Performance of Concurrency Control Methods in Distributed Database Management Systems (Timestamp Ordering, Two-Phase Locking, Optimistic Scheme, Restart, Transaction Blocking)

167 p.Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1985.The effects of transaction restart overhead on the performance of concurrency control methods (CCMs) have been studied using the simulation approach. In order to do this, detailed simulation models for a representative sample of concurrency control mechanisms, namely Basic Timestamp Ordering (BTO), Basic Optimistic scheme (BOPT), Basic Two-Phase Locking (B2PL), and Conservative Timestamp Ordering (CoTO), were developed, and the performances of those CCMs in distributed database management system (DDBMS) were simulated with some care to get statistically significant results. The performances of the CCMs were then compared with respect to several measures. This study systematically investigated effects of major factors on CCM performance in a quantitative manner, whereas few quantitative results have been obtained in the past work on CCM performance.Based upon the simulation results and results from the analysis of effects of restart overhead, the transaction restart overhead is found to have a determining effect on the performance of CCMs in DDBMS in most cases.Qualitatively, the CCMs that use mainly a blocking mechanism to handle conflicting transactions, for example CoTO and B2PL, generally perform better, with respect to most performance indices, than the CCMs that use mainly a restarting mechanism, for example BTO and BOPT.U of I OnlyRestricted to the U of I community idenfinitely during batch ingest of legacy ETD