Implementing FIFO queues and stacks

Abstract

The cost of implementing FIFO queues and stacks is studied under two consistency condiions for shared memory multiprocessors, sequential consistency and linearizability. The cost measure is the worst-case response time in distributed implementations of virtual shared memory supporting one of the two conditions. The worst-case response time is very sensitive to the assumptions that are made about the timing information available to the system. All the results in this paper assume that processes have clocks that run at the same rate as real time and that all message delays are in the range [d−u, d] for some known constants u and d, 0≤u≤d.

If processes have perfectly synchronized clocks or if every message has delay exactly d, then the response time of a dequeue operation is at least d, for any sequentially consistent implementation of FIFO queues. This matches exactly an upper bound in which an enqueue operation is performed instantaneously and the response time of a dequeue operation is d; this upper bound implements linearizability. If clocks are not perfectly synchronized and if message delays are variable, i.e., u>0, then, for any linearizable implementation of a queue, the response time of an enqueue operations is at least Ω(u). In contrast, we present sequentially consistent implementation for this weaker timing model in which an enqueue operation is performed instantaneously, and the worst-case response time of a dequeue operation is 2d. (This algorithm is completely asynchronous and does not rely on any timing information.)

Similar results are proved for implementing stacks, with the pop operation playing the role of the dequeue operation, and the push operation playing the role of the enqueue operation.

A complete version of this paper appeared as Technical Report #672, Department of Computer Science, The Technion, Haifa, Israel, May 1991.