A NEW DYNAMIC VOTING ALGORITHM FOR DISTRIBUTED DATABASE-SYSTEMS

We consider the problem of keeping a distributed database system that has been partitioned because of site or communication link failures pa rtially operable while ensuring data consistency. A dynamic-voting-con sistency algorithm is proposed, and its correctness is demonstrated. T he proposed algorithm results in improved efficiency in executing read requests by not requiring a read quorum. This algorithm is effective in environments where the majority of user requests are ''read'' types of requests. Furthermore, the proposed algorithm results in efficient recovery by avoiding updating those data objects that are still curre nt. Under the proposed algorithm, the majority partition would be avai lable even if changes in the network topology take place at a higher r ate than the update rate, as long as only simple partitioning takes pl ace.