultaneously satisfies both user QoS requirements and network provider's capacity and routing constraints and ii) engineers the usage of network resources efficiently by load-balancing and preventing hotspots from developing in the network, b) computationally efficient to make decisions within real-time call (flow) setup durations, c) scalability to network size and complexity and d) lends itself to a distributed implementation. The proposed mechanism runs in two distinct phases - phase-1 is the admission control and feasible route determination step and phase-2 is the optimal routing step. This structure provides flexibility to the service provider to implement this mechanism in its complete form or even implement it solely as an admission control/feasible route determination algorithm (i.e. use phase-1 only). Experimental results on large-size service provider network with 56 nodes, 117 links and 14 LSPs between an ingress-egress pair demonstrate that, for incoming traffic with diverse QoS requirements, the proposed mechanism is very efficient in determining admission and allocating network resources by optimal routing, all within real-time call setup times.