The classical Ricart-Agrawala algorithm (RA) has long been considered the most efficient fair mutual exclusion algorithm in distributed message-passing systems. The algorithm requires 2(N — 1) messages per critical section access, where N is the number of processes in the system. Recently, Lodha-Kshemkalyani proposed an improved fair algorithm (LK) that requires between N and 2(N — 1) messages per critical section access, and without any extra overhead. The exact number of messages depends on the concurrency of requests, and is difficult to prove or analyze theoretically. This paper shows the superior performance of LK over RA using extensive simulations under a wide range of critical section access patterns and network loads.
展开▼
