DO-WF2Q: delay-optimised WF2Q packet scheduling

Packet scheduling is an essential element of any QoS enabled packet-switched network. (QoS provisioning and packet scheduling play important roles in present and future Internet applications.) In terms of packet fair queueing, worst case fair WFQ (WF²Q) is better than weighted fair queueing (WFQ) in that it can provide a bounded worst-case fairness index (WFI). The delay property of these two queueing schemes are studied. To do this, the delay index (DI) and the worst-case delay index (WDI) are defined to measure the discrepancy between the delay using these queueing schemes and that obtained using generalised processor sharing (GPS). Simulations show that in some circumstances WF²Q has higher DI than that using WFQ. The reason why WF²Q sometimes experiences higher DI is analysed, and based on this, an adaptive packet queueing algorithm termed delay-optimised WF²Q (DO-WF²Q) is proposed. This algorithm aims to improve the delay index (DI) while maintaining the WFI bound provided by WF²Q. Verifications and simulations show that DO-WF²Q can reduce the DI while maintaining the WFI bound, and at the same time retaining the same complexity as WF²Q.