In this paper, we study resource management models and algorithms that satisfy multiple performance objects simultaneously. We realize the proportional fairness principle based QoS model, which defines both delay and loss rate requirements of a class, to include fairness, which is important for the integration of multiple service classes. The resulting Proportional Fairness Scheduling model formalizes the goals of the network performance, user’s QoS requirement and system fairness and exposes the fundamental tradeoffs between these goals. In particular, it is difficult to simultaneously provide these objects. We propose a novel scheduling algorithm called Proportional Fairness Scheduling (PFS) that approximates the model closely and efficiently. We have implemented the PFS scheduling in Linux. By performing simulation and measurement experiments, we evaluate the delay and loss rate proportional fairness of PFS, and determine the computation overhead.
