Throughput and Latency of Acyclic Erasure Networks with Feedback in a Finite Buffer Regime

Abstract

The exact Markov modeling analysis of erasure networks with finite buffers is an extremely hard problem due to the large number of states in the system. In such networks, packets are lost due to either link erasures or blocking by the full buffers. In this paper, we propose a novel method that iteratively estimates the performance parameters of the network and more importantly reduces the computational complexity compared to the exact analysis. This is the first work that analytically studies the effect of finite memory on the throughput and latency in general wired acyclic networks with erasure links. As a case study, a random packet routing scheme with ideal feedback on the links is used. The proposed framework yields a fairly accurate estimate of the probability distribution of buffer occupancies at the intermediate nodes using which we can not only identify the congested and starving nodes but also obtain analytical expressions for throughput and average delay of a packet in the network. The theoretical framework presented here can be applied to many wired networks, from Internet to more futuristic applications such as networks-on-chip under various communication and network coding scenarios.