Re: squid neigbour response

From: Richard Munro <>
Date: Thu, 23 Oct 1997 10:28:40 -0400 (EDT)

On Thu, 23 Oct 1997, Dancer wrote:

> Yes...And we have the same complaints from our users. Your summary neatly
> covers my understanding of the process.

Us too. However, I have stumbled on something that has made a big
difference for us.

For a long time I was using the standard VM version of Squid. Everything
worked well. Then at some point a few months ago I decided to use the NOVM
version of Squid. From that point on, things were noticeably slower and
unusual things started happening that I didn't pin down to the NOVM version
at first (silly me, I chalked it up to other changes in the code).

What I notice, for instance, is that graphics were handled differently.
With the VM version, a client would check a page and if the material was
already in the cache, the proxy server would send out an if-modified-since
request. All is well so far. However the difference between the VM and
NOVM versions seemed to be this: if the client already had the material
*AND* it hadn't been changed, the VM version didn't resend it to the
client. The NOVM version seemed to want to refresh everything. Is was
*THIS* that was causing people to perceive very slow response, especially
on dialup lines.

A few days ago I switched back to the VM version and the complaints
stopped. In fact, people have been stopping me in the hall asking what I
did that improved the performance so radically. Nothing, other than
switching versions.

I'm just mentioning this in case it helps...

> Matthew Flanagan wrote:
> > Hi All,
> >
> > I just thought I'd fire a couple of questions off squid and neighbour
> > interaction. I'll give you the low down first.
> >
> > I'm getting complaints from users that accessing the web via the cache is
> > too slow. If they turn the cache off in their browser it is much faster, so
> > this is what they tend to do :(. I've had a look at the server and there
> > doesn't seem to be any real problems.
> >
> > btw. the server is a dedicated Ultra 2170 384MB RAM, 16GB cache-swap, 128MB
> > cache-mem, connected at 100Mbps running squid 1.1.15.
> >
> > It occurred to me that I may be being slowed down by my neighbours but I
> > just wanted to verify the way I thought ICP queries worked.
> >
> > Now I thought that when an ICP query was sent out that the cache wouldn't
> > go direct until all replies had returned AND they were all negative, but if
> > a positive reply was returned at any time it would act on that.
> >
> > So, if some of my neighbours are slow to respond, and all my ICP replies so
> > far are negative, then my cache will take a while before it will finally
> > decide to go direct.
> >
> > Is this correct?
> >
> > If not, then what is going on here?!!?
> >
> > --
> > Matthew Flanagan
> >
> > Network Administrator - Information Technology Division
> > University of Technology, Sydney.
> > phone://61-2-9514-2141 fax://61-2-9514-1994
> --
> Why are doves always the symbol of peace?
> Why not pillows? They have more feathers, and they don't have a dangerous beak.
> ICQ UIN: 3225440

Unix System Administration Humber College Network Services Department
Received on Thu Oct 23 1997 - 07:37:49 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:37:19 MST