You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by "Jason A. Dour" <ja...@bcc.louisville.edu> on 1997/06/10 17:47:05 UTC
PCWeek article...
-----BEGIN PGP SIGNED MESSAGE-----
First off...congrats, guys! AG folk on the cover of PC Week? It's almost
like the release of 1.2 -- I thought it might never happen. ;) What a
motley collection of geeks; I'm so proud. 8)
Second...what is up with that "performance" chart they touted? They're
reputing IIS to TRIPLE the number of requests per second that Apache can
serve? Is it just me, or is this just...ummm...wrong?!?!
Please tell me their benchmarking is off! I don't think I could handle it
if a Microsoft product outperformed a piece of freeware. My view of the
universe would shatter. HELP! (Well, maybe it's not THAT important to
me...but it does worry me none the less.)
Anyone have any info/answers?
Jason
# Jason A. Dour <ja...@bcc.louisville.edu> 1101
# Programmer Analyst II; Department of Radiation Oncology; Univ. of Lou.
# Finger for URLs, PGP public key, geek code, PJ Harvey info, et cetera.
-----BEGIN PGP SIGNATURE-----
Version: 2.6.2
iQCVAwUBM512/Jo1JaC71RLxAQGvLQQAhyvSa9v6BcLLGHPANdvaeZIYMgFclKjN
P5pUQwkTtdSjLAvg1WzKWZFq5bITuJTFRDqIa8yu1Wo1PZ7zerfeOOya86cQ7S1o
PHnEEBiRBcslh1i4HBwVmcQrF+wCGLLxRoJkcLr1ZPLC+AJLsGFb9U15yyuvpPO1
gzOGGXIyMK0=
=vey5
-----END PGP SIGNATURE-----
Re: PCWeek article...
Posted by Chuck Murcko <ch...@topsail.org>.
Jason A. Dour wrote:
>
> -----BEGIN PGP SIGNED MESSAGE-----
>
> First off...congrats, guys! AG folk on the cover of PC Week? It's almost
> like the release of 1.2 -- I thought it might never happen. ;) What a
> motley collection of geeks; I'm so proud. 8)
>
> Second...what is up with that "performance" chart they touted? They're
> reputing IIS to TRIPLE the number of requests per second that Apache can
> serve? Is it just me, or is this just...ummm...wrong?!?!
>
> Please tell me their benchmarking is off! I don't think I could handle it
> if a Microsoft product outperformed a piece of freeware. My view of the
> universe would shatter. HELP! (Well, maybe it's not THAT important to
> me...but it does worry me none the less.)
>
> Anyone have any info/answers?
>
I think those were old numbers using 1.1.x on an untuned solaris x86
system. We may well see some new 1.2.0 numbers soon, from a Linux system
running a P6 with some reasonable tuning.
Hard core UNIX programming? Hmmm. 8^)
--
chuck
Chuck Murcko
The Topsail Group, West Chester PA USA
chuck@topsail.org
Re: PCWeek article...
Posted by Marc Slemko <ma...@worldgate.com>.
On Wed, 11 Jun 1997, Dean Gaudet wrote:
> On Wed, 11 Jun 1997, Marc Slemko wrote:
> > Make it threaded.
>
> I'm not convinced this is a huge win. On Solaris sure, where thread
> creation time is fast and process creation time is slow. But on linux?
> Not likely. process creation is fast, and thread creation is actually
> the same operation. What's freebsd's thread creation speed relative
> to process creation?
Dunno, but you can have a lot more threads hanging around by default
without worrying about them eating as many resources.
Just because Linux sucks... <sigh> I think you will find that Linux is
somewhat unique.
Re: PCWeek article...
Posted by Dean Gaudet <dg...@arctic.org>.
On Wed, 11 Jun 1997, Marc Slemko wrote:
> Make it threaded.
I'm not convinced this is a huge win. On Solaris sure, where thread
creation time is fast and process creation time is slow. But on linux?
Not likely. process creation is fast, and thread creation is actually
the same operation. What's freebsd's thread creation speed relative
to process creation?
It opens up more optimizations though. But I'm doubting that IIS
implements any of them...
I get a new machine tomorrow to tinker with. I'll start doing what
benchmarks I can.
Dean
Re: PCWeek article...
Posted by Marc Slemko <ma...@worldgate.com>.
On Wed, 11 Jun 1997, Jason A. Dour wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
>
> I'm responding to a gazillion of PCWeek posts...
>
> On Tue, 10 Jun 1997, Marc Slemko wrote:
> > I can't speak for this particular article, but the typical comparison is
> > done by someone without a clue about Unix or configuring anything.
