Opportunistic Scheduler Evaluation Using Discriminatory Processor Sharing Model

This paper studies the flow-level performance of a special family of weight-based opportunistic scheduler using discriminatory processor sharing (DPS) model. It is known that this family of schedulers can achieve any feasible long- term throughput vectors by the variation of its weights. The optimal weight setting problem under DPS model is studied by decomposing it into two subproblems. In each subproblem, a guideline as to how the weight should be chosen is derived. The usage of these two guidelines in analyzing the original problem is discussed via theoretical and simulation results.