Implementation and verification of high data availability on database

Abstract

This study provides a low-cost and high-availability database management system architecture for general Small/Medium Enterprises (SMEs) to backup database data access. To prove that the proposed architecture can support the high availability of the database, and can effectively avoid data loss in memory caused by failovers, this study applies the main test method of powering off the virtual machine and verified three cases on two commonly used databases MySQL and PostgreSQL: Case 1 proves that this study combines the database native disaster recovery mechanism to effectively achieve high availability of the database. Case 2 proves that it effectively controls the WAL (Write Ahead Log) of the PostgreSQL database and Redo log mechanism of the MySQL database, so that data correctness is maintained during failovers. Case 3 proves that it can analyze and control the timing of the database in writing data in the cache memory to the hard disk. This study also designed a failover process to avoid data loss during failovers due to no enough time to write the data in the cache memory back to the hard disk; and finally to realize the high-availability of the database management system architecture in a practical way.   Keywords: Database, DBMS, High Availability, Failover, DRBD