Determining optimal weights for Generalized Processor sharing

Generalized Processor Sharing (GPS) is a simple mechanism to provide fair quality-of-service differentiation between heterogeneous application classes. In GPS packets of the different classes are backlogged in different queues and the different queues are served according to predetermined weights. Choosing these weights to optimize a certain objective function is in practice very hard, resulting from the notoriously difficult analysis of GPS systems. Using results from strict priority scheduling we derive theoretical bounds on when GPS is more optimal than strict priority. This way we minimize the necessary simulation effort to find the optimal weights