Randomized Mutual Exclusion with Constant Amortized RMR Complexity on the DSM

In this paper we settle an open question by determining the remote memory reference (RMR) complexity of randomized mutual exclusion, on the distributed shared memory model (DSM) with atomic registers, in a weak but natural (and stronger than oblivious) adversary model. In particular, we present a mutual exclusion algorithm that has constant expected amortized RMR complexity and is deterministically deadlock free. Prior to this work, no randomized algorithm with o(log n/log log n) RMR complexity was known for the DSM model. Our algorithm is fairly simple, and compares favorably with one by Bender and Gilbert (FOCS 2011) for the CC model, which has expected amortized RMR complexity O(log² log n) and provides only probabilistic deadlock freedom.