In the studies of real-time concurrency control protocols (RT-CCPs), it is always assumed that earliest deadline first (EDF) is employed as the CPU scheduling algorithm. However, using purely (ultimate) deadline for priority assignment may not be suitable for distributed real-time database systems (DRTDBS) in which there exist different kinds of transactions, such as global and local transactions. In order to improve the performance, different priority assignment heuristics have to be used. In this paper, we have investigated the performance of different priority assignment heuristics for subtransactions in DRTDBS with optimistic concurrency control (OCC) protocol. It is found that the heuristics, which are suitable for distributed real-time systems, are not suitable for DRTDBS. We find that our proposed heuristic, which considers both deadline constraint and data contention, can give the best performance.
展开▼
