A Slice-Based Method to Speed Up Join View Maintenance for Transactions

Abstract

With the increments of data volumes and user numbers, big data applications require higher transaction throughput but lower query latency for database systems. The materialized view accelerates analytical queries by trading space for query efficiency. Nevertheless, it has to be updated under transactional workloads to obtain up-to-second results. Unfortunately, the cost of view maintenance is expensive, which requires examining its maintenance strategies carefully. In this paper, we redesign the view maintenance strategy from the transaction perspective. Compared with conventional methods that compute the modifications of different operations separately, we implement a slice-based method that maintains the updates of several base tables with join relations in one transaction as the increments of a slice. Then we optimize the view maintenance process based on the slices such as avoiding invalid expression evaluation and base table access. We conduct experiments in PostgreSQL under CH-benCHmark. Experiments show that our method can increase transaction throughput by 17%–121%, reduce query latency by 30%–85%, and achieve 1.9$\times$ higher query throughput than those of conventional methods.