Modeling and simulation of a LFVC scheduler

Abstract

Service disciplines that approximate GPS have had a lot of success satisfying QoS isolation among connections in integrated services networks. In the past years, several of such disciplines were developed. However, to maintain a delay little bigger than GPS, with a fair throughput and a good worst-case fairness while reducing computational overhead is a big challenge. LFVC appears to be the first service discipline that optimizes these factors. This motivated us to implement an original LFVC model in the context of an ATM network model, originally developed to trustworthily evaluate QoS in ATM networks. LFVC expands this model and improves its quality, since LFVC is capable to isolate traffic and capture service differences among connections. In addition, LFVC interacts with other models of this model set, producing better simulation results. The paper also presents a model validation and a performance evaluation of the LFVC scheduler.