How much can knowledge of delay model help chunked coding over networks with perfect feedback?

In this work, we consider the problem of designing efficient feedback-based scheduling policies for chunked codes (CC) over single-path (line) networks with stochastic (queuing) delay. The state of the art in such policies are random push (RP) and local-rarest-first (LRF), which outperform the original policy of CC, namely the uniformly-at-random policy, in terms of the expected throughput even without any knowledge about the delay model. To our knowledge, however, this work is the first attempt to discover how much better one policy can do in an ideal case with perfect feedback when the model of delay is perfectly known. Towards this goal, we propose a new policy, referred to as transmitted-innovation-maximizer (TIM), based on the expected number of innovative packet transmissions at each transmitting node of the network by the next transmission time given the feedback information from the receiving node about the received packets. Our simulations show that TIM provides significantly larger (tighter) lower bounds on the maximum expected throughput (compared to the tightest existing bounds provided by LRF and RP), and thus it can be considered as the newest benchmark in this emerging line of research.