I'm running TSM on Solaris 9, version 5.3.1.2 and have experienced an
extreme slowdown in performance when writing data to any collocate
sequential storage pool. FILE or tape.
Server setings:
movebatchsize=1000, txngroupmax=2048, movesizethresh=2048
This happens both with client backups, pool migrations and move data's
within the same pool.
However, if a very large file is being written, it streams well for the
duration of that large file.
The slowdown has something to do with a pause between each batch of small
files.
Here's some stat's around this.
random disk pool -> Collocate=filespace pool performance
3:20pm up 9 day(s), 1:43, 3 users, load average: 1.47, 1.32, 0.99
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
6.2 0.0 1587.2 0.0 0.0 0.6 0.0 97.9 0 6 sd144
I just updated tape pool collocate=no and waited a minute and we have fast
write speed again.
3:23pm up 9 day(s), 1:46, 3 users, load average: 0.65, 1.03, 1.24
r/s w/s kr/s kw/s wait actv wsvc_t asvc_t %w %b device
171.8 0.0 41541.9 0.0 0.0 0.7 0.0 4.0 0 67 sd144
Notice the higher cpu load as well.
no other processes were running in TSM or on host during this.
Has anyone else seen this? Is it a 5.3 issue or maybe 5.3.1.2?
I placed a call in, seems they are swamped today with other calls at IBM
Tivoli HQ.
Thanks
Matt G.
|