Intelligent Packet Discard Policies for Improved TCP Queue Management

Recent studies have shown that suitably-designed packet discard policies can dramatically improve the performance of fair queueing mechanisms in internet routers. The Queue State Deficit Round Robin algorithm (QSDRR) preferentially discards from long queues, but in-troduces hysteresis into the discard policy to minimize synchronization among TCP flows. QSDRR provides higher throughput and much better fairness than simpler queueing mech-anisms, such as Tail-Drop, RED and Blue. However, because QSDRR discards packets that have previously been queued, it can signficantly increase the memory bandwidth require-ments of high performance routers. In this paper, we explore alternatives to QSDRR that provide comparable performance, while allowing packets to be discarded on arrival, saving memory bandwidth. Using ns-2 simulations, we show that the revised algorithms can come close to matching the performance of QSDRR and substantially outperform RED and Blue. Given a traffic mix of TCP flows with different round-trip times, longer round-trip time flows achieve 80% of their fair-share using the revised algorithms, compared to 40% under RED and Blue. We observe a similar improvement in fairness for long multi-hop paths competing against short cross-traffic paths. We also show that these algorithms can provide good performance, when each queue is shared among multiple flows