A queue buffer management algorithm based on traffic prediction in DiffServ architecture

Abstract

In the DiffServ architecture, a queue buffer management algorithm-PARIO-C algorithm based on traffic prediction was proposed. The algorithm used Recursive Least Squares (RLS) method to predict the change of average queue length in the network, and combined Adaptive Random Early Detection (ARED) algorithm to adjust the parameter adaptively, thus achieved better control of network congestion. Experimental comparison with Random Early Detection with In/Out and Couple queue (RIO-C) algorithm on latency, jitter, packet loss rate, the results show that PARIO-C algorithm lower the scope of network jitter, and reduce the network latency and packet loss rate, so improve network QoS to a certain extent.