Analysis of Complex Interactions among Transport Service Protocols in a Heterogeneous Network Environment

The congestion control algorithm is the core of transport service protocol in the Internet. In utilizing network resources efficiently and fairly, and at the same time assuring network stability, the congestion control algorithms implemented at the end hosts play an important role. In recent years, many different networks with varying characteristics (optical networks, wireless networks, etc.) have become parts of the Internet. Different queue mechanisms (Drop Tail, RED, etc.) have also been widely deployed. As a result, network pipes of the current Internet may have different characteristics in terms of bandwidth, delay, and PER, etc. Routers of a network pipe may also provide different feedback (packet loss, queue delay, ECN, etc.) to end hosts. TCP’s AIMD algorithm can not work well on all of these network pipes, and a large number of different congestion control algorithms have been proposed for different kinds of network pipes.

As a result, many TCP implementations, with different algorithms, will coexist in the heterogeneous Internet. Most of these algorithms still have not been investigated thoroughly and a consensus is still far away. But OS providers are eager to adopt these new protocols for acquiring high throughput and attracting more users. For example, Microsoft Vista adopts Compound TCP and BIC-TCP is used by Linux in default. Within such a heterogeneous environment, different congestion control algorithms may simultaneously be used by hosts and their cross interactions are not desirable as it would result in premature termination of some flows and result in chaos in some cases.

In this project, we plan to analyze and evaluate their interactions thoroughly and systematically to enable end-hosts utilize network resources efficiently. In particular, our work will focus on interactions among congestion control algorithms for high speed networks and wireless networks since the Internet is becoming faster and more users are expected to access it through wireless networks. Simulation, emulation, and experiment will be used, and some models will also be built for analyzing their interactions theoretically.

The study would be very useful in the context of increase use of data transmission using 3G and other wireless devices, and if adopted we expect to bring in savings for the end users.

Next >