DRIFT MATCHING CONTROL OF REPLICATOR PROCESSES IN IP MULTICASTING NETWORKS
Alexandru Murgu · 2004
Abstract: In this paper we consider the problem of multi-service quality networking for which we analyze feature estimates for user populations behavior in terms of the long-term evolution of trends and biases using the group communication services concept. For this purpose, a drift matching control strategy modelled as generalized Polya urns process is formulated. Statistical nature of the multi-service compound process is equivalently described by a mean limit ordinary differential equation called the replication equation. This equation allows the synthesis of evolutionary selection strategies in a connected internally recurrent chain set which is the reflection of the multi-service class transition. Drift matching is in fact a natural selection of populations behaviours where the pairwise interactions lead to highly variable outcomes, but optimal behaviours emerge in the mean limit sense. Large randomly mixing population of flows with overlapping replication patterns are balanced in order to ensure a reliable multi-class service quality. Alexandru Murgu 1