| 1 |
On Wed, 2006-02-01 at 07:54 -0500, Sean Cook wrote: |
| 2 |
> -----BEGIN PGP SIGNED MESSAGE----- |
| 3 |
> Hash: SHA1 |
| 4 |
> |
| 5 |
> > They are all different questions: |
| 6 |
> > |
| 7 |
> > * Hardware RAID vs Software RAID |
| 8 |
> > Hardware RAID offers more performance and often more sophisticated RAID |
| 9 |
> > features. |
| 10 |
That is debatable :-) |
| 11 |
For really high-end systems hardware raid (3ware, SAN box, ...) is more |
| 12 |
performant, but even on a modestly loaded Athlon64 you'll (usually) be |
| 13 |
I/O limited before CPU limits become visible |
| 14 |
|
| 15 |
> > Another advantage is that Hardware RAID controllers often offer you the |
| 16 |
> > opportunity to extend your RAID array beyond the usual 4 SATA |
| 17 |
> > interfaces. Depends on the card though. |
| 18 |
I'm running a software raid spanned over 3 controllers ... I'm not aware |
| 19 |
of a hardware-based solution that even comes close to that flexibility. |
| 20 |
|
| 21 |
> Back in the day... we were running the LSI MegaRAID controllers with |
| 22 |
> RAID 1 on 9gig SCSI disks, we were choking our application |
| 23 |
> (mysql/mod_perl) we took the MegaRAID controller out and performance |
| 24 |
> increased by about 25%. (note: this is a real world situation) |
| 25 |
Even if this is anecdotal - my experience is very similar. |
| 26 |
Software raid5 on an Athlon-1000 (basically a box of leftovers) takes |
| 27 |
~7% CPU to saturate a 100Mbit line. An older 3ware controller maxed out |
| 28 |
at similar speeds, slightly lower CPU load - but adding more RAM and a |
| 29 |
slightly faster CPU is in this case much cheaper and faster than buying |
| 30 |
a new controller. Also changing controllers might not be possible |
| 31 |
without wiping and rebuilding the array. |
| 32 |
|
| 33 |
But if you need a disk array with maximum performance I'd still suggest |
| 34 |
a hardware-based solution. |
| 35 |
-- |
| 36 |
Stand still, and let the rest of the universe move |
Archives