Statistical Traffic Regulation at the Network Edge

Abstract

In order to increase bandwidth utilization, next generation communication networks could offer a Quality of Service (QoS) to active connections in a statistical sense rather than deterministic sense. For this to be practical the presence of new regulation algorithms at the network edge would be required. These new algorithms would be designed to bound incoming traffic flows with predetermined statistical descriptors. Here, we propose a novel yet simple regulation algorithm which bounds an arrival process with the statistical traffic descriptor known as exponentially bounded burstiness (EBB). We show that by setting the parameters of the algorithm appropriately, the arrival process can be regulated so as to bound the flow with another statistical descriptor---the effective bandwidth. Both of these statistical descriptors have a well-developed network calculus related to bandwidth utilization for a requested statistical QoS metric. Our algorithm allows us to apply with confidence the EBB and effective bandwidth network calculus to traffic flows within the network core.