A fluid model approximation to quantitative information feedback in congestion control

Abstract

A new approach for deriving quantitative information for rate-based congestion schemes is presented. Using the fluid model, it is shown that the bottleneck capacity can be estimated from round trip delay when the sender is increasing its traffic rate linearly. The exact solution for a time-dependent M(t)/M(t)/1 queue verifies that the approximation is accurate when the traffic load is high. The analytical results were verified with simulation experiments, and the practical issues in applying the estimation techniques in congestion control schemes are discussed.<>