Concurrency control in real-time, active database systems

Abstract

In the past few years, Active, Real-Time Database Systems (ARTDBs) have been considered as an interesting research area. Although, Real-Time database systems and Active database systems have been discussed and implemented respectively to support non-traditional applications. However, only few researches concentrate on the integration of active and real-time database systems and it is very useful for some applications like network management, computer integrated manufacturing (CIM), and automated air-traffic control. In this final year project, we will focus on the concurrency control (CC) of this new and popular research topic. A number of new problems are evolved in CC of active, real-time database systems. For example, conventional CC protocols are more concerned about the serializability but real-time database systems also focus on transaction deadline. Unfortunately, the situation is more complicated when real-time databases integrated with active characteristic. ARTDBs must not only respond to the external transactions but also the internal triggered events. Due to the triggering structure in ARTDBs, a dynamic CC algorithm is needed. If we simply apply existing conventional database or real time database CC protocols, a lot of CPU processing time will be wasted and transactions may not able to complete before their deadlines. We have intensively investigated current RTDBs and ARTDBs OCC algorithms, OCC-BC, OCC-TI, and OCC-APFO. However, the simulation result shows that the existing ARTDBs CC algorithm, OCC-APFO, does not outperform other CC algorithms. Besides, a new concurrency control protocol “OCC-Dynamic Wait” is proposed to serve for ARTDBs and it is compared with existing algorithms. We have built a simulation model to evaluate the performances of the CC protocols in different system loadings. It appears that the proposed concurrency control mechanism is tailor-made for ARTDBs and it has a better performance than OCC-APFO.