httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ben Laurie <>
Subject Re: strange performance delays
Date Fri, 31 Oct 1997 09:22:50 GMT
Brian Behlendorf wrote:
> I'm running a benchmarking program against an apache 1.3 server (actually,
> against Hyperreal itself, during off hours :) and there are some things I'm
> seeing which give me, uh, cause for concern.  I'm going to an SSI page
> which involves one search for an htaccess and I've got followsymlinks on.
> I'm going to the exact same page, so it should be at the top of the disk
> cache.  I'm getting roughly 50 per second, with latencies around 15-20 ms
> on average (from a machine over a packet-filtering firewall, which added
> about 4 ms based on ping).  However, once every 200 requests or so, it
> takes well over a second to service the page.  Once every 1000 requests, I
> see an even larger hit; the largest I've see was 46 seconds (!!!).
> Now, I accept the chaos of a multiuser system, especially hyperreal, where
> a qmail delivery process to a big list can fork() 256 separate deliveries
> in 3 seconds.  :)  I also accept the possible error within the firewall
> between us.  But can anyone think of any part of apache which could cause a
> gap like this?
> By the way, the benchmarking program was opening several parallel threads;
> the more parallel threads, the greater the frequency of these strange delays.
> I suppose a fork() could take an arbitrary long time.  But 46 seconds?
> Anyways, any thoughts would be cool.

Could you be dropping the odd packet? These could just be TCP retry



Ben Laurie            |Phone: +44 (181) 735 0686|Apache Group member
Freelance Consultant  |Fax:   +44 (181) 735 0689|
and Technical Director|Email: |Apache-SSL author
A.L. Digital Ltd,     |
London, England.      |"Apache: TDG"

View raw message