Re; ZFS Performance FreeBSD 9.0 vs. Openindiana

grarpamp grarpamp at gmail.com
Thu Mar 29 10:51:19 UTC 2012


Knowing of no better... I find Iozone quite nice.

However... I would first resolve these differences in
the test methodology before trying to compare
anything 74 revisions apart... and perhaps the
address width as well.

Version $Revision: 3.323 $
Compiled for 32 bit mode.
Build: Solaris10cc-64
     File size set to 4194304 KB

Version $Revision: 3.397 $
Compiled for 64 bit mode.
Build: freebsd
     File size set to 4194304 KB

Then try to see where the average 45% speedup is occuring.

> to run not only in cache
>  Processor cache size set to 1024 Kbytes.

128k could very well fit in a cache somewhere.
Such as say, the one above.

As shown on the iozone site, I'd try graphing the full range
of record sizes as a first baseline comparison.

And poke around the sysctls to see how the zfs parameters
match up.

> -c Include  close() in the timing calculations. This is useful only
>     if you suspect that close() is broken in  the  operating  system
>     currently  under test.

Really, you do?


More information about the freebsd-fs mailing list