Analysis of locking policies in database management systems

Abstract

Consistency control has to be enforced in database management systems (DBMS) where several transactions may concurrently access the database. This control is usually achieved by dividing the database into locking units or granules, and by specifying a locking policy which ensures integrity of the information. However, a drawback of integrity enforcement through locking policies is the degradation of the global system performance. This is mainly due to the restriction imposed by the locking policies to the access of transactions to the database, and to the overheads involved with the management of locks. A framework for the quantitative analysis of the impact of these factors on the performance of DBMS is presented in this paper. In a first step, the main factors which determine the behavior of these systems are pointed out and analyzed independently. The results hereby obtained are aggregated in a second step to yield a global performance evaluation. Throughout this hierarchical modeling approach various analytical techniques are used and the results are illustrated by numerical examples. The paper concludes by pointing out the final results' sensitivity to some basic assumptions concerning transaction behavior and the need for more experimental studies in this area.