While tape-based backup approaches may still be optimal for long-term data retention, more and more database storage/backup administrators are moving from tape to disk technology for everyday data backup and recovery. This is not surprising as disk is faster to write to and faster to recover from, as well as more reliable overall.
A further advantage of disk backup is the ability to perform deduplication on the data being backed up, which reduces both backup/recovery times and – perhaps most importantly – data volumes. Deduplication usually takes place during data transfer to disk. Some tools ship with the disk hardware and others with backup software.
Many products are currently available that offer deduplication at the backup server level, which reduces network traffic between the backup server and the backup target but not between the database server and the backup server. Some products still perform deduplication post-transfer, at the backup target. While slower overall, this approach eliminates the chance that CPU-intensive deduplication processes will create a bottleneck between the backup server and the secondary storage target.
Emerging technology enables deduplication at the source (that is, on the server that hosts the database). Ultimately, deduplication will likely take place as a function of primary storage rather than as a backup/archiving function.
Like many new technologies before them, deduplication and disk-based backup/recovery fill a long-standing need for improved speed and reduced cost. But many DBAs remain skeptical, and debate exists about how best to use these solutions and their impact on data protection and the rest of the IT environment.
The problem many DBAs have with disk-based backup is that it is not specific to the database. Unlike traditional logical/physical, file-based backups performed with database tools, with disk-based backups the database is not “aware” that a backup is occurring, so its state at the time of the backup may be unknown. Therefore, it is important to test and confirm that your databases can recover and come back up after being backed up in this new way.
While cutting-edge disk-based recovery approaches can restore gigabytes of data in seconds, they generally can’t restore a single database table. In other words, disk-based backup doesn’t (yet) replace the logical backup methods that DBAs have used for years.
At least one product family, EMC RecoverPoint, addresses the need for logical backups by enabling “any-point-in-time recovery to protect data against loss or corruption.” RecoverPoint’s features include policy-based management. Data files can be isolated in logical groupings within the backup, to facilitate recovery of specific data.
A holistic approach to protecting your data can include both logical backups at the database level, combined with disk-based backups of your raw data.
For more information on what disk-based backup and deduplication approach is right for your business, contact Buda Consulting.