Grimes Gripes About Backup Software

InfoWorld columnist Roger Grimes gripes about backup software in his Security Advisor column [ “Backup Software: Bah Hambug” on Infoworld.com]. Some of the stuff he complains about are situations we find ourselves in ever so often – the bad tape drives, the dumb backup software that’s supposedly getting smarter every day, expensive technical support from backup software vendors, and above all – bad backups!

Grimes believes 20 to 40 percent of backups are bad, and yet a good number of sysadmins at organizations of all sizes continue to backup every day and store those tapes offsite, without ever testing them.

On Exchange backups, Grimes says “Want to restore your Exchange server to working condition after a fatal crash? Good luck unless you really knew what you were doing ahead of time, before the crash”.

Exchange Server 2007’s Database Continuous Replication feature that creates a replica of the Store on another volume of a standalone Exchange server (called “Local Continuous Replication” or LCR), or on the passive node in the case of a clustered deployment (dubbed “Cluster Continuous Replication” or CCR), should go a long way in alleviating those backup woes as far as Exchange is concerned. The ability to switch to a replica of the Store provides an amazing recovery capability without having to look for those backup tapes. In fact, unless you’ve got the tape with a good backup already loaded in the tape drive or sitting on your desk, you can be up and running before you can even find the tape!

Easy as it is to get excited about Database Continuous Replication, it’s important to remember that replication is not a substitute for a good backup strategy, and neither does it replace backups completely. What it does provide is the ability to recover quickly without restoring from tape, and reduces the need for frequent full backups. Additionally, it allows you to perform backups from the replica instead of the production Store that the users connect to – reducing the performance impact of backups.

If you are planning a clustered deployment, CCR provides automatic failover to the replica on the passive node (in case of standalone servers and LCR, switching to the replica is a manual process), and removes the requirement to use shared storage, eliminating the single point of failure common to Exchange clusters before Exchange Server 2007.

Tags
Administration   Exchange 2007