Optimal Strategies for Spinning and Blocking

In parallel and distributed computing environments, threads (or processes) share access to variables and data structures. To assure consistency during updates, locks are used. When a thread attempts to acquire a lock but nds it busy, it must choose between, spinning, which means repeatedly attempting to acquire the lock in the hope that it will become free, and blocking, which means suspending its execution and relinquishing its processor to some other thread. The choice between spinning and blocking involves balancing the processor time lost to spinning against the processor time required to save the context of a process when it blocks (context switch overhead). In this paper, we investigate a model that permits us to evaluate how long a process should spin before blocking. We determine conditions under which the extreme cases of immediate blocking (no spinning) and pure spinning (spin until the lock is acquired) are optimal. In other cases, we seek ways of estimating an optimal limit on spinning time before blocking. Results are obtained by a combination of analysis and simulation.