We are trying to provide highly available, quickly recoverable SQL2000 databases on a W2K advanced server. We are using MSCS to tackle the availability angle but are currently in a quandary over quick recoverability with, in particular, site failure in mind. Are you able to describe best practices in for both availability and recoverability? I note that Microsoft is advocating stretched clusters to provide the recoverability but I think this space is immature and poses a high level of risk.
Since your main concerns are availability and recoverability, I suggest using a mirrored/Business Continuance Volume (BCV) configuration. The mirrored protection (RAID-1) will provide you with the highest level of availability and performance.
The BCV is a logical hardware mirror that can be synchronized and split from the primary mirrored volumes to provide little to no disruption in your database during backup. Yes, you will need more storage but if you do run into a problem and need to restore the last backup you are restoring from disk, not tape. You can also create multiple BCVs to provide multiple backup copies. Before a new backup is required, the BCV data can be moved to tape for long-term storage.
When configuring your storage subsystem mirror the primary volumes. Then, set aside enough standard devices to use as BCVs. Since the BCV is a standard logical device, you can mix and match drives. You might want to spread your database across more 36GB drive spindles and use more cost-effective 73GB drives for the BCVs.
Several manufactures including EMC ( http://www.emc.com/), IBM ( http://www.ibm.com/) and HDS ( http://www.hitachi.com/) offer this BCV option on their hardware platforms. The BCV products fall under the trade names of TimeFinder, SnapShot and ShadowImage respectively. Check their Web sites for the latest innovations.
This was first published in June 2001