Low-latency guaranteed-rate scheduling using Elastic Round Robin

Packet scheduling algorithms in switches and routers will likely play a critical role in providing the Quality-of-Service (QoS) guarantees required by many real-time multimedia applications. Elastic Round Robin (ERR), a recently proposed fair scheduling discipline designed for best-effort traffic, is very efficient with an O(1) dequeuing complexity and, in addition, has better fairness characteristics than other algorithms of equivalent complexity. In this paper, we analyze ERR for guaranteed-rate services, and obtain an upper bound on its latency. We further show that the bound obtained in this paper is tight. Our analysis shows that ERR, in comparison to other scheduling disciplines of equivalent complexity, also has significantly better latency properties. The combination of fairness, efficiency and low-latency makes ERR an attractive scheduling discipline for both best-effort and guaranteed-rate services.