End-to-End Statistical Delay Guarantees using Earliest Deadline First (EDF) Packet Scheduling

This paper develops a framework for statistically guaranteeing end-to-end delay bounds to envelopeconstrained connections transporting real-time traffic in a network of switches employing the Earliest Deadline First (EDF) packet scheduling scheme in conjunction with per-hop traffic shaping. The performance of an isolated EDF scheduler fed by markovian sources has been studied earlier in the literature; we extend the analysis to the multi-node setting in the presence of per-node traffic shaping. By making conservative approximations, we show that the analysis is tractable, and offers large performance gains over frameworks which offer deterministic services (the difference is as large as 25% for reasonably realistic traffic mixes). We also investigate the advantages of smoothing the traffic at the ingress to the network. Our framework, we believe, could be of great use in networks which statistically guarantee end-to-end delay bounds to real-time applications.