Calculation of the optimal timeout of executing a SQL query for multi-master replication system

Abstract

Many multi-master replication systems are based on a two-phase commit mechanism. One of its problems is waiting for a response from the database servers by the coordinator, which may not come due to problems on the network or on the database server itself, which leads to delays in the activity of the multi-master replication system. Usually, such problems are solved by introducing a timeout, but if its value is too large, the problem may not be noticed in a timely manner, and too small a timeout will be processed for SQL queries, which will cause an increase in the load on the network due to additional service messages. The paper analyzes the execution time of a SQL query on several servers in order to calculate the optimal timeout for a multi-master replication operation. Based on the results obtained, a module for calculating the optimal query execution timeout for multi-master replication systems has been developed. © 2022 Author(s).