On Wednesday, Jan 29, 2003, at 06:14 Australia/Sydney, Lee, Gary D.
wrote:
Currently running tsm 4.2.1 on solaris 8. Have approx 6gb database
split into two volumes, each mirrored.
Using standard filesystem based volumes. a few questions.
1. Is the performance increase worth the time and hastle to convert
to raw volumes?
IMHO, yes.
2. Can I mix and match? i.e. create a temp volume, delete dbvol1,
create a raw volume in its place, then delete the ttemp volume to move
the data back?
Yes - been there, done that.
3. Just how do you specify the raw partition? Tried it before
/dev/rdsk/c2t3d0 for example, but no joy.
You may want to create a disk slice that starts a small way into the
disk
so that TSM doesn't try to write over the disklabel. 32-64k should be
enough.
On both the Sun TSM servers I manage, all dbvols, logvols and disk
stgpool
volumes are Veritas Volume Manager logical volumes, a mix of stripes,
concatenation and RAID5 volumes. The RAID5 stgpool volumes are split
over
two FC-AL loops, and can sustain sequential I/O at around 120 MB/s -
tested
outside of TSM.
The one big advantage I see in going this way, is that all I/O bypasses
the
Solaris buffer cache, which results in a performance gain and frees up
RAM
for other purposes - in my case, we've devoted more RAM to TSM.
Cheers,
--
Paul Ripke
Unix/OpenVMS/DBA
101 reasons why you can't find your Sysadmin:
68: It's 9AM. He/She is not working that late.
-- Koos van den Hout
|