Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 3 Aug 1997 13:12:16 -0700 (PDT)
From:      Doug White <dwhite@gdi.uoregon.edu>
To:        Katherine Nenno <spam_test@dittosrush.com>
Cc:        freebsd-questions@FreeBSD.ORG
Subject:   Re: FreeBSD and large #s of http requests
Message-ID:  <Pine.BSF.3.96.970803130736.4730S-100000@localhost>
In-Reply-To: <3.0.2.32.19970803145947.00920210@mail.gte.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 3 Aug 1997, Katherine Nenno wrote:

> We host a web site for a popular radio show host (#2 in the US, 1 in
> Canada).  Recently (a few months ago), we planned to have a interactive
> chat with her where people could ask her questions and she could respond.
> What we didn't count on was the number of http requests this would generate.  

:-)  This is 1997, the Year of the Surfers.  

> The site is hosted on a Pentium Pro 200 machine with 64 meg RAM running
> Apache 1.2.0 and FreeBSD 2.1.7.  The chat script was a custom perl script
> that we contracted out to have written.  We didn't anticipate the number of
> hits that this would generate.  Within the first minute the server crashed.

Ouch. :(  When you're this big and have an online event, you have to be
ready for everyone in the US to hit you at once.    (did it just drop to a
crawl or did you get some sort of error message?)

> Now, we have been asked if we can fix the problem and have another chat
> that will work.  Does anyone have any suggestions on the best way to do
> this?  Is it even possible without buying tons of expensive equipment?

I think it's possible, but some tuning is necessary.

1.  The Perl script should be as minimal as possible.  Each hit will spin
off a copy of the script and that builds up fast.  More memory will head
off the inevitable, however. 

2.  It may be necessary to tune some of the kernel parameters to
accomodate a high number of http hits.  I suggest asking in
hackers@freebsd.org, our kernel developer's list.  They should be able to
suggest some parameters to tune before your next trial.

Hope this helps and good luck :)

Doug White                              | University of Oregon  
Internet:  dwhite@resnet.uoregon.edu    | Residence Networking Assistant
http://gladstone.uoregon.edu/~dwhite    | Computer Science Major
Spam routed to /dev/null by Procmail    | Death to Cyberpromo




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.96.970803130736.4730S-100000>