Performance improvements on Self-Similar Traffic using Measurement-Based Admission Control

Abstract

In this paper we consider an admission-controlled traffic scenario, in which non-adaptive connections are offered to the network. We show the superiority of Measurement Based Admission Control (MBAC) algorithms over Parameter Based Admission Control (PBAC) algorithms, in particular when Long Range Dependence (LRD) or asymptotic second order self-similarity arises in the offered traffic. We show that a pure MBAC approach is capable of significantly reduce the Long Range Dependence of the accepted traffic aggregate, thus yielding a considerable performance improvement, in terms of Quality of Service (QoS) experienced by each flow sharing the media. Moreover the MBAC approach results in a powerful tool, robust to traffic correlations properties and burstiness, and able to overcome the horizon of PBAC scheme, despite of the traditional meaning of MBAC role, commonly intended as an even coarse approximation for PBAC.