You've elected to go the route of disk-based snapshots, which implies that you are looking for a recovery time objective and recovery point objective of minutes to hours in the event of full system failure. This is a good approach, and I would agree that a backup-to-tape solution would be hard-pressed to provide this level of recoverability.
However, as you've discovered, local snaps do not provide recoverability in the event of total site failure. But take heart, you have laid the foundation to extend this method to address your site recovery need. Most disk-based snapshot technologies offered by storage array vendors do have a way of replicating their snapshot data out to an alternate site, and usually over an IP WAN. Both EMC and NetApp offer this capability in their storage array solutions. I'm not sure what storage array vendor you've invested in, but you should discuss this capability with them. Of course, this you need a second storage array at your alternate site to act as a destination for the replication operation.
If this snapshot replication capability is not available or your recovery objectives in the event of a site disaster are not as stringent as required for local system recovery, you could consider a backup-to-disk solution that incorporates not only replication to an alternate site, but also data deduplication. Data deduplication is a "compression" algorithm that weeds out redundant blocks/segments of data and only keeps unique data blocks/segments with pointers that describe how to assemble the disassembled data from the blocks/segments that have been stored. This results in compression ratios on the order of 20:1 and many times as high as 100:1 for databases, and 200:1 for virtual machines files -- allowing the ability to utilize backup to disk at almost the price per gigabyte of tape. Most hardware-based deduplication storage systems have replication built into them.
If you elect to go this route, the news for you is even better because you're using TSM, which already has the ability to utilize disk in a seamless way. By implementing a backup-to-disk solution that incorporates deduplication, you can achieve fast recovery capability as well as the benefits of replication to an alternate site for restoration. Examples of storage appliances that utilize deduplication are Data Domain Restorers, Quantum DXi, and the newly announced EMC Disk Library 3D products.
Note on TSM: IBM has announced that it is planning on introducing deduplication as a process within TSM within 12 to 18 months. However, expect to build out a substantial TSM server in order to undertake this, and there is no word on whether there will be any replication capability from this feature. I would suggest that you would probably get more gains with a backup to disk appliance approach because if you decide instead to invest this heavily in TSM to incorporate deduplication, then you should probably be investing this heavily to solve your TSM backup to tape problem as well.
This was first published in June 2008