Feasibility of using stochastic adaptive congestion control in computer networks

Abstract

The paper presents a novel approach for congestion control and avoidance in computer networks. The approach is based on stochastic search, as opposed to systematic-like search employed by existing mechanisms. Our approach forms basis for more informed heuristics and promises to be more adaptive. We employ the Weibull distribution density function to continuously define the search space. In this paper we develop a single simple heuristic, based on the approach, and evaluate its performance in terms of congestion window evolution, response time and throughput.