Efficiency/friendliness tradeoffs in TCP Westwood

In this paper, we propose a refinement of TCP Westwood allowing the management of the Efficiency/Friendliness-toNewReno tradeoff. We show that the refined TCP Westwood is able to achieve higher efficiency yet at the same time maintain friendliness. TCP Westwood (for short, TCPW) implements a novel window congestion control algorithm based on available bandwidth estimation. The performance of TCPW has been promising, exceeding that of TCP NewReno in “large leaky pipes”; i.e. over pipes with high bandwidth delay product and suffering some (minimal) amount of random errors. However, under certain circumstances, TCP NewReno may experience some performance degradation because TCPW possesses more information and thus can take better advantage of available bandwidth. In this paper we propose combining the original TCPW sampling strategy which produces Bandwidth Share Estimates (BSE), with a new strategy which produces Rate Estimates (RE). The two sampling techniques are identical if the ACKs are uniformly spaced in time. Simulation and measurements, however, show that ACKs tend to cluster in bursts. As a result, BSE may overestimate a connection fair share. On the other hand, RE tends to be closer to the achieved rate of a connection, but may underestimate the connection fair share in certain conditions. The question is: which estimate – achieved rate or bandwidth share – yields better throughput results, and more generally, better throughput/friendliness tradeoffs? Our studies show that RE works best when packet losses are mostly due to congestion. If, on the other hand, the packet losses are mostly due to link errors, BSE gives better performance. Further, under congestion losses, RE provides better friendliness to NewReno connections. To achieve the “best of all worlds”, we introduce a method we call Combined Rate and Bandwidth estimation (CRB.) A connection first infers the predominant cause of packet losses, and then uses the most appropriate estimation method. We also introduce the Efficiency/Friendliness Tradeoff Graph that provides better tradeoff visualization. In our experiments, we found that CRB provides a better compromise between efficiency and friendliness, and the means to manage such a tradeoff.