TRANSACTION SCHEDULING WITH TEMPORAL DATA IN REAL-TIME DATABASE SYSTEMS

摘要

In a real-time database system (RTDBS), transactions are associated with timing constraints and data accessed by transactions may be valid only for specific time intervals. A correct transaction processing means that a transaction completes its processing before its timing constraint and uses data that are both absolutely and relatively timing-consistent. Previous research in RTDBS's focuses on scheduling transactions with absolute deadlines as the sole correctness criterion. No results have been reported for guaranteeing both the absolute and relative consistency of data. In this article, we explore new transaction-scheduling algorithms that attempt to guarantee both the absolute and relative consistency of data while keeping the transaction time constraints satisfied. The simulation results show that taking advantage of the data temporal consistency constraints in transaction scheduling improves the data consistency significantly. However, a trade-off with the improvement of performance metrics (e.g., percentage of missed transactions and percentage of absolute inconsistency) is needed.