When it comes to backup strategies, many organizations face a bit of a paradox. The volume of an organization's data grows exponentially, but this ever-expanding dataset still must be backed up within a window that usually does not expand. There are several possible strategies for dealing with this problem, but one solution is to use System Center Data Protection Manager 2007 (DPM 2007) to decrease the amount of time it takes to produce a backup.
DPM 2007 is a continuous data protection solution. In other words, it doesn't typically rely on making a single monolithic backup late at night (aside from an initial synchronization). Instead, DPM 2007 creates a series of incremental backups throughout the day. These incremental backups are based on the Volume Shadow Copy Service, and occur nearly instantly.
The backup schedule isn't set in stone. Instead, DPM 2007 requires you to create a protection group that consists of the resources you want to back up. From there, you would create a backup schedule for the protection group based on your backup requirements.
Typically, a protection group is designed to perform a synchronization every 15 minutes and create three recovery points per day (although this schedule is fully customizable). What this means is that data is backed up incrementally every 15 minutes.
Recovery points can be thought of as restore points. In other words, to restore data you will have to select which recovery point you want to restore the data from. As I mentioned previously, far fewer recovery points are created than the number of synchronizations that DPM performs. For example, even if data is being synchronized every 15 minutes, it may be several hours between the creation of recovery points.
While this type of design may seem awkward and counterproductive, it actually helps to improve the backup's performance. The backups are incremental, so the actual volume of data being transferred to the server during a synchronization is usually very small. Since it doesn't take much computing power to process small chunks of data, it is far more efficient for DPM 2007 to receive frequent (but usually small) updates throughout the day than to perform a single backup once per day.
Performing synchronizations on such a frequent basis also ensures that if a protected resource (DPM 2007 speak for a protected volume or application) were to fail, then no more than a few minutes worth of data would be lost. Granted, the odds of a recovery point existing for the most recently synchronized data are slim, but recovery points can be manually created at any time, ensuring that you can always recover the most recent data if necessary.
Express full backups
While DPM 2007's overall approach to protecting server resources lends itself to speedier backups, the approach described in the section above can't always be used. For example, performing frequent, incremental backups tends to work really well for file servers, but databases can't usually be backed up in this way. That being the case, DPM 2007 periodically performs an express full backup.
As the name implies, an express full backup is similar to a regular full backup -- only faster. The initial express full backup of a protected resource takes roughly the same amount of time to complete as any other backup would. Subsequent express full backups, however, only copy the blocks that have changed since the previous express full backup was made.
Backing up a DPM server
One cause of time-consuming backups involves limitations in a tape drive's throughput. Although tape drive throughput has improved over the years, it is not uncommon for organizations to perform parallel backups to multiple tape drives as a way of decreasing the amount of time the backup takes to complete.
Although this approach works, it tends to be expensive to implement. DPM 2007 offers an ideal solution since backups are initially written to disk before being written to tape for long-term storage. What this means is that the backup window no longer matters because DPM 2007 is writing the data that it has already collected to tape. This process has no impact on the protected resources.
The only downside to creating tape backups in this way is that the process only copies the most recent data to tape. Tape backups do not contain individual recovery points. Therefore, if your DPM 2007 server were to fail, you would lose all of your recovery points. The only thing that you would have to fall back on would be the data that had been written to tape.
The best way around this problem is to get a second DPM 2007 server and use it to back up your primary DPM server. That way, if your primary DPM server were to fail, you can restore it using the secondary DPM server. This would allow you to protect all of your existing recovery points.
As you can see, Data Protection Manager 2007 can greatly expedite the backup process and eliminate the need to worry about backup windows. Even so, it is important to design your DPM 2007 architecture in a way that protects against the failure of the DPM server itself so that you do not risk losing your backups.
ABOUT THE AUTHOR
Brien M. Posey, MCSE, has received Microsoft's Most Valuable Professional Award four times for his work with Windows Server, IIS and Exchange Server. He has served as CIO for a nationwide chain of hospitals and healthcare facilities, and was once a network administrator for Fort Knox. You can visit his personal Web site at www.brienposey.com.