Synchronized update over distributed and parallel database

Abstract

To achieve synchronized update, two-phase commit protocol (2PC) has been broadly used for distributed databases. However, this is not enough for distributed and parallel database system (DPDBS), which has a distributed and parallel I/O interface and absolute peer to peer management among server nodes. Therefore, a distributed and parallel commit protocol (DP2PC) is proposed to improve the 2PC effectiveness. Transactions in inconsistent environment are successfully filtered. The performance is improved by taking consistency check of correlative data-catalog items into voting message, differentiating read-only-transaction from write-transaction and selecting the optimal service nodes. Further, our analysis and experiments indicate its adaptability to data environment with desirable performance.