Книга: Distributed operating systems

4.5.3. Synchronous versus Asynchronous Systems

4.5.3. Synchronous versus Asynchronous Systems

As we have just seen, component faults can be transient, intermittent, or permanent, and system failures can be fail-silent or Byzantine. A third orthogonal axis deals with performance in a certain abstract sense. Suppose that we have a system in which if one processor sends a message to another, it is guaranteed to get a reply within a time T known in advance. Failure to get a reply means that the receiving system has crashed. The time T includes sufficient time to deal with lost messages (by sending them up to n times).

In the context of research on fault tolerance, a system that has the property of always responding to a message within a known finite bound if it is working is said to be synchronous. A system not having this property is said to be asynchronous. While this terminology is unfortunately, since it conflicts with more traditional uses of the terms, it is widely used among workers in fault tolerance.

It should be intuitively clear that asynchronous systems are going to be harder to deal with than synchronous ones. If a processor can send a message and know that the absence of a reply within T sec means that the intended recipient has failed, it can take corrective action. If there is no upper limit to how long the response might take, even determining whether there has been a failure is going to be a problem.

Оглавление книги