COMMENT: In The Field
New Strategies Simplify BackupsBy Ari Kaplan
Innovative technologies speed up tasks and reduce costs.
Backing up data used to be simple. At night, when system use was low, you physically copied everything to tape. If you needed more space, you bought more. Backups might take several hours, but that didn't bother most people. If you ever had to restore the data, you just copied everything back into place. Life was good.
Now things aren't so simple. Data size and the costs to store backups have grown by double digits—but your enterprise's income and profits probably haven't. That once-small backup window threatens to dwarf the active production window, especially since your enterprise now handles transactions 24/7. If a disaster requires restoring all that data, it had better be done yesterday.
Now new backup technologies are making it possible to back up data more completely and economically, using less space, with simpler and faster restore procedures and intriguing side benefits. These backup technologies don't require faster drives, although better hardware helps. Instead, these technologies involve better ways of handling the data you want to back up. Best of all, they also work well in Oracle environments.
How good are the new backup solutions? In the past, recovering one terabyte of backed-up data might take up to 15 hours and require considerable manual handling of tapes. Now, recovering that same amount of data requires only minutes and is largely automated. Take Oracle Recovery Manager (Oracle RMAN). Before Oracle RMAN was introduced, backup required physically copying each database-related file, writing every byte. Oracle RMAN is far more clever. Instead of writing every byte, Oracle RMAN backs up only blocks of data that have changed, resulting in backup sets that are a fraction of the size of the original database, require fewer disks or tapes, and are faster to restore.
Snapshots are another approach to backups. This system of pointers saves the original backup information in one area and points to new blocks when data is changed. If recovery is necessary, you just select which set of pointers to use as production, resulting in a near-immediate recovery. Since the snapshot consists of pointers, it can require far less space than a full backup, and because you are handling pointers to data rather than data itself, you can restore terabytes of data in seconds rather than hours.
Creating a snapshot is also fast compared to physically copying data, with essentially no performance hit to the production system. As a result, you can do more-frequent (and more-granular) full backups of your data, which reduces the risk of not being able to recover completely. Imagine how much simpler and quicker it is to restore from hourly full backups instead of nightly incrementals with weekly full backups.
Using snapshots offers another bonus to shops that create test and development systems. In the past, you had to duplicate a production database on a separate system to test new development. This took a long time and required a lot of hardware. As a result, many development teams cut corners and did inadequate testing with incomplete data.
However, snapshots let you use the same pointers to backup data in your test systems without moving the actual data, so you can set up test systems in minutes rather than days, on a system that is more compact and economical. As a result, cloning full production data to test and development systems is far easier. Indeed, removing this bottleneck permits multiple test environments simultaneously, reducing time to deploy applications and freeing up database management and resources.
Deduplication is another approach to backups. This technology looks for duplication in data that is headed for backup and records only one copy of similar data. Suppose an e-mail arrives in your inbox (which requires a backup), and you forward it to colleagues (which also requires a backup). Come backup time, deduplication solutions will recognize the similar data, save only one copy, and use pointers to that one copy for any similar material. The upshot: far less space is required for backups. The process is invisible to both the user and the application—all they see is faster backups and less tape to manage.
How much can you save by changing your backup habits? Using this combination of technologies, the amount of data being moved for backup can be reduced by more than 90 percent.
Oracle's backup solutions, including Oracle RMAN and Oracle Data Guard, incorporate some of these strategies. Smart use of these solutions can simplify backup and restore tasks, save costs, and improve performance. Life is better.
Ari Kaplan (firstname.lastname@example.org) is president of the Independent Oracle Users Group (IOUG) and senior consultant at Datalink. He founded Expand Beyond, a leader in mobile IT software. He has been involved in Oracle technology since 1992.