Improving the Performance of Read-Only Transactions Through Speculation

A read-only transaction (ROT) does not modify any data. The main issues regarding processing ROTs are correctness, data currency and performance. Two-phase Locking (2PL) protocol is widely used for concurrency control with serializabilty as correctness criteria. Even though 2PL processes ROTs correctly with no data currency related issues, the performance deteriorates as data contention increases. To improve the performance over 2PL, snapshot isolation (SI)-based protocols have been proposed. SI-based protocols process ROTs by reading from a snapshot of the committed data and ignoring the modifications produced by the concurrent active transactions. Even though SI-based algorithms improve the performance of ROTs, both data currency of ROTs and correctness (serializability) are compromised. In this paper, we propose an approach to improve the performance of ROTs using speculation without compromising data currency of transactions and correctness. The proposed approach improves the performance of ROTs by trading extra computing resources without violating serializability as correctness criteria. The simulation results show that with the proposed protocol the throughput performance is improved significantly over 2PL and SI-based approaches with manageable extra resources.