An Augmented Version of the Chang & Roberts Algorithm

This project and the accompanying research article was created for coursework in the Distributed Algorithms & Systems course at the University of Glasgow. I participated in this exercise as part of a 5-person team.

Leader election algorithms are crucial mechanisms which distributed systems use to achieve coordination and agreement. Several mutual exclusion algorithms need to elect a leader and use a leader election algorithm to accomplish this. One such algorithm is the Chang and Roberts algorithm. In this project and its accompanying research article, an augmented version of the Chang and Roberts algorithm is proposed that achieves significantly better performance in a single-election worst-case scenario. The implementation of the two versions is then discussed, using Java RMI. Furthermore, an evaluation was carried out that illustrates the predictions of the algorithms’ complexity. The augmented algorithm sends 2N messages for a single election in both the worst-case and best-case scenarios, while the original algorithm has a worst-case complexity of 3N − 1.

A research article was produced as a result of this coursework. Please see the Research Article. This article will give some useful insight before exploring the code in the project's GitHub repository.