实时数据库中具有时态限制的事务调度算法

Transaction Scheduling Algorithm with Temporal Constraints in Real-Time Database Systems

在实时数据库中,事务对时态数据对象的访问在很多领域的应用日益广泛.目前实时数据库中的事务调度算法大多仅考虑数据的逻辑一致性,而忽略了满足时态的一致性.本文提出了临时数据截止期的概念,以定理的形式证明了低成本的预测算法的可行性,尽可能早地终止或推迟无法满足时态一致性要求的事务的执行,节省了计算资源以供其它事务执行.在此基础上提出了新的实时数据库中具有时态限制的事务调度算法TDDBPA(TemporaryDataDeadline-BasedPredictiveAlgo-rithm).通过与目前所知的有效算法比较,本算法在性能指标上(事务失败率、浪费的CPU时间)明显优于这些算法.

Temporal data object accessed by transactions of real-time database systems have been applied in diverse domains. However, almost all of the current transaction scheduling algorithms only take logic consistency of object into account, while ignoring temporal consistency. This paper presents the concept of temporary deadline, and proves the feasibility of low-cost predictive algorithm in form of theorem in order to abort or postpone those transactions that cannot satisfy temporal consistency as soon as possible, therefore, computation resources are saved to accommodate for other transactions. A new transaction scheduling algorithm with temporal constraints, called TDDBPA, is proposed based on strategies mentioned above in this paper. The simulation results show that performance of our algorithm significantly outperforms that of well-known algorithms so far.