Adaptive rate control in high-speed networks: performance issues

It is well established that increased network transmission speed reduces the e€ectiveness of feedback-based adaptive rate control mechanisms, due to the increased bandwidth propagation delay product. In this environment, di€erences in the propagation delays associated with the controlled sources become larger when measured in slots (transmission time of information unit), potentially inducing substantially diversi®ed performance metrics. One of the objectives in this work is to quantify the potential unfairness as well as the reduced e€ectiveness of this rate control scheme in the presence of non-zero propagation delays. As the network speed increases, it is also expected that trac sources will seem to be slower due to the relative increase of the information generation time and/or the decrease of the portion of the bandwidth required by a speci®c source. The other objective in this work is to investigate and quantify the expected increased e€ectiveness of this rate control scheme in the presence of non-zero propagation delays as the trac sources become slower. The studies are carried out by formulating queuing models and evaluating the per-session cell loss probabilities and they are supported by numerical results.