Re: [squid-users] Mixed environment performance

From: Nathan Bell <nathanb@dont-contact.us>
Date: Fri, 19 May 2006 11:48:39 -0600

Kevin wrote:

> On 5/18/06, Nathan Bell <nathanb@actarg.com> wrote:
>
>> Top download speed is the same for both
>> linux and windows boxes, but overall browsing is significantly slower on
>> the windows boxes. At first I thought the problem was with the windows
>> boxes not receiving the dns information quickly, but tests have shown
>> otherwise.
>
>
> I assume this is a non-transparent proxy? How are you configuring the
> proxy settings on the clients? What browsers are used? Are different
> browsers (Firefox vs IE vs Opera) from the same workstation equally
> slow?
>

I am using squid as non-transparent proxy. Mostly everyone is using
Firefox and IE, and both are effectively the same speed.

> When I configure Windows clients for non-transparent authentication, I
> set them up with a PAC script so they don't perform DNS lookups unless
> they are going to an intranet destination. This improves performance
> and reliability -- how much depends on how bad your DNS servers are.
>
> (The one place we cannot use PAC is on Mac.)
>
I was setting everyone up manually with proxy:8080, but now I've set up
our proxy host as wpad.actarg.com with a wpad.dat file (a PAC script as
you suggested) and now things are running smoothly. Thanks for your help.

For other people with the same problem, this website seems to help the most:
http://nscsysop.hypermart.net/proxypac.html

>> Is there a performance penalty for having pc_hosts that don't
>> authenticate along side unix_hosts that do? Is squid trying to access a
>> non-existant ident server on the windows stations?
>
>
> I believe so, but this is one area where my skills are rusty.
> A sniffer capture would show you what is really going on in the inside
> of the network.
>
> Can you provide more details on the browser configuration? There are
> tunables in Firefox for how clients make connections through the
> browser, these can make a significant difference to performance.
>
> Kevin
Received on Fri May 19 2006 - 11:48:48 MDT

This archive was generated by hypermail pre-2.1.9 : Thu Jun 01 2006 - 12:00:02 MDT