Re: Squid performance problems

From: Chris Tilbury <>
Date: Wed, 28 Apr 1999 17:43:37 +0100

On Wed, Apr 28, 1999 at 08:32:12AM -0700, Scott Hess wrote:

> Since we're talking about a pretty specific connectivity problem (you
> either see the US or you don't), could the problem be addressed by using a
> sacrificial Squid box of some sort? Perhaps you could route all requests
> to US sites through a parent proxy who's entire job is to proxy for
> requests to US sites. Then if the US goes away, _that_ proxy gets wedged
> up, but the others should be able to continue proxying non-US sites.
> [If that won't work, let me know why not, because it would indicate a
> misunderstanding on my part about how Squid clusters might work,]

The reason it won't work is nothing as such to do with squid, but more
to do with the simple fact that there is no practical way to tell squid
(or browsers) to send all traffic destined to the US through one proxy
and not another. AS number routing isn't really practical here; there's
no sensible mechanism to recurse the routing databases to identify which
networks are reached via different providers, and they're not
completely accurate anyway.

Bear in mind that for UK providers, there are going to be a
large collection of sites based in Europe (poor national peerings
being a particular problem in many european countries) reached via the
US. If the provider has no presence in Europe, then _all_ of Europe
is probably going to be reached via the US.

There is also the whole Asia Pacific region, which is also bound
to be reached via the US for ISPs in the UK.

I'm afraid it's not as simple as saying ".com == US". If it was, an awful
lot of other things would be much easier, too ...

Ultimately, the "core" internet is in North America. If providers in
countries outside North America lose their connectivity to it, then
they're probably going to be stuffed, one way or another.



Chris Tilbury, UNIX Systems Administrator, IT Services, University of Warwick
EMAIL: PHONE: +44 1203 523365(V)/+44 1203 523267(F)
Received on Wed Apr 28 1999 - 10:35:50 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:46:01 MST