Coordinators are used in distributed systems to optimize resource distribution and maintain consistency. This is implemented with a leader election algorithm to elect new coordinators when the existing leader has failed. The Bully algorithm is a popular approach that re-elects the node with the highest identifier as the leader. However, this algorithm involves a lot of message passing to find the highest available identifier, which imposes heavy traffic. This project aims to improve the bully algorithm by reducing the number of messages exchanged between the nodes by dividing them into two groups to limit the number of nodes involved. By halving the nodes involved, the evaluation of the improved algorithm results in a 75% reduction in the number of messages sent, compared to the original bully algorithm.

свернуть