The consistency control problem for replicated data in adistributed computing system (DCS) is considered. An algorithm isproposed to dynamically regenerate copies of data objects in response tonode failures and network partitioning in the system. The DCS is assumedto have strict consistency constraints for data object copies. Thealgorithm combines the advantages of voting-based algorithms andregeneration mechanisms to maintain mutual consistency of replicateddata objects in the case of node failures and network partitioning. Thealgorithm extends the feasibility of regeneration to DCS on wide-areanetworks (WANs), and is able to satisfy user queries as long as there isone current partition in the system
展开▼