Optimistic Concurrency Control based on Cache Coherency in Distributed Database Systems

Optimistic concurrency control highly takes advantage of parallelism because there is no delay by lock. Unfortunately, I/O operations by transactions and cache operations are delayed in the optimistic concurrency control. In order to reduce such delay, we combine cache coherency control and optimistic concurrency control rather than they operate in separate. In this paper, we propose optimistic concurrency control, which includes cache control in parallel database system. Two basic concurrency control schemes, direct validation scheme and reduced validation scheme are devised according to the degree of cache activity. Based on these basic schemes, we designed four integrated concurrency control algorithms: GSO, GMS, GMM, and LMM. Experimental results show that GSO, GMS and GMM that use global cache show higher performance than LMM that uses local cache. GSO that is a direct validation scheme has larger communication overhead than reduced validation scheme. So, GMS shows the best performance among other algorithms as the portion of communication overhead is large.