Abstract
Starting from the premise that modern routers are not protected from aggressive and unresponsive flows, we define a new, almost stateless, active queue management scheme, called Prince. The basic idea is to protect the fair share of well-behaved flows. We adopt a game theoretic view, where incentive is given to the majority flow by dropping its packets at congestion. In order to find the majority flow, we focus on the queue of the router and detect the flow with the most packets in it. From a game-theoretic point of view, Prince manages to track and bound aggressive flows and favor socially responsible ones. Our results show that in this context Prince resembles MaxMin Fairness allocation. Finally, we also examine a streaming version of the algorithm that can be fine-tuned to any desired performance/accuracy trade-off point.