Experiences applying game theory to system design

Abstract

We applied techniques from game theory to help formulate and analyze solutions to two systems problems: discouraging selfishness in multi-hop wireless networks and enabling cooperation among ISPs in the Internet. It proved difficult to do so. This paper reports on our experiences and explains the issues that we encountered. It describes the ways in which the straightforward use of results from traditional game theory did not fit well with the requirements of our problems. It also identifies an important characteristic of the solutions we did eventually adopt that distinguishes them from those available using game theoretic approaches. We hope that this discussion will help to highlight formulations of game theory which are well-suited for problems involving computer systems.