On the Gain of Statistical Multiplexing Over Traffic Grooming

In optical networks that are based on provisioned or switched lambda-paths there is no support for sub-lambda granularity, and therefore, the network is not efficient for carrying demands of bandwidth requirement significantly smaller than the capacity of a lambda-path. Even if we add a layer over the lambda-layer to improve granularity and also add grooming capability the network will not be efficiently utilised unless the added upper layer supports statistical multiplexing. The focus of this paper is on showing that in such a two-layer network neither statistical multiplexing in itself nor the traffic grooming with deterministic multiplexing itself can utilise the resources efficiently but only the joint use of both. We perform a simulation study to compare deterministic multiplexing to three different models for statistical multiplexing for two cases, namely, when all nodes are capable of grooming and when no node is capable of grooming at all. The lowest blocking and highest throughput has been achieved for the case of joint use of grooming and statistical multiplexing, particularly when the bandwidth requirements of demands are significantly smaller than the capacity of lambda-channels