Re: Squid Performance Issues.

From: Bill Wichers <billw@dont-contact.us>
Date: Tue, 9 Dec 1997 23:09:23 -0500 (EST)

You are correct that WinNT does do some compression. When a WinNT/95
machine dials up a PPP connection to a WinNT server they do do (provided
it's enabled, and I believe it is by default) some serious compression. I
have seen up to almost 300 kb/s over a 24 kb/s modem (phone lines here are
less than amazing) when downloading a copy of the squid access.log file
for analysis. The lag from the compression algorithm is negligable, and
I've never noticed it.

If you connect to the NT machine to test it, and connect to the
Linux/Solaris machines to test them, then NT will have an advantage do to
its link compression. A better cache-specific test would probably be a few
machines on a LAN running wget or something similar, and I think someone
on the list ran SGI's WebBench suite through Squid once.

        -Bill

On 9 Dec 1997, Stefan Monnier wrote:

> Chris Keladis <chrisk@tpgi.com.au> writes:
>
> > AHA-2940 PCI Controllers, and 10mbit ISA Ethernet cards. I have timed the
> > test loading identical web pages over a 26.4k modem connection. (Always
> > clearing the local browser cache).
>
> That doesn't sound like a proxy benchmark but a communication benchmark.
> Are you sure the communication protocol used with NT is the same as with Linux?
> Isn't the NT protocol using some compression (or some *better* compression)?
>
>
> Stefan
>
Received on Tue Dec 09 1997 - 20:13:12 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:37:53 MST