Modeling TCP-Vegas under On/Off Traffic

Adam Wierman
Takayuki Osogami
Jorgen Olsen

There has been a significant amount of research toward modeling variants of the Transmission Control Protocol (TCP) in order to understand the impact of this protocol on file transmission times and network utilization. Analytical models have emerged as a way to reduce the time required for evaluation when compared with more traditional evaluations performed using event driven simulators such as ns. In addition, when designed carefully, analytical models help researchers make design decisions about novel TCP mechanisms.

A wide variety of techniques have been applied to the problem of TCP modeling with a fair amount of success. These techniques range over renewal theory, Markov chains, and fluid models. However, until the recent seminal work of papers using fix-point methods, no modeling technique was able to mimic both the structure of a TCP source and the interaction a source has with the network. The fix-point methods take the novel approach of separating the modeling of network behavior from the modeling of the behavior within a TCP source, and then allowing the two to tune each other via feedback. The fix-point framework allows general network topologies to be analyzed, and issues such as the interpretation of end-to-end loss rates in multiple bottleneck networks are addressed in. In this paper, we generalize the framework based on a fixed point method introduced by Casetti and Meo in order to allow us to model TCP-Vegas connections. The framework, which uses a Markov chain to model the TCP source, has a few advantages over other fixed point methods: (1) it can model explicit details of TCP, making it possible to distinguish different flavors of TCP; (2) it allows modeling on-off traffic sources; (3) it gives the fraction of time that TCP spends in each state, from which we can evaluate the effectiveness of each mechanism of the protocol.