In this thesis, the performance of a mixed concurrency control method(CCM), which combines the locking and the timestamp ordering, and the performance of the basic two-phase locking (B2PL) method, have been studied using the simulation approach. For this purpose, a slightly modified version of the original LOCK/TO was proposed to reduce an unnecessary message, called insurance, used for transaction commitment.
Based on the simulation results, the average response times of the two CCMs were found to be largely affected by the degree of conflicts among transactions. When all transactions were read-only, the response times of the two CCMs were always the same, since each transaction was processed only in its local site by the same locking rule. When most of transactions were read+write, the response time of the LOCK/TO was smaller than that of the B2PL, since the effect of the restart overhead on the response time of the LOCK/TO was smaller than that of the communication overhead and the blocking overhead on the response time of the B2PL. However, when the read-only transactions and the read+write transactions were moderately mixed, the response time of the LOCK/TO was larger than that of the B2PL, since the effect of the restart overhead on the response time of the LOCK/TO was larger than that of the communication overhead and the blocking overhead on the response time of the B2PL.
In addition, it was also shown that the mean interarrival time and communication time had effects on the response time. In particular, since the B2PL had much more communications to process transactions than the LOCK/TO, the response time of the B2PL was more sensitively changed than that of the LOCK/TO by the changes in the communication time.
