Resource-contention cognizant priority assignment in real-time database system

Abstract

After a long period studies on Concurrency Control (CC) in real-time database system (RTDBS), concurrency control such as Optimistic Concurrency Control (OCC) have been pushed to their limits and it is difficult to extract additional meaningful performance. Although there are some potential research areas such as priority consideration in real-time CC, unfortunately, studies in [1] concluded that for a variety of reasons, priority cognizant optimistic concurrency control (OCC) algorithms cannot significantly perform better than their priority incognizant relatives in real-time database systems. Although the priority cognizant OCC – OCC-APR proposed in [1] cannot effectively reflect the superiority of priority cognizant OCC, its performance has stimulated my thinking. We can change our focus from only considering conflict reduction to cooperating with reducing resource contention. Based on the idea of restart time estimation technique from OCC-APR, my proposed Early-Abort mechanism has produced. It can reduce the resource contention in the system to allow more transactions to complete its operations before deadline. Based on this premise, I analyze the issue of priority cognizance and method of reducing resource contention. Subsequently, I have designed a smart mechanism – EA with incorporation of OCC-APR. I will analyze the performance of OCC-APR+EA, as well as that of several other OCC algorithms across different system loading and data contention parameters. Surprisingly, OCC-APR+EA turns out an amazing result especially in using pessimistic approach. Therefore, in this project, I will explain why this occurs and conclude that EA mechanism can improve the real-time OCC performance. The contribution of this project is to study the issue priority cognizance and explore the effect of reducing resource contention with regard to real-time concurrency control mechanisms.