Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 13 Sep 2000 00:20:59 -0400
From:      Garance A Drosihn <drosih@rpi.edu>
To:        Stan Osborne <stan@craigslist.org>, billf@FreeBSD.ORG
Cc:        freebsd-advocacy@FreeBSD.ORG
Subject:   Re: advocacy/21238: poor performance; missed opportunities
Message-ID:  <v04210102b5e49ec3483e@[128.113.24.47]>
In-Reply-To:  <Pine.LNX.4.21.0009121913380.30467-100000@cnewmark.craigslist.net>
References:   <Pine.LNX.4.21.0009121913380.30467-100000@cnewmark.craigslist.net>

next in thread | previous in thread | raw e-mail | index | archive | help
At 7:34 PM -0700 9/12/00, Stan Osborne wrote:
>Your tone/attitude may explain why FreeBSD is not keeping up with
>the Linux hord.  Me, I'm trying to let the FreeBSD developer
>community know what is happening in the real world.  I'm the
>internal advocate for FreeBSD and one of the reasons we went to
>the effort to evealuate FreeBSD.  I am very dissapointed that
>after trying to justify switching to FreeBSD we could not.

The text in your initial PR was such that it was very easy to
assume it was a troll.  Very very easy.

>As for your claim that I did not provide specifices, do you need
>more information than my statement that we used FreeBSD 4.0?

I will believe that you are a sincere advocate of freebsd in
your environment, and that your wording was merely a result
of genuine disappointment.

That said, you really did give close to zero useful information
for anyone who might look into why your results were so poor.
You said that your search code was three times faster under
linux.  Okay, so what can anyone do about that?  Nothing. Why?
Because we don't have a clue what your special search code
does.  Maybe some autoconfigure script in YOUR (unnamed) package
just picked a bad choice for running under freebsd, and IF we
knew what was going on, someone could fix that choice and get
"your search code" (which is never identified) to run much faster.

What hardware are you running?  Maybe it's a bad driver for
some particular piece of hardware.  Now it would be nice if
freebsd beat linux on every kind of hardware, but realistically
that just ain't going to happen any time soon (not for EVERY kind
of hardware).  It might be that by spending $40 for a alternate
PCI card, freebsd would have done much much better, and compare
favorably to linux on the same hardware (ie, linux would also
have the same alternate card).  But, well, we can't possibly
consider that, because you didn't mention a single thing about
your hardware.

Perhaps it is a problem with SMP ... but you don't say if you
have a multi-processor box.  If it IS a multi-processor box,
then there have been a lot of developments since 4.0 which
would make a big difference.  Especially in the 5.0-current
branch, where there has been a MASSIVE amount of SMP-related
work in the past few months.  Perhaps someone is ALREADY
addressing the very problem you ran into, but you don't say
anything that might give us that clue.

You then said that the gdbm libraries were very slow.  How slow?
You don't say.  Slow at doing WHAT?  You don't say.  Reading?
Writing?  You don't say.  How big a database were they slow
with?  You don't say.  One big database, or many small databases
open at the same time?  You don't say.

Maybe you just made a poor choice in the kernel you built for
freebsd.  Did you include softupdates?  Now maybe you are very
expert, and are insulted that I would even ask such a dumb
question.  However, I don't know you from any of the other
50,000 people who post here, and IN YOUR PR you do not give
any useful configuration information.  If you are using 4.0,
then maybe you didn't add in softupdates, and maybe adding
that in would have made a big difference.  Unfortunately, we
do not know, because you did not say anything.

There are people using freebsd who are very interested in real
world benchmarks, because their livelyhood depends on fast and
reliable results from their OS.  These people, in fact, are
some of the main driving forces behind freebsd development.

However, any analysis requires information.  I haven't even
scratched the surface of issues which MIGHT have effected you,
and MIGHT have been easily addressed once they were known.
Were we all supposed to engage in a game of 20,000 questions,
just because you were disappointed?

I am quite willing to believe that you honestly would have
liked freebsd to have compared better to linux in your testing,
but the fact still is that you did not provide much useful
information.  You did not even offer to provide information if
anyone were to contact you.  In fact, you explicitly stated that
the "environment" was "no longer available".  That implies you
have no way of getting useful information, even if someone were
interested in some details.  The "fix" you explicitly offered
was to "use linux", which is another statement that just slams
a door on the idea of any discussion.

In short, all your PR did was complain.

I realize this message is not very helpful, but all I am trying
to do is explain why your original PR did certainly seem like
a "troll" for some flame war.  There are so many details you
COULD have said, and did not say, that it was pretty hard to
take the PR seriously.


---
Garance Alistair Drosehn           =   gad@eclipse.acs.rpi.edu
Senior Systems Programmer          or  drosih@rpi.edu
Rensselaer Polytechnic Institute


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-advocacy" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?v04210102b5e49ec3483e>