poor performance in LTO6 to LTO6 copy/move data

[Alexandre Razera]

First of all, excuse my english...
I have a bad situation in two customers that I have checked this.
When I try to copy some data, specifically small number of larger files, like exchange or oracle databases, the speed limit reported in windows Performance Monitor and math with select on tsm process is around 40MB/s. If I copy/move the same data to disk (disk pool or filepool) and back to tape later, in the both situations the speed is around 100MB/s.
My conclusion is that the problem is the transfer between drives, because from disk to tape/tape to disk, the bottleneck probably is the disk.

The configuration is similiar, in both costumers, X3650M4, 32GB RAM, 16 and 24 cores, 1 FC adapter DUAL PORT 8Gbit., windows 8 R2 and windows 12 R2, both up to date in windows, drivers and firmwares with windows update and IBM update express express

Tape is TS3200, 2 drives LTO6 HH FC, both with latest firmwares.

TSM is 7.1.0.0 and 7.1.0.100

LTO Tapes are new.

When I did the tests, I had stopped all sessions and processes, including db auto reorganize.

I do some tests with movabathsize and txngroupmax values but no significantly changes in performance results.

Any sugestions or ideas?

 

[ShovelMan]

Greetings!

I am in exactly the same boat.

When I go from Disk > LTO6 I can enjoy 100-160MB/sec without drama.

I can restore from LTO6 to disk at around 100-120MB/sec as well (on incompressible data, much more on highly compressible stuff).

But if I do a LTO6 > LTO6 stg backup I creep along at exactly 34.15 MB/sec. There are small dips and spikes in the trace (read/write performance trace from the library) but it makes a tabletop-level line at 34.15 MB/sec * however many processes are running.

Anyone have any clue? I can provide whatever details that would be helpful but here are the basics:

Server: Windows 2012, TSM 6.3.4, 256GB, 16 cores, 4 x 8Gb HBA ports
Tape: Ultrium 6 with latest firmware into fabric switches and zoned to TSM server.

I also see this behavior when doing NDMP-image stg pool backups, so I am concerned that it may be a drive-specific issue rather than a TSM issue.

Any help is appreciated!