Why do database concurrency control studies contradict each other? This paper critically examines the assumptions underlying various concurrency control algorithms for database management systems, aiming to reconcile seemingly contradictory performance results reported in recent studies. The authors employ a comprehensive model of a database environment to compare three different concurrency control approaches under various modeling assumptions. These approaches represent extremes in how transaction conflicts are handled. The study analyzes the impact of assumptions related to database system resources, transaction restart modeling, and the amount of information available to the concurrency control algorithm. By demonstrating how differences in these underlying assumptions explain the seemingly contradictory performance results, the paper offers valuable insights for database system designers and researchers. The research helps determine how realistic the various assumptions are for actual database systems.
Published in ACM Transactions on Database Systems, this paper fits directly within the scope of the journal, focusing on concurrency control, a key topic in database systems. By analyzing various assumptions and their impact on performance, the paper contributes to the ongoing development and optimization of database technologies.