Performance Comparison of Packet-Level Multiplexing Algorithms with Bursty Traffic

Abstract

Packet-based networks have opened the possibility of carrying a great variety of applications with different traffic profiles. Traffic profiles such as bursty traffic is essential for the analysis and evaluation of network performance. This paper is a simulation study devoted to comparing the performance of a range of packet-level multiplexing algorithms with bursty traffic. Four multiplexing algorithms are studied: round robin, random selection, weighted random selection and longest queue. Bursty traffic in this study is specified using two parameters: average packet rate and mean burst length. The algorithms performance has been measured in terms of three performance measures: average delay, delay variation and packet loss. The performance of multiplexing algorithms deteriorated when traffic intensity and burstiness of traffic increase. The algorithms competed between each other. Each algorithm showed better performance for a certain measure, however, with the expense of deterioration in other performance measures.