Sorry if this is a ramble, I'm just trying to find somebody with a simliar enviroment to compare notes.

We recently migrated a large number of systems from another backup/archive vendor to TSM. For various reasons Linux was the selected platform.

Our largest system has roughly 211 clients and a 76GB database. Performance is fair enough with one exception; backup sets.

For SLA reasons, we need to keep data for a minimum of 12 months offsite. I would assume that we would use a copy pool for this normally, but the issue is that we need to capture files that a client may delete which expires past the retain only version date. (i.e., client deletes file in January, but doesn't discover that they needed it until May. Retain only version is currently 30 days.)

In any event, backup sets were decided as the mean to accomplish this. Great in theory, but nodes with numerous small files seem to tie up tape drives for days at a time. Even generating a backupset for nodes that have filesystems consumed with one or two large files runs at 12-14MBytes per second.

My question is:

Does anybody eles have an enviroment like this? Suggestions? If I increase the retain only version to a larger number (say 365 days) then my fear is that the database will grow to an exponential size. I tossed around the idea of creating policies for "important" files that would have an extended retain only version, and just using that policy in the include list. The problem is that with the number of clients in question, we can't gaurntee that we'll always know where "important" data is stored.

Largest system is SUNV40Z w/ 2 Opteron 870 dual core, 4GB RAM, 2 QLA2310 HBA's, 300GB EXT3 filesystem on RAID1_0 for DB/logs , 5 200GB ext3 filesystems RAID3 for disk backuppool. Tape pools onsite are CDL virtual libraries with >50TB of disk, secondary tape pool is an STK silo with 4 Ultrium3 drives.