performance test by Phoronix

Eric Sandeen sandeen at redhat.com
Sun Dec 21 23:58:15 UTC 2008


Michal Schmidt wrote:

> But I don't trust Phoronix. I guess I'll have to see if I can
> reproduce their results (not on the same hw, alas).

Nor do I.  lame mp3 encoding for filesystem perf tests?  please.

and:

http://www.phoronix.com/data/img/results/ext4_benchmarks/1.png

is completely utterly wrong and busted.  There is no test in bonnie++
which does a "4GB random delete" by default - and ext3 cannot delete 210
4G files per second anyway, which would be obvious to anyone who's tried
deleting a single DVD iso on ext3, for example.

What they're actually reporting from bonnie++ is seeks per second on the
sequential input test (and I'd have to dig to see what that even means.)

So they've misunderstood what the delete test does (it deletes empty
files not 4G files by default), and reported a number that is impossible
to anyone who's tried, due to sloppy test output parsing.

I let them know about this - fairly politely even! - but they apparently
don't care to update the site.

But since it's on the internets it must be true, and this kind of crap
will be taken as the gospel truth for the next decade.  Sigh.

Anyway, I don't put much stock in phoronix test reports at this point.

-Eric




More information about the devel mailing list