Fast Incremental Backups with Percona Server and Percona XtraBackup

New Features
3 April 11:10AM - 12:00PM @ Ballroom B

Experience level: 
50 minutes conference
This session will describe the fast incremental backup feature, that is available in Percona XtraBackup when used with Percona Server. Regular incremental backups necessitate full data file scans, constraining the backup frequency/scheduling. To support fast incremental backups, Percona Server has implemented a feature--redo log tracking--that writes out changed data page ids to a new separate compact disk file. Percona XtraBackup then reads that file, removing the need for full data file scans during incremental backups. This makes the incremental backup I/O cost to be proportional to the actual changed data amount since the last backup. When there is little changed data, this results in massive I/O savings, making frequent incremental backups cheap, as will be show by performance results. Moreover, the server itself is able to read the data page id file and present it as a new INFORMATION_SCHEMA table. Querying this table provides new ways to get insight into database write patterns: find tables that have been written to, hottest tables, hottest pages, etc. The complexity of topics discussed will vary from simple fast backup setup descriptions to sample INFORMATION_SCHEMA queries to XtraDB implementation details.


Percona Server Lead, Percona
Laurynas joined Percona in 2011 as a member of the Percona Server, Percona XtraDB and Percona XtraBackup development group. Before coming to Percona, Laurynas was a PhD student at Aalborg University, in Denmark, where he did research on physical database indexes.