RE: [SQU] Performance Problem

> -----Original Message-----
> From: Rainer Ginsberg [mailto:rainer.ginsberg@de.bosch.com]
> Sent: Monday, 20 November 2000 11:29 PM
> To: Squid-Users
> Subject: Re: [SQU] Performance Problem
>
>
> On 18.11.00 10:22 +1100 Robert Collins
> <robert.collins@itdomain.com.au>
> wrote:
>
> >> Here is some data about our proxy usage:
> >> - User authentication is done via LDAP to a directory on
> >> localhost. I have specified 16 instances of squid_ldap_auth.
> >> - At peak times we have about 100 requests per second.
> >> - We use only five ACLs with up to 50 regular expressions.
> >>
> >> What are the main parameters that I should tune? What are the
> >> parameters that I should look for in the cachemgr?
> >
> > Are you part of a hierarchy? If so are you using ICP/Digests? Most
> > environments don't need ICP (and it introduces delays
> before requests get
> > fired off).
>
> No, I'm not part of a hierarchy. Though I have configured cache_peers,
> which are Gauntlet firewalls.
>
> > When you say the WWW was slow; do you mean that large files
> (say 1+Mb)
> > have a slower download rate, or that requests are not
> initiated to the
> > origin servers as quickly?
>
> I mean that the request doesn't get sent, or the answer doesn't get
> delivered to the client. When I try to access a web site, the browser
> displays a white screen, and then nothing happens for a few minutes.
> Then the site comes up, but with delays for every image.
>
> Cheers,
> Rainer
>

Can you send you cache_peer lines? The ip/name can be masked if you
wish.

When you have a hung request, is it listed under active client requests
in the cachemgr? and do the gauntlet firewalls see the requests? Does
cache.log or access.log give any clues? (ie error 4xx or 5xx in
access.log)..

Rob

--
To unsubscribe, see http://www.squid-cache.org/mailing-lists.html