Hello all,
I have been experimenting with different disk layouts/volume sizes
for DB (on RS6000 M80, SSA) and want to accomplish two things with
this attempt : up the volume size to 4GB (from 2GB) and move from
raid-10 to JBOD and then mirror with TSM. Simple, I thought, set
up new volumes, define to TSM and then delete old ones - voila,
TSM moves the data ... I've done this before but not for a long time.
But, it is taking approximately 40 minutes for each 2GB volume to move
and that is way longer than I would expect. The DB is 100GB and is
currently on two raid-10 volumes. On watching iostat as it goes, it
indicates that is reading from one raid-10 set (fine), and writing to
the new disk volume ... but as well it is also writing almost as much
to each of the two raid-10 sets (that make up the TSM DB). I didn't
expect that to happen.
For comparison, when I have mirrored DB volumes, with 4 processes
running concurrently, I would see about 3-5 minutes for moving 8GB
- so 40+minutes for one delete dbv does seem excessive.
I have also done a little test on another box, without raid-10 and
that was much faster. Is there anything about the raid-10 that I am
missing?
I'm giving up for today, and will try again next week. If anyone has
any experience with this, particularly going from raid-10 to something
else, or a hint of what might be going on, I'd be glad to hear from you.
(TSM server level is 4.1.0.9 - not yet braved 5.1 in production ...
the little test was at 5.1.5.1)
Many thanks, Sheelagh
--
Sheelagh Treweek
Oxford University Computing Services
Email: sheelagh.treweek AT oucs.ox.ac DOT uk
Phone: +44 (0)1865 273205 Fax:-273275
|