Tuan A Trinh
Budapest University of Technology and Economics
The Internet has been a huge success since its creation in the early 70's. It has a big impact on the way we interact and communicate. As the Internet evolves, it is shared, used by millions of end-points and many kinds of applications. They compete with each other for the shared resources and their demand for resources (such as bandwidth) is growing rapidly. As a result, congestion at certain points of the network is inevitable. The TCP protocol suite was originally designed to control congestion in the Internet and to protect it from congestion collapse.
Basically, TCP is a closed loop control scheme. Congestion in the network is fed back to the source in the form of losses (Reno-like versions) or delay (such as TCP Vegas) The source then reacts to the congestion signal from the network by reducing its transmitting rate. In other words, we can consider packet loss and high queueing delay as the cost of (aggressively) sending packets into the network. The higher the rate, the higher the cost (certainly, the relationship is not necessarily linear in nature), given a fix network. Furthermore, as the Internet has been gradually transforming from a government sponsored project to a private enterprise (or even a commodity), the economics of the Internet becomes more and more important issue. Recent projects such as FIND (Future Internet Network Design) and EuroNF (Network of the Future) also place these issues as of first order concern. Consequently, Internet connectivity and services will have to confront issues of pricing and cost recovery. In this perspective, the cost of congestion can be in monetary form. Introducing cost of congestion into the network creates balance, stability and high utilization of resource usage. These issues involve understading and designing strategies for congestion control of today/future Internet. Game theory (and its companion Mechanism Design) provide us the useful tools to view the Internet congestion control problem from a different perspective.
In this seminar, we discuss how Game Theory can be applied to understand Internet congestion control and to design efficient/incentive compatible congestion control schemes for the (future) Internet. Recent state-of-the-art results are reviewed.
Finally, we address open issues and future directions of "game-theory-minded" Internet congestion control.