DC-SFQ: An improved Stochastic Fairness Queuing algorithm

Abstract

Stochastic Fairness Queuing scheduling scheme resolved the connection level fairness problem and are widely used in traditional network, however, with rapid development of P2P services in internet, this classical algorithms failed in maintaining fairness among hosts. In this paper we proposed an improved SFQ algorithm: Dual Class SFQ to ensure the fairness among hosts using different type of services. In DC SFQ, periodical perturbed hash functions originally used in SFQ are modified into Bloom Filter to identify the high load P2P hosts. DC SFQ uses 2 classes of queues to separate the P2P hosts and normal hosts to guarantee fairness among hosts. By adjusting the scheduling weight between these 2 classes of queues we can achieve an ideal balance between the traditional network services and P2P services. Analysis and simulation results show that DC SFQ can solve the packet scheduling problem for hybrid traffic at the cost of acceptable increase of complexity.