CONTENTION RESOLUTION WITH CONSTANT THROUGHPUT AND LOG-LOGSTAR CHANNEL ACCESSES

For decades, randomized exponential backoff has provided a critical algorithmic building block in situations where multiple devices seek access to a shared resource. Despite this history, the performance of standard exponential backoff is poor under worst-case scheduling of demands on the resource: (i) subconstant throughput can occur under plausible scenarios, and (ii) each of N devices requires Omega(log N) access attempts before obtaining the resource. In this paper, we address these shortcomings by offering a new backoff protocol for a shared communication channel that guarantees expected constant throughput with only O(log(log* N)) channel accesses in expectation, even when packet arrivals are scheduled by an adversary. Central to this result are new algorithms for approximate counting and leader election with the same performance guarantees.