Hi All!

This is my first posting so Hello! I have multiple file system
resident databases (no boos/hisses please!) and am evaluating
performance in anticipation of moving to IBM's FAStT600 drives. I
have run tests on FS's blocked @ 8K and 16K (from servers with page
sizes @ 2K and 16K) and can find no clear-cut performance metrics that
would point to any particular blocking size(s). I tested using both
fast and slow bcp's (500,000 wide rows) and it looked like the 16K was
a little better, but not much.

Anyone have any experience with this hardware?? I am running
12.5.0.3/EBF 10973 ESD# on Sun 5.9.

Thanks and Hi again!

Sara ...

saradba (AT) hotmail (DOT) com (Sara Law) wrote in
news:b67ab85d.0404151052.1ec09277 (AT) posting (DOT) google.com:

Hi Sara,

Because you're using FS (DSYNC is off for the dataserver?), try to
increase the blocksize as large as possible. Perhaps 32 or 64K.
--
Pablo Sanchez - Blueoak Database Engineering, Inc
http://www.blueoakdb.com

Pablo Sanchez <honeypot (AT) blueoakdb (DOT) com> wrote

Thanks Pablo - I am definitely considering the higher blocking factor
.... I just wish I could justify it with some metrics ... the ones I
generated show mixed results ... and dsync is off for all my tempdb
devices but not the other customer db devices.

Sara ...

saradba (AT) hotmail (DOT) com (Sara Law) wrote in
news:b67ab85d.0404161010.1282b001 (AT) posting (DOT) google.com:

YW

If you have a repeatable benchmark, I'd run it using different
blocking factors.

Why are you using file system though? Your best write performance
will be using raw devices. (note: I'm not booing nor hissing!
--
Pablo Sanchez - Blueoak Database Engineering, Inc
http://www.blueoakdb.com