Max-min fair self-randomized scheduler for input-buffered switches

We consider self-randomized scheduling policies for input buffered switches. We provide a general architecture for the design of self-randomized algorithms. The common trend in the design of self-randomized schedulers is to use the number of backlogged cells as the weight function and to use an instantaneous cell arrival graph to generate a candidate matching for scheduling. We discuss some of the shortcomings of this approach, and introduce the total arrival graph as an alternative to the instantaneous arrival graph to improve performance. We also introduce the concept of max-min fair self-randomized scheduling algorithms. The idea here is to introduce self-randomized algorithms that can provide QoS by sharing the switch bandwidth proportional to some assigned weights. In order to study and compare the performance of the proposed scheduling algorithms, several simulations are carried out. Their results are provided and discussed.