A solution to the distributed mutual exclusion problem in k-groups

Abstract

The authors consider a distributed system, DS, which consists of a collection of k distinct groups. Each group is a set of processes which have distinct identities, and overlapped groups have at least one process in common. The concept of groups would be appropriate in situations in which processes belong to the same group share a particular property. Moreover, k resources are available, each one in a unique instance and is assigned to one group at the same time. Throughout the paper they describe an algorithm to cope with the problem of the distributed mutual exclusion within the system DS. This algorithm makes use of a logical structure called k-rooted trees and k tokens. They study the performance of this algorithm in terms of the number of messages exchanged per critical section. They discuss process failures and loss of tokens on the algorithm and propose method for recover from these failures.