Communication Delay in Hypercubic Networks with LRD Traffic

摘要

Realistic network traffic can exhibit the Long-Range Dependent (LRD) feature which is characterised by a hyperbolically decaying correlation function and has strong effects on the performance of communication networks. As the convergence of simulations to a steady state under LRD workloads is often very slow, analytical models become cost-effective and versatile tools that can help designers to investigate system performance. This paper designs an analytical performance model for computing communication delay in adaptively routed hypercubic networks in the presence of LRD traffic. The model is derived in the context of pipelined circuit switching. The tractability and reasonable accuracy of the analytical model make it a practical and cost-effective evaluation tool to study the performance behaviour of hypercubic networks under LRD traffic.