An effective bandwidth allocation approach for self-similar traffic in a single ATM connection

The cell loss probability and cell latency have been two major approaches to estimate the effective bandwidths of communication network connections. Unfortunately, most of methods calculate the effective bandwidths under some unrealistic assumptions that sometimes results in underestimated values specially when the traffic is self-similar. Another consequence of this is the failure of maintaining the contracted QoS. In this study, we generalize the effective bandwidth estimation method proposed by Kesidis providing a more realistic estimate whether the traffic is self-similar or not. Experimentally the estimated effective bandwidth is optimized to reduce the impact provoked by highly self-similar traffic. For this end, analytical, heuristical and simulation methods were jointly used both in the design and validation of the proposed method