Re: SATA II and PCI-X vs PCI for Areca ARC-1110 RAID Adapter

[Jukka Marin <jmarin%embedtronics.fi@localhost>]

On Wed, May 07, 2008 at 04:27:43PM -0400, George Georgalis wrote:
> On Wed 07 May 2008 at 11:41:55 AM -0700, Brian Buhrow wrote:
> >     Hello.  My suggestion is to check the performance, if possible, of a
> >hardware raid card versus raidframe in a given eenvironment.  The reason is
> >that I've run into a number of situations where raidframe out performs
> >hardware raid cards by a good margin, and this was somewhat of a surprise
> >until I started asking around, and found that, in fact, my experience
> >wasn't that unique.  In fact, I think if you google around for hardware
> >raid versus software raid, you'll find quite a few people hav had the
> >experience that I have.
> 
> this sounds rather contrary to what I'm familiar with... especially
> since this is a high end raid card. can you provide any references?
> a google does not confirm your assertion.

My ARC-1130 is much faster than any raidframe I have ever used (even in
RAID6 mode).

Under xen dom0:

# time dd if=/dev/zero of=foo bs=1m count=10000
10000+0 records in
10000+0 records out
10485760000 bytes transferred in 61.904 secs (169387438 bytes/sec)
0.029u 33.433s 1:01.92 54.0%    0+0k 127+325io 1pf+0w
# time dd of=/dev/zero if=foo bs=1m count=10000
10000+0 records in
10000+0 records out
10485760000 bytes transferred in 39.992 secs (262196439 bytes/sec)
0.030u 19.542s 0:39.99 48.9%    0+0k 0+0io 0pf+0w

domU, a different partition:

# time dd if=/dev/zero of=foo bs=1m count=10000
10000+0 records in
10000+0 records out
10485760000 bytes transferred in 68.653 secs (152735641 bytes/sec)
0.000u 39.113s 1:08.65 56.9%    0+0k 0+173io 0pf+0w
# time dd of=/dev/zero if=foo bs=1m count=10000
10000+0 records in
10000+0 records out
10485760000 bytes transferred in 34.276 secs (305921344 bytes/sec)
0.049u 21.457s 0:34.27 62.7%    0+0k 0+1io 0pf+0w

This system is not completely idle which may affect the results
(running several domU's with web and sql servers, innd, etc.).

  -jm