Fuzzy Logic Token Bucket Bandwidth Predictor for Assured Forwarding Traffic in a DiffServ-Aware MPLS Internet

Abstract

The foundation of the Internet network architecture, which revolves around the transfer of packets over multitude of networks, need to be improvised to support multimedia communications. To this end the Internet community has proposed several architectures that potentially can address this requirement. Two promising architectures are Differentiated Services (DiffServ) and Multiprotocol Label Switching (MPLS). This paper describes a fuzzy logic token bucket bandwidth predictor for Assured Forwarding (AF) traffic aggregates at the DiffServ level in a DiffServ-aware MPLS Internet network. The AF traffic class comprise of variable bit rates traffic. The fuzzy logic bandwidth predictor facilitates bandwidth predicting and dynamic policing at the DiffServ level. A simulation study was performed for the fuzzy logic predictor using Network Simulator-2. The network and traffic models were developed and validated before the performance study is performed. The results show that the fuzzy logic predictor gave commendable bandwidth prediction compared to actual bandwidth utilization of the AF traffic class. © 2007 IEEE.