Coordinated Backup and Recovery between Database Management Systems and File Systems

We consider a network of computers consisting of file servers and a Database Management System (DBMS) where a linkage is maintained, with referential integrity, between data in the DBMS and files in the file servers which are external to the DBMS. We present algorithms for performing backup and recovery of the DBMS data in a coordinated fashion with the files on the file servers. When a file is associated (linked) with a record in the DBMS, certain constraints are applied to support referential integrity, access control, and coordinated backup and recovery as if the file is stored in the DBMS. Backup of a referenced file is initiated when the file is linked. The file backup is performed asynchronously to the linking process so that the linking transaction is not delayed. In a typical scenario, when a database backup operation starts, all unfinished file backups are ensured to be completed before the database backup is declared successful. When a database is recovered to a state which includes references to files in one or more file servers, the DBMS ensures that the referenced files are also restored to their correct state in those file servers. However, since database backup and recovery are critical for database availability, the presence of an unavailable file server is tolerated during the database backup and recovery operations. Our algorithms for coordinated backup and recovery have been implemented in the IBM DB2/DataLinks product. The DataLinks concept is also part of the ISO SQL/MED standard [ISO00].