Fixed Priority SWAP scheduling policy with differentiated services under varying job size distributions
Michael Okopa, Hudson Okii · 2012
The importance of service differentiation has been recognized in a number of networking environments. SWAP, a non-preemptive threshold based policy has been proposed to approximate the optimal Shortest Job First (SJF) policy by introducing service differentiation between short and large jobs such that short jobs are given service priority over large jobs. However, SWAP cannot differentiate the service of jobs using an attribute other than their sizes, implying that it cannot guarantee the service quality of mission critical jobs or users especially during time of congestion. In this paper, we propose and evaluate a variant of SWAP called Fixed Priority SWAP policy that offers service differentiation so as to improve the performance of high priority jobs without appreciably degrading the performance of low priority jobs. The service differentiation attributes can be selected depending on which traffic the operator defines as more important such that the more important job takes a higher priority over the less important job. We evaluate Fixed Priority SWAP under an M/G/1 queue for job size distributions with varying coefficient of variation (CoV) to analyze the impact of service differentiation on high and low priority jobs. Our detailed numerical analysis reveals that Fixed Priority SWAP policy improves the performance of high priority jobs at the expense of little or no penalty for the low priority jobs depending on the threshold value.