A Performance Analysis of Timed Synchronous Communication Primitives

Two algorithms for timed synchronous communication between a single sender and single receiver have recently appeared in the literature. Each weakens the definition of "correct" timed synchronous communication in a different way, and exhibits a different "undesirable" behavior. In this paper, their performance is analyzed, and their sensitivity to various parameters is discussed. These parameters include how long the processes are willing to wait for communication to be successful, how well synchronized the processes are, the assumed upper bound on message delay, and the actual end-to-end message delay distribution. We conclude by discussing the fault tolerance of the algorithms, and propose a mixed strategy that avoids some of the performance problems.