Abstract

Consider a distributed mutual exclusion algorithm fur processes ar-ranged in a ring network in which mutual exclusion is guaranteed by means of a token that is passed around the ring ( [6], [10], [12]). How can we determine that such a system of processes is correct? Our first attempt might be to consider a reduced system with one or two processes. If we can show that the reduced system is correct and if the individual processes are really identical, then we are tempted to con-elude that the entire system will be correct.. In fact, this type of informal argument is used quite frequently by designers in constructing systems that contain large numbers of identical processing elements. Of course, it is easy to contrive an example in which some pathological behavior only occurs when, say, 100 processes are connected together. By ex-amining a system with only one or two processes it might even be quite difficult to determine that this behavior is possible. Nevertheless, one has the feeling that in many cases this kind of intuitive reasoning does

Citations