Effects of bursts in effective bandwidth estimation

Bursty traffic has been known to affect network performance in various ways such as causing low resource utilization or high cell loss ratio (CLR). In order to achieve controlled quality of service (QoS) requirements as well as optimal utilization of network resources, the concept of effective bandwidth was proposed to provide a useful practical framework for connection admission control (CAC) and capacity planning in high-speed communication networks. However, the effectiveness of the effective bandwidth is affected by different kinds of bursty traffic. In this paper, we investigate the characters of two different bursty traffic, short-range bursty (SRB) traffic and long-range dependent (LRD) traffic, and their effects on the estimation of effective bandwidth. We get the conclusion that different kinds of bursty traffic can affect the estimation of effective bandwidth in two different ways: the conservative degree and the CLR slope