Performance analysis of TCP with nonpersistent sessions

Pr. Anurag Kumar

Indian Institute of Science, Bangalore, India


Most studies of implicit or explicit feedback control of elastic sessions consider only persistent sources which have an infinite backlog of data to send. In real applications, however, TCP sessions are nonpersistent, i.e., a new TCP connection starts, transmits some finite number of packets and closes the connection. This paper is about an analytical model for calculating the average bandwidth shares obtained by Additive Increase Multiplicative Decrease (AIMD) controlled sessions that arrive randomly, share a single (bottleneck) link, and then depart after transferring their volumes of data. In particular, we consider the following traffic model: file transfer requests arrive in a Poisson process, each request requiring the transfer of an independent and exponentially distributed volume of data, which we view as a fluid. We develop an analysis of the AIMD controlled rate evolution process, with arrivals and departures, and then use simulations to examine how well this model predicts the performance of TCP controlled sessions.

We also provide results on the usefulness of the simple Processor Sharing model for predicting TCP throughputs.

[Pr. Anurag Kumar]
[Indian Institute of Science, Bangalore, India]