>
> That's apparent...but it does bother me that IIS "out of box" is
> supposedly *three times* faster than Apache "out of box." Is there
> nothing we can do about this, even documentation-wise?
Make it threaded. Honestly, I think that is one of the biggest things.
The biggest overhead comes from configuring it correctly with number of
processes. Oh, and make it run on an OS that has right around zero
protection from programs doing stupid things.
Re: PCWeek article...
Posted by "Jason A. Dour" <ja...@bcc.louisville.edu>.
-----BEGIN PGP SIGNED MESSAGE-----
I'm responding to a gazillion of PCWeek posts...
On Tue, 10 Jun 1997, Marc Slemko wrote:
> I can't speak for this particular article, but the typical comparison is
> done by someone without a clue about Unix or configuring anything.
That's apparent...but it does bother me that IIS "out of box" is
supposedly *three times* faster than Apache "out of box." Is there
nothing we can do about this, even documentation-wise?
> BTW, did you know that, according to
> http://www8.zdnet.com/pcweek/news/0609/09apache.html :
>
> Anyone unfamiliar with hard-core Unix programming will not be
> able to get Apache running, said users.
PAH! I'm by far not a "hard-core" Unix programmer...but I do well
with Apache. 8) Of course, I'm just Hard Core(tm)(or "Punk As Fuck" or
"Punk Rock" or...), though, so maybe that's why Apache is easy for me. ;P
On Tue, 10 Jun 1997, Chuck Murcko wrote:
> I think those were old numbers using 1.1.x on an untuned solaris x86
> system. We may well see some new 1.2.0 numbers soon, from a Linux system
> running a P6 with some reasonable tuning.
According to the graph, they tested IIS 3.0 for NT 4.0, Apache
1.1.1 on RedHat Linux 4.1, and Apache 1.2b11 on RedHat Linux 4.1. Order
of performance? IIS. 1.1.1. 1.2b11. The graph is on page 18 of the
June 9th, 1997 PC Week...the one with the AG geeks on the cover.
On Tue, 10 Jun 1997, Brian Behlendorf wrote:
> I'm rather unhappy that despite my best efforts to represent otherwise,
> the server is still seen as for Unix gurus only. And the chart showing
> speed lagging behind IIS is an embarrassment.
Well, technically, the article touting Apache as "for UNIX nerds
only" was a *different* article from the one featuring AG members. This
article of "It's Not Easy. So What?" was a "PC Week Labs benchmark."
Obviously, they don't know their benches from their arses.
> If someone wants to write a rebuttal, I am almost positive we could
> get it printed.
I'd be happy to give it a try. I'm in an expository mode at the
moment with several other projects. I could fit a rebuttal in. Should I
word it from the AG as a whole, or make it more personal? Which would
everyone rather see? I give "from the AG" a BIG +1.
On Tue, 10 Jun 1997, sameer wrote:
> I told our marketing dept to write a rebuttal saying, "Apache
> kicks ass and is easy to use." I don't think we've gotten a chance to
> see the article yet though.
Or should we let Sameer's folk write it?
ohhhh,PCWeek,you'regoingDOWN!
Jason
# Jason A. Dour <ja...@bcc.louisville.edu> 1101
# Programmer Analyst II; Department of Radiation Oncology; Univ. of Lou.
# Finger for URLs, PGP public key, geek code, PJ Harvey info, et cetera.
-----BEGIN PGP SIGNATURE-----
Version: 2.6.2
iQCVAwUBM56ap5o1JaC71RLxAQHetQP/ZKkvtSqPAkWHKPpxEE/vkrD6Zvj/eFC8
LyLcbvBtUkAuOfdGgWejNxf2VjyGZ7XVksXu5p27zHJCRJix6rAK0fGi1iRZfSr6
WV7pIVashC+wqcQec53h6c7m0/u81GcQgjwD6HGf7wQTuoqiQYB0043KTXBysRmt
CJvg5SLxatI=
=TkSd
-----END PGP SIGNATURE-----
Re: PCWeek article...
Posted by Brian Behlendorf <br...@organic.com>.
On Wed, 11 Jun 1997, Chuck Murcko wrote:
> Brian Behlendorf wrote:
> >
> > I'm rather unhappy that despite my best efforts to represent
> > otherwise, the server is still seen as for Unix gurus only. And the
> > chart showing speed lagging behind IIS is an embarrassment.
> >
> > If someone wants to write a rebuttal, I am almost positive we could
> > get it printed.
>
> Is this graph the one in the 5/27 ZDNet article?
I've only seen the online text, I thought someone said there was a
performance graph with the print article too...
Brian
--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com www.apache.org hyperreal.com http://www.organic.com/JOBS
Re: PCWeek article...
