Simulating Few by Many : Limited Concurrency = Set Consensus ( Extended

Perfection is the enemy of the good. Most backoff schemes wait until a single contending candidate survives. But it is “the last mile,” getting contention down from few to one, which is usually the most costly: backoff systems which wait on some fixed k > 1 perform better than those that require perfection, i.e. k = 1. This paper asks what tasks are read-write solvable “k-concurrently ” i.e. tasks where progress can be guaranteed when contention goes below k+1. That is, what good is backoff to k > 1? While backoff to k = 1 is indeed omnipotent and can guarantee progress on any task, what set of tasks can be guaranteed progress with k > 1? We show that the set of these tasks is exactly the set of tasks that are solvable wait-free with the availability of k-set consensus. It is now for the system designer to decide whether good is good enough.