Concurrency control in distributed real-time database systems
Student thesis: Doctoral Thesis
Related Research Unit(s)
|Award date||1 Oct 1994|
Concurrency control in real-time database systems (RTDBS) is complicated by the requirement of maintaining database consistency and at the same time satisfying the timing constraints of the transactions. These two goals are difficult to achieve at the same time because the two scheduling for these two purposes are incompatible. As a result, the timing constraints of transactions in a RTDBS are more difficult to be satisfied. The problem is even more difficult to solve in distributed real-time database systems (DRTDBS). Additional considerations on integrating concurrency control protocols with different distributed transaction management algorithms, management of replicated data and resolution of distributed deadlocks must be taken into account. It is the objective of this thesis to develop efficient deadlock-free real-time locking protocols for aperiodic single-site and distributed RTDBS. The protocols take the real-time priorities of the transactions in scheduling data to them and in resolving lock conflicts between transactions with different priorities. The importance of the concurrency control protocols on system performance as compared with the real-time CPU scheduling algorithm has been identified and the factors affecting their roles are examined. The relative performance of the new protocols is studied extensively with the use of simulation models and compared with other well-known real-time locking protocols. The effect of different factors such as system workload, transaction deadline constraints, deadline distributions and proportion of read/write operations, on their performance have been investigated. Because of the importance of satisfying the timing constraint of real-time transactions, attempts are also made to release the constraint of serializable protocols with lesser restrictive protocols. Two new locking protocols based on epsilon serializability have been proposed for different types of aperiodic RTDBS. For DRTDBS, the effect of additional factors such as data distribution, degree of data replicated and the number of sites in the DRTDBS, on the performance of the new protocols are also studied.
- Real-time data processing, Distributed databases