Re: The results of my PostgreSQL/filesystem performance - Mailing list pgsql-performance

I need to step in and do 2 things:

First, apologize for posting inaccurate test results.

Second, verify that Sean is absolutely correct.  FreeBSD 4.8 was accessing
the drives in PIO mode, which is significantly lousier than DMA, which
RedHat was able to use.  As a result, the tests are unreasonably skewed
in favor of Linux.

The only thing that the currently posted results prove is that Linux is
better at dealing with crappy hardware than BSD (which I feel we already
knew).

I did some rescrounging, and found some newer hardware stuffed in a
corner that I had forgotten was even around.  I am currently re-running
the tests and will post new results as soon as there are enough to
be interesting to talk about.

In an attempt to avoid the same mistake, I did a timed test with dd(1)
to a raw partition on both Linux and FreeBSD to ensure that both systems
are able to access the hardware at more or less the same speed. The
results of this will be included.

I'm also gathering considerably more information about the state of
the system during the tests, which should answer a number of questions
I've been getting.

To the many people who asked questions like "why not try filesystem x
on distribution y" and similar questions, the answer in most cases is
time. I've pared the tests down some so they run faster, and I'm hoping
to be able to run more tests on more combinations of configurations as
a result. Also, I never intended for anyone to assume that I was _done_
testing, just that I had enough results for folks to talk about.

I'll post again when I have enough results to be interesting, until then,
I apologize again for the inaccurate results.

Sean Chittenden wrote:
>>What it still leaves quite open is just what happens when the OS has
>>more than one disk drive or CPU to play with.  It's not clear what
>>happens in such cases, whether FreeBSD would catch up, or be "left
>>further in the dust."  The traditional "propaganda" has been that
>>there are all sorts of reasons to expect PostgreSQL on FreeBSD to
>>run a bit faster than on Linux; it is a bit unexpected for the
>>opposite to seem true.
>
> Let me nip this in the butt before people run away with ideas that
> aren't correct.  When the tests were performed in FreeBSD 5.1 and
> Linux, the hard drives were running UDMA.  When running 4.8, for some
> reason his drives settled in on PIO mode:
>
> ad0s1a: UDMA ICRC error writing fsbn 1458368 of 729184-729215 (ad0s1 bn 1458368; cn 241 tn 12 sn 44) retrying
> ad0s1a: UDMA ICRC error writing fsbn 1458368 of 729184-729215 (ad0s1 bn 1458368; cn 241 tn 12 sn 44) retrying
> ad0s1a: UDMA ICRC error writing fsbn 1458368 of 729184-729215 (ad0s1 bn 1458368; cn 241 tn 12 sn 44) retrying
> ad0s1a: UDMA ICRC error writing fsbn 1458368 of 729184-729215 (ad0s1 bn 1458368; cn 241 tn 12 sn 44) falling back to
PIOmode 
>
> The benchmarks were hardly conclusive as UDMA runs vastly faster than
> PIO.  Until we hear back as to whether cables were jarred loose
> between the tests or hearing if something else changed, I'd hardly
> consider these conclusive tests given PIO/UDMA is apples to oranges in
> terms of speed and I fully expect that FreeBSD 4.8 will perform at
> least faster than 5.1 (5.x is still being unwound from Giant), but
> should out perform Linux as well if industry experience iss any
> indicator.

--
Bill Moran
Potential Technologies
http://www.potentialtech.com