SAF-PS: Starvation Avoidance for Priority Scheduling

Conventional priority-based queuing is regarded as one efficient way of imposing hierarchy on disparate traffic when it comes to delay, jitter, size etc... The basic method involves having multiple queues at the processing node, built atop a heap structure, to cater to different priority traffic. The latter is then channeled through the media, one queue/priority at a time in decreasing/increasing order depending on the classification used. However, the downside of such an algorithm arises when the network experiences high load from high priority traffic. This case results in the buffer queues being filled up fast enough leaving no time to flush the lower priority queues over the transmission lines, thus holding up some of the traffic indefinitely; a nuisance referred to as starvation. This paper presents a dynamic scheme aimed at starvation avoidance for priority scheduling (SAF-PS), while preserving the integrity of the priority algorithm. The results obtained in terms of decreased drop rate and average waiting time will demonstrate the advantages of the developed scheme.