Concurrency Control for Replicated Data in Distributed Real-Time Systems

The design and implementation of time-critical schedulers for distributed replicated database systems must satisfy two major requirements: transactions must be able to meet the timing constraints associated with them, and mutual and internal consistency of replicated data must be preserved. In this paper, we present a replication control algorithm, which integrates real-time scheduling and replication control. The algorithm adopts a majority consensus scheme for replication control and attempts to balance the criticality of real-time transactions with the conflict resolution policies of that scheme. The algorithm employs epsilon-serializability (ESR), a correctness criterion which is less stringent than conventional one-copy-serializability, to guarantee the robustness of the scheme. The performance of the algorithm is evaluated and compared with a real-time token-based algorithm. The algorithm is extended to use the notion of quorum consensus, and the effects of read quorum on the performance is investigated.