Performance evaluation of a fair fault-tolerant mutual exclusion algorithm
Résumé
This paper presents an efficient and fair fault-tolerant token-based algorithm for achieving mutual exclusion. It is an extension of the Naimi-Trehel algorithm that uses a distributed queue of token requests and a dynamic tree. In case of failures, our algorithm tries to recover the requests' queue by gathering intact portions of the one which existed just before the failure. Thus, fairness of token requests is preserved despite failures. Furthermore, the use of broad- cast is minimized when rebuilding the dynamic tree. Experiment results with different fault injection scenar- ios show that our approach presents a fast failure recovery and low message broadcast overhead.