Posted by Chuck Murcko <ch...@topsail.org>.
Brian Behlendorf wrote:
>
> I'm rather unhappy that despite my best efforts to represent
> otherwise, the server is still seen as for Unix gurus only. And the
> chart showing speed lagging behind IIS is an embarrassment.
>
> If someone wants to write a rebuttal, I am almost positive we could
> get it printed.
>
> Brian
Is this graph the one in the 5/27 ZDNet article?
--
chuck
Chuck Murcko
The Topsail Group, West Chester PA USA
chuck@topsail.org
Re: PCWeek article...
Posted by Chuck Murcko <ch...@topsail.org>.
Brian Behlendorf wrote:
>
> I'm rather unhappy that despite my best efforts to represent
> otherwise, the server is still seen as for Unix gurus only. And the
> chart showing speed lagging behind IIS is an embarrassment.
>
> If someone wants to write a rebuttal, I am almost positive we could
> get it printed.
>
> Brian
Um, was there a chart I missed in the PC Week article? Or are you
referring to the untuned 1.1.3 chart from the ZDNet 5/27 article?
--
chuck
Chuck Murcko
The Topsail Group, West Chester PA USA
chuck@topsail.org
Re: PCWeek article...
Posted by Marc Slemko <ma...@worldgate.com>.
On Tue, 10 Jun 1997, Brian Behlendorf wrote:
>
> I'm rather unhappy that despite my best efforts to represent
> otherwise, the server is still seen as for Unix gurus only. And the
> chart showing speed lagging behind IIS is an embarrassment.
I did some benchmarking of Ambarish's NT port of Apache and compared it
against IIS 3.0 and Netscape Suitespot 3.0 beta or whatever version.
Under NT4.0. I didn't take any formal results, but my first impressions
are that Apache was around the same level of performance as the other two.
Not bad for a first try port.
Re: PCWeek article...
Posted by Brian Behlendorf <br...@organic.com>.
I'm rather unhappy that despite my best efforts to represent
otherwise, the server is still seen as for Unix gurus only. And the
chart showing speed lagging behind IIS is an embarrassment.
If someone wants to write a rebuttal, I am almost positive we could
get it printed.
Brian
Re: PCWeek article...
Posted by Marc Slemko <ma...@worldgate.com>.
I can't speak for this particular article, but the typical comparison is
done by someone without a clue about Unix or configuring anything.
Several times I have seen zd make comparisons between systems running on
completely different hardware and use them to say "this bit of software is
faster".
Apache does have more overhead in some situations and does require more
careful tuning because of its process based model instead of a threads
based model.
BTW, did you know that, according to
http://www8.zdnet.com/pcweek/news/0609/09apache.html :
Anyone unfamiliar with hard-core Unix programming will not be
able to get Apache running, said users.
Wow. I would have never guess that most of the people in the world
running webservers are familiar with hard-core Unix programming. Hell, I
am not familiar with hard-core Unix programming. <g>
On Tue, 10 Jun 1997, Jason A. Dour wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
>
> First off...congrats, guys! AG folk on the cover of PC Week? It's almost
> like the release of 1.2 -- I thought it might never happen. ;) What a
> motley collection of geeks; I'm so proud. 8)
>
> Second...what is up with that "performance" chart they touted? They're
> reputing IIS to TRIPLE the number of requests per second that Apache can
> serve? Is it just me, or is this just...ummm...wrong?!?!
>
> Please tell me their benchmarking is off! I don't think I could handle it
> if a Microsoft product outperformed a piece of freeware. My view of the
> universe would shatter. HELP! (Well, maybe it's not THAT important to
> me...but it does worry me none the less.)
>
> Anyone have any info/answers?
>
> Jason
> # Jason A. Dour <ja...@bcc.louisville.edu> 1101
> # Programmer Analyst II; Department of Radiation Oncology; Univ. of Lou.
> # Finger for URLs, PGP public key, geek code, PJ Harvey info, et cetera.
>
> -----BEGIN PGP SIGNATURE-----
> Version: 2.6.2
>
> iQCVAwUBM512/Jo1JaC71RLxAQGvLQQAhyvSa9v6BcLLGHPANdvaeZIYMgFclKjN
> P5pUQwkTtdSjLAvg1WzKWZFq5bITuJTFRDqIa8yu1Wo1PZ7zerfeOOya86cQ7S1o
> PHnEEBiRBcslh1i4HBwVmcQrF+wCGLLxRoJkcLr1ZPLC+AJLsGFb9U15yyuvpPO1
> gzOGGXIyMK0=
> =vey5
> -----END PGP SIGNATURE-----
>