> The storage is SATA, SCSI, and iSCSI attached to Linux and/or FreeBSD,
> UFS and ext 3 or 4 on a variety of i386 hardware. I'm not especially
> interested in which ones /should/ perform better, I want to measure
> which ones /do/ perform better. (And so I can tell my applications
> team that they are hallucinating when they blame performance on the
> file system.)
> 
> What tools can I use to measure disk performance?

That's a thin line.  I have, in the past, built ext3 on raid-5 saying how
fast it is, and then seen it outperformed by a single disk mirror.  Because
the single disk mirror is faster for teeny tiny operations, while the raid-5
is faster for larger sequential stuff.  Guess what type of usage the users
predominantly do ...  Now I use a stripe of mirrors instead of raid-5.

BTW, ext4 for certain operations like rm'ing or touching many files or large
files, is a bazillion times faster than ext3.  Likewise, ZFS for other
operations like backups assuming snapshots, also a bazillion times faster
than either of the above.

You can't get really good reliable numbers on real systems that are in use.
I think you should treat as suspect any result you're getting in a user's
real live environment.  In order to get reasonable results reasonably fast,
you need to yank the RAM out of the system (otherwise you'll just be reading
from your RAM cache during the read operations)...  Conversely, you need to
install maximum ram on all your production machines.  Huge difference.

I use iozone.  Please see below.  I wrote the precise iozone commands into
each of the raw results files, so there would never be any question how the
results were generated.  I think my results are irrelevant to you, except to
demonstrate what a huge variation your configuration can make.  Mostly this
could be of value to you, as a good demo of what to expect from iozone.  I
think the measure that you care about is Bob's method, Random readers,
Random mix, and Random writers, Compared to a single disk.
Hmmm...  Rather than post the 4 huge links, please just go to
http://nedharvey.com and see the "RAID Benchmarks" section.

_______________________________________________
Tech mailing list
[email protected]
http://lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
 http://lopsa.org/

Reply via email to