Simulating Synchrony on Top of Asynchronous Systems

Abstract

Synchronous distributed algorithms are easier to design and analyze than their asynchronous counterparts. Unfortunately, they do not work when executed in an asynchronous system. Hence, the idea to simulate synchronous systems on top of an asynchronous one. Such a simulation algorithm is called a synchronizer. First, this chapter presents several synchronizers in the context of fully asynchronous systems. It is important to notice that, as the underlying system is asynchronous, the synchronous algorithms simulated on top of it cannot consider physical time as a programming object they could use (e.g., to measure physical duration). The only notion of time they can manipulate is a logical time associated with the concept of a round. Then, the chapter presents synchronizers suited to partially synchronous systems. Partial synchrony means here that message delays are bounded but the clocks of the processes (processors) are not synchronized (some private local area networks have such characteristics).