Heterogeneity-aware shortest path routing: flow holding time, user demand and network state

We investigate possible performance improvements by exploring heterogeneity of traffic characteristics when designing a shortest path routing scheme. First we focus on the effect of the maintenance of the link metrics for connections with different holding times. We found that by using a differentiated routing scheme with respect to connection holding times, one can enhance network performance for a range of traffic loads. Second we propose a selective routing control scheme which determines whether to accept a shortest path routing decision based on user demands in the (source, destination, bandwidth request) tuple. Simulations were conducted to exhibit the effectiveness of such a routing algorithm. Finally, realizing the difference in routing "local" versus "transit" traffic, we present a novel approach to aggregate network state by drawing on an analogy to circuit theory. The proposed "effective capacity" abstraction is not only efficient in terms of the signaling savings, but also maintains a compatible routing metric to represent aggregated and local states.