Limitations of high availability technology
HA is often perceived as the ultimate in data protection, particularly when the primary and backup servers are separated geographically. But there is a gap in this protection. The types of threats that HA and tape backup solutions address are not the only data risks companies face today.
For example, if someone accidentally deletes a file, the HA replicator will immediately repeat the deletion on the backup server. Likewise, if a virus or a malicious individual alters data in a way that misrepresents the truth (while still adhering to all the rules enforced for that field by the system), the replicator will blindly replicate the destructive update to the backup server. In these and similar cases, there should be a way to rapidly revert to the prior state of the data. HA software alone does not offer this option.
Tape-based backups are not a solution to this problem. Although they make it possible to revert to the state of the data at the time of the last backup job, that moment might be well before the desired recovery point. The database's journaling feature might allow a rolling forward to a point immediately before the data was corrupted, but that is typically a time-consuming, labor-intensive and error-prone process.
Continuous data protection offers another option
The solution is Continuous Data Protection (CDP), which is available in both standalone products and as a feature in some HA products. While various CDP technologies differ in their specific features and options, viewed from a high level, CDP comes in two flavors: True CDP and Near CDP. True CDP captures all data writes, transfers them to a secondary disk and stores each update independently. Using True CDP, one can "undo" data updates and additions by recovering the data to any point in time.
Near CDP differs from True CDP in that one can recover only to specific points in time. For example, Near CDP may copy data to the secondary disk only when a file is saved or closed. Two benefits of this approach are that it reduces the amount of data transmitted and every available recovery point is "clean" (that is, no transactions will be caught in mid-flight). The drawback is that, in some cases, this can result in recovery points being spaced at intervals of several hours or more. This may not be adequate in many environments.
It's best to work together
The optimum choice of tape backup, HA or CDP options is often "all of the above." Tape provides a reliable off-site backup should all else fail. Combining tape with HA and CDP goes further and addresses all data corruption and availability issues. HA ensures that operations will not be interrupted by disasters or maintenance activities, without the need to perform cumbersome, time-consuming recovery operations after a disaster. Finally, CDP enables the quick restoration of data to a point in time prior to its corruption. Thus, the combination of tape-based backups, HA and CDP is the best way to provide seamless protection against data loss, data corruption and system downtime.