An efficient distributed concurrency control algorithm using two phase priority

Abstract

Many concurrency control algorithms for distributed real-time database systems have been proposed. But there isn’t a representative concurrency control algorithm for replication environment. In this paper, we propose an efficient concurrency control algorithm for distributed real-time database systems in replication environment. The main ideas of this paper are promoting priority and trading of data. Promoting priority is that the priority of the transaction that enters into voting phase is elevated. Trading of data is that a holder in the voting phase can lend holding data to other transactions. The proposed algorithm does not cause priority inversion. Therefore it decreases the ratio of restarting transactions and guarantees a transaction to commit at its maximum. Also to reduce blocking times of transactions, it permits data of a transaction in the voting phase to lend to the others. It is shown through the performance evaluation that the proposed algorithm outperforms the existing algorithms such as DO2PL_PA and MIRROR.