Learning Automata-Based Adoptive Bandwidth Provisioning in Differentiated Services

Abstract

In this paper we propose dynamic bandwidth provisioning method in differentiated services (DiffServ) based on learning automata. This mechanism tries to maximize bandwidth utilization and not breech quality of services (QoS) levels contracted in service level agreement. These two objectives may can contrary to each other; a trade-off has to be made. In proposed mechanism, the amount of bandwidth to provision for each per hop behavior (PHB) adjusts at regular intervals based on environment feedback. The result of multiple simulations presents using this method improves QoS in the term of loss rate, delay and throughput in comparison to static provisioning. The results show that proposed method is able to adopt to converge rapidly to optimal policy under changing traffic conditions, as well as pricing plans and QoS requirements.