Abstract

While there has been significant progress in recent years in the development of TCP congestion control algorithms for high BDP paths, consensus remains lacking with regard to a number of basic issues. The aim of the present paper is to highlight some of these key bottleneck issues and present a number of new results with a view to promoting discussion and fostering progress. Issues highlighted include: impact of shape of cwnd evolution (concave, convex etc), increased variability in throughputs in unsynchronised environments when more aggressive algorithms are used, impact of proposed changes on convergence rates and network responsiveness and the associated impact on user experience.