Reclamations using LT04 Drives

[jrd31]

Hi

Has anyboy come across the problem we are experiencing reclaiming tapes using LT04 Drives.

We have a storagetek library and trying to reclaim LT04 tapes from within the same library and finding that we are only able to reclaim two tapes over a period of 8 hours. It seems to be running about 5MB/S per drive during the reclaim process. We have looked at the tape drives and performing migrations from disk to LT04 we can get up to 40 to 50MB/s per drive. Backup Storage pools either go from LT02's to LT04's or LT04's to LTO2 drives and we can get about 35MB/s obviously due to the limitations of LT02 drives. We are using 4GB switches and 4GB HBA's so we have plenty of bandwidth.

We do not do co-location on these tapes due as it got to expensive using too many tapes so all types of file sizes are on these drives but we are increasingly finding that we are getting very low on scratch tapes and trying to find a resolution of increasing the number of tapes we can reclaim, we have about 100 odd LT04 tapes with less than 50% of data in a full state.

Thanks in advance.

 

[moon-buddy]

I can safely say that the 'slowness' is due largely to fragmented files. You mentioned that you backup from mixed sources (LTO2 and 4) and most likely from mixed environments. These alone would cause numerous source tape mounts when TSM tries to reconstruct as much as possible the reclaimed data.

 

[smccambly]

I agree, fragmentation kills the streaming features of the tape media.
If your concern is that the drives are being tied up by the reclamation process for too long then consider using a Reclaim Stgpool.
This feature was added to support single drive reclamation, but even with multiple tape drives it can be used as a way to buffer fragmented data tape reads & writes. The reads won't run too much faster but you won't be tying up the target drives during this time, and then the writes will transfer at the speeds you are used to seeing because the data is now coming from disk.

Scott McCambly