Statistical services for multiple-time-scale policed traffic

In order for a network resource reservation scheme to efficiently support bursty real-time traffic streams, it must exploit the effects of statistical multiplexing. Moreover providing statistical services with deterministic traffic models has the advantage that deterministic client-specified traffic parameters may be efficiently policed or enforced by the network. We study the impact of the traffic´s time scales on the effectiveness of such enforceable network services. In particular we use single and dual time-scale traffic sources to illustrate the importance of explicitly incorporating the source´s multiple-time-scale nature into both the deterministic traffic model as well as the admission control algorithm. By investigating the errors introduced with a single time scale scheme, we show that even with an ideal choice of traffic parameters, a single-time scale approach can significantly under-estimate the true admissible region for dual time-scale traffic