An efficient parallel optimization algorithm for the token bucket control mechanism

Abstract

The Token Bucket algorithm, one of the most widely used control mechanism nowadays, has been widely studied to ensure the QoS needs of various applications. However, one main drawback of current models of this algorithm is that most of them have focused on a single Token Bucket system. In this thesis, based on previous research efforts, we propose a parallel solution to the multiple Token Bucket model. We also develop a Reduced Memory Algorithm to decrease the algorithm's memory requirements at the cost of extra computation time. We test our parallel processing algorithm using two sets of traces. Our numerical results show that the model can effectively solve the multiple Token Bucket problems. Besides showing the benefits of using a parallel processing platform, our results also provide us with the guidelines to configure the parallel processing platform.