Toward a Differentiated-Service Enabled Parallel Packet Switch

Parallel packet switch (PPS) is aimed to scale the switching capacity. Basic PPS architectures perform uniform load balancing of aggregated traffic to achieve high throughput, but it fails to support differentiated QoS for various classes of traffic flows. As the necessity of providing differentiated QOS in the switching facilities is obvious, the goal of this paper is to provide a QOS enabled parallel packet switch architecture which supports guaranteed-delay (GD), guaranteed-loss (GL) as well as best-effort (BE) class of service. Several QoS architectures are proposed and their main properties are studied under various scenarios. Backpressure, speedup and adaptive rate shaping are the three main mechanisms used in the proposed architectures. Simulations show that by using central backpressure, which is complex in implementation, or distributing backpressure from output multiplexers to input demultiplexers in conjunction with using adaptive rate shapers in input demultiplexers for controlling rate of loss-sensitive and best-effort traffic, we can reduce average delay for guaranteed-delay packets considerably as well as keeping loss threshold for loss sensitive traffic, with very little reduction in utilization.