TCP Reno with random losses: latency, throughput and sensitivity analysis

Current models for TCP performance focus mainly on the steady-state throughput of infinite flows, making them inappropriate for the short TCP flows which dominate the transfers over the Internet. We present a model for the latency of finite TCP Reno flows with independent losses for which currently no models exist. The proposed model can also estimate the steady-state throughput of long flows. We also present a sensitivity analysis and empirical models which investigate the effects of various factors such as loss rates, window limitation, delayed acknowledgments and packet size on TCP latency and quantify and isolate their individual contributions. Our model captures the effect of timeouts and slow start phases which occur anywhere during the transfer and uses a more accurate model for the slow start phase, leading to a more accurate estimation of TCP latencies.