Many new products on the market and new and evolving technologies offer data-protection and disaster-recovery capabilities (especially in the wake of last September 11). It's a good time for companies to reassess their methodologies for data protection and disaster recovery.
The debate about whether tape is dead continues, kept alive by improved storage-product functionality such as disk-to-disk backup. If a company can implement disk-to-disk backup, for example, recovery times are faster and the company has a reduced need for expensive and less reliable hardware such as tape libraries, drives, and robotics.
And what about mirroring and snapshot technologies? Can and should such technologies replace traditional tape backup as the method of choice for data protection? Although these newer technologies have a place in any good data-protection and disaster-recovery configuration, companies shouldn't necessarily replace tape backup but rather complement tape with other technologies to provide different recovery functionality.
Disk mirroring has been available since the early days of mainframe computing. Mirroring is an effective way to protect data from certain error conditions and events. In particular, mirroring protects against online failure of storage devices—when one mirrored disk breaks, the other disk can provide the application with data without interruption. Mirrored copies of data are interchangeable, and access to mirrored copies of data is transparent to the user. However, although mirroring is effective in recovering from hardware failure, it's not a solution for data recovery. Mirrored copies of data are just that—mirrors—which implies that any corruption or user error perpetrated on the original will be reflected on the mirrored copy. In addition to these limitations, mirroring isn't a good solution for copies of data that must reside at geographically distant locations.
Several vendors and customers advocate using snapshots to ensure quick data recovery. Snapshot copies of data provide a fast way to make a copy of data; however, snapshot technology generally isn't effective for disaster-recovery purposes unless you combine it with backup. Two examples of the different methodologies you can use to deliver data recovery are software snapshots and split-mirror snapshots. Each vendor generates software snapshots differently. In some cases, snapshots don't actually make a copy of data but make a point-in-time copy of pointers to the data. Subsequently, when new blocks are written to existing files or databases, the new blocks are written to locations that existing snapshots don't reference. Because the snapshot copy references the location of the original blocks of data on disk, applications such as backup can use this input without the need to shut down critical applications.
Split-mirror snapshots consist of two or more mirrored copies of data that you can split at any time to create a coherent point-in-time snapshot. After you've performed a backup on the split-mirror copy, you can bring the copies back together for resynchronization. Although split-mirror snapshots provide excellent data protection, supporting this configuration might cost more than traditional tape backup, and resynchronizing copies after a split might involve significant overhead.
Snapshots and disk-to-disk backups are extremely useful for retrieving individual files or file systems when you need quick recovery. These technologies effectively provide near-online storage capabilities, and if you have the disk space to support the technologies, you can get extremely satisfactory recovery times.
However, you shouldn't use these methodologies alone to ensure data protection because when it comes to disaster recovery, tape backup continues to provide benefits that you'll find unavailable or difficult to achieve with mirroring or snapshots. Only with a secondary copy of data stored at a remote location can you truly protect an IT environment in the event of a significant disaster—for example, the destruction of a primary data center or its equipment. You can transport backup tapes to any location and retrieve data without regard to network connectivity or failed primary hardware. A tape copy of data is portable. Traditional tape backups also let you store and transport data securely if data is encrypted.
Of course, tape backup presents some problems too. It's complex, often requiring complicated policies and expert technicians to maintain them. Unlike with other data-protection technologies (e.g., split mirror, snapshots), systems administrators must ensure that they've backed up all data (all applications, all files) regularly. Even though a good backup ensures a coherent copy of data, recovery can be difficult and time-consuming.
Any method of data protection and disaster recovery has its pros and cons. In general, depending on the IT funds available, it makes sense to use a combination of technologies that provide quick recovery in the event of data corruption or user error and provide complete recovery in the event of a disaster.