On mutual exclusion in faulty distributed systems

In this paper, we present an extension of the fault-tolerant mutual exclusion algorithm of [1] to make it deadlock-free when a node requesting critical section or when a node executing its critical section fails. The proposed algorithm requires <i>O</i>(&radic;<i>tn</i>) messages per critical section entry (in a network of <i>n</i> nodes) and <i>O</i>(<i>t</i>&radic;<i>tn</i>) messages when a node fails (<i>t</i> &minus; 1 is the upper bound of the node failures).