lkml.org 
[lkml]   [1997]   [Jul]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: InfoWorld web server shootout
We (apache group) deal with this sort of "reporting" all the time.  In
this case I believe one of our members even offered to tune the server for
them, or at least point them at the documentation. I could be mistaking
it for another article. Had they:

- used Apache 1.2 (1.1.3 is way too unstable under high load)
- read the htdocs/manual/mod/core.html documentation
- tuned the Max/MinSpareServer, and MaxClients settings

They would easily have served several hundred hits per second. Apache as
"shipped" is tuned for the average web server -- probably less than 200k
hits per day. It tunes itself dynamically, but does so in a way that
doesn't swamp the machine if there's a quick load spike. So if you want
it to serve 200 hits per second, you'll have to hit it that hard for
probably 5 minutes before it'll be "tuned". Or you could change the
settings beforehand. This totally screws up people doing naive
benchmarking.

However I have doubts that at the top end Apache actually keeps up with
the netscape and microsoft servers. There are various reasons for this,
including the fact that most people have less than 10Mbits to fill, and
it's very easy to fill that, so it's never been a priority. I have plans
for the top end, the new development tree has improvements in it
already...

Finally, it's bunk to compare different OSs/application combos across
vastly different hardware (in vastly different price categories) and draw
any sane conclusions. There is absolutely nothing you can do about crap
reporting like this.

Dean

P.S. I speak for myself of course, these are my opinions.

On Mon, 7 Jul 1997, Michael Brennen wrote:

>
> I've cut out most of the article below; the full text is online at the URL
> listed. Infoworld did a Web server comparison, and to my surprise the
> Apache / Linux configuration did not do well at all. Considering what it
> was up against in the Netscape and M$ configurations, something must have
> been seriously wrong with the Apache / Linux test rig. BTW I don't know
> how a Linux Hardware Solutions 2000P platform is built.
>
> Infoworld is widely enough read that the article will be negative to
> corporate management. Since David Miller posted some Web benchmarks after
> reworking the networking code (which I don't remember in detail) and got
> strong results, I thought I would post this for comment from others that
> might have insight into the poor performance shown.
>
> -- Michael
>
> ---------- Forwarded message ----------
>
> http://www.infoworld.com/cgi-bin/displayTC.pl?/970707comp.htm
>
> Web platform solutions
>
> ------------------------------
>
> Big Blue déjà vu
>
> Web servers are now strong enough to be the core of your
> enterprise. But whether they're used for an intranet or a
> commercial site, hearty administration tools are key for
> managing your platforms.
>
> COMPARED
>
> Apache solution
> Apache 1.1.3
> Red Hat Linux 4.2
> Linux Hardware Solutions 2000P
>
> IBM solution
> Internet Connection Secure Server 4.2
> AIX 4.2
> IBM RS/6000 Internet Power Solutions F50
> Netscape LiveWire 1.1
>
> Microsoft solution
> Internet Information Server 3.0
> Windows NT Server 4.0 with Service Pack 2
> Active Server Pages 1.0
> FrontPage 97 2.0
> Index Server 1.1
> Seagate Crystal Reports for IIS 4.5
> HP NetServer LX Pro 6/200 SMP
> Microsoft Resource Kit for Windows NT Server 4.0
>
> Netscape solution
> Enterprise Server 3.0
> Windows NT Server 4.0 with Service Pack 2
> Compaq ProLiant 5000
>
> ------------------------------
>
> Getting the goods
>
> We used virtual test clients to send HTTP requests for URLs
> and images to the Web servers. All of the servers didn't
> generate the same number of requests per minute, but the
> error rate increased for most servers as clients were
> added. Only ICSS stayed strong.
>
> Number of virtual
> test clients 240 960 1,440 1,920
>
> Apache solution
>
> Total attempted Stopped
> requests 2,342 2,534 2,740 serving
>
> Total good gets 2,311 1,857 1,645 [Image]
>
> Error rate 1.32% 26.72% 39.96% [Image]
>
> IBM solution
>
> Total attempted
> requests 3,796 14,964 22,345 29,674
>
> Total good gets 3,796 14,964 22,345 29,674
>
> Error rate 0.00% 0.00% 0.00% 0.00%
>
> Microsoft solution
>
> Total attempted
> requests 3,888 15,419 23,095 19,043
>
> Total good gets 3,888 15,352 22,963 17,390
>
> Error rate 0.00% 0.43% 0.57% 8.68%
>
> Netscape solution
>
> Total attempted
> requests 3,295 11,332 19,212 26,967
>
> Total good gets 3,295 10,481 10,429 9,523
>
> Error rate 0.00% 7.51% 45.72% 64.69%
>
> Lotus Notes
>
> Total attempted
> requests 3,779 14,953 N/A(*) 30,584
>
> Total good gets 3,779 10,150 N/A(*) 9,850
>
> Error rate 0.00% 32.12% N/A(*) 67.79%
>
> (*) Due to time constraints, no number was available for
> Lotus Notes at 1,440 clients.
>
>


\
 
 \ /
  Last update: 2005-03-22 13:39    [W:0.045 / U:0.516 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site