March 2003
Intermediate to advanced
912 pages
27h 17m
English
Consider n processes executing on n processors with no shared memory. A problem discussed in the literature is to devise an algorithm to implement a critical region within each process. As discussed in Chapters 7 and 16 any such algorithm must take into account the possibility of independent failures of the components of distributed systems, the absence of global time and the delay involved in communications between components.
In order to explore the characteristics of applications for which a distributed mutual exclusion protocol may be appropriate we shall start from an example of a centralized data object with distributed processing.
Assume that n processes are cooperating to process data in a ...