Majority & Stabilization in Population Protocols
share
Population protocols are a distributed model focused on simplicity and robustness. A system of n identical nodes must perform a global task like electing a unique leader or determining the majority opinion when each node has one of two opinions. Nodes communicate in pairwise interactions. Communication partners cannot be chosen but are assigned randomly. Quality is measured in two ways: the number of interactions and the number of states per node. Under strong stability requirements, when the protocol may not fail with even negligible probability, the best protocol for leader election requires O(n ·( n)^2) interactions and O( n) states [Gasieniec and Stachowiak; SODA'18]. The best protocol for majority requires O(n ·( n)^2) interactions and O( n) states [Alistarh, Aspnes, and Gelashvili; SODA'18]. Both bounds are known to be space-optimal for protocols with subquadratic many interactions. We present protocols which allow for a trade-off between space and time. Compared to another trade-off result [Alistarh, Gelashvili, Vojnovic; PODC'15], we improve the number of interactions by almost a linear factor. Compared to the state of the art, we match their bounds and, at a moderate cost in terms of states, improve upon the number of interactions. Our results extend to a slightly weaker "stability" notion known as convergence.
READ FULL TEXT