hello all,
I recently got access to an old box (P-II 200) with SCSI drives. so i
thought that I'd play around with software RAID. i may play with hardware
RAID later, but for now I'll just do software RAID.
I'm testing for performance, so I'm looking at RAID-0 (striping with no
redundancy). how much of a performance gain (i know it's application
dependent, but give me a range, perhaps qualified with what kind of
application gives that range) can i expect?
I've done a few tests already (single-disk, no RAID at all, and i've
completed two different RAID-0 tests with different chunk sizes).
I'm not seeing much difference though.
my setup:
1. /dev/sda is the root filesystem and it is not part of the array.
2. /dev/sdb, sdc and sdd are part of the array and are mounted
under /var
3. the test involves:
1. restoring a postgres database (psql --command "\\i db.dump").
2. do a select which basically does a full table scan of a table
with close to 2million records.
3. do a select which joins two tables (one of them the large table
from #2).
4. #1 doesn't create any indexes, so #2 and #3 don't benefit
from indexes at all, basically everything involves full
table scans. In this test, i create the indexes (this takes
about an hour).
5. do a complex 5 table select, this takes advantage of the
indexes.
4. all tests are timed. before each test is performed, i run a program
that flushes the disk caches (well, it does enough disk access that
the disk caches don't hold any of the postgres data in the caches
anymore, although i don't stop and restart the postgres daemon, so
*that* might hold some data in its caches still.
6. i format /dev/md0 as xfs. i may test later with ext2 and ext3.
how much of a performance gain would i be getting with RAID-0 over
three disks? i thought i might get at least 100% (things would go twice
as fast), but i'm not seeing that. the best i'm getting is around 25%.
is it maybe that postgres' CPU requirements are what's slowing me down?
i could download some benchmarks that just torture test the drives, but
i'm using postgres to benchmark because that's what i do that really
needs the performance boost. everything else i've got runs fast enough
even on slow hardware, except for the database (which runs well enough,
but i'd like it to run even faster).
tiger
--
Gerald Timothy Quimpo gquimpo*hotmail.com tiger*sni*ph
http://bopolissimus.sni.ph
Public Key: "gpg --keyserver pgp.mit.edu --recv-keys 672F4C78"
Whom the mad would destroy, first they make gods.
Bernard Levin.
--
Philippine Linux Users' Group (PLUG) Mailing List
[EMAIL PROTECTED] (#PLUG @ irc.free.net.ph)
Official Website: http://plug.linux.org.ph
Searchable Archives: http://marc.free.net.ph
.
To leave, go to http://lists.q-linux.com/mailman/listinfo/plug
.
Are you a Linux newbie? To join the newbie list, go to
http://lists.q-linux.com/mailman/listinfo/ph-linux-newbie
