On optimizing token bucket parameters at the network edge under generalized processor sharing (GPS) scheduling

In this paper, we consider the case where non-linear traffic bounds are provided for traffic sources which share a link operating under a generalized processor sharing discipline. We consider the problem of searching for parameters for token bucket traffic shapers which provide linear bounds for the non-linear traffic bounding function in order to make use of results for traffic delay bounds which require a linear traffic bounding function, expressed in the form of token bucket shaper parameters. We formulate an optimization problem to obtain the parameters (i.e., bucket size and token generation rate) with the objective of minimizing a delay bound for a particular traffic source. This method can be used iteratively to obtain good delay bounds for a number of sources. Some typical numerical results obtained from the optimization model are presented. We also propose an alternate method, which we refer to as the composite delay envelope method.