A NEW ALGORITHM FOR INCREASING FAULT-TOLERANCE OF DISTRIBUTED SYSTEMS

摘要

Existing solutions to improve availability and fault-tolerance of distributed systems suffer from extra added complications and overheads to these systems. Some of these solutions use object replication for achieving their objective and thus entail these overheads to control object replicas. This paper presents a similar solution, but uses a novel mechanism for the pessimistic control of object replicas, namely the primary server approach. The message overhead in this approach is considerably less than related solutions, leading to more improved system fault-tolerance and availability.