|
Thomas Proell wrote:
>
> Hi!
>
> > I assume what you're trying to do is to put an LVS director in
> > front of several local squids.
>
> Not exactly. The squids are not local, I have to use tunnelling.
> This should run on the french university net (fiber) where each
> university has one proxy cache.
that's local enough, all machines are at the client end.
The scheduling won't be "Round
> Robin", but a hashing function which chooses the proxy by hashing
> the IP-adress (I'll have to implement that, too).
> And, it seems to be a great disadvantage to loose performance,
> since it's a national project...
can't see that you'll loose any performance.
> > both the fwmark type of setup (by Horms) and policy routing (by Julian)
> > will do the same thing.
>
> Can you write a little more about this? I don't understand what
> you mean.
it's new stuff and we're still finding out what we can do with it.
I haven't written it up for the HOWTO yet, but it is mentioned on the LVS web
site
the fwmark is discussed, although not in much detail on the ultramonkey
site which you can get to from
http://www.linuxvirtualserver.org/software/index.html
there is stuff on transparent proxy (called redirect on this page)
and policy routing on
http://www.linuxvirtualserver.org/arp.html
you'll have to read up on ipchains for any of this to make much sense.
None of this is step by step HOWTO grade documentation, so you'll
have to scratch your head a bit before you get it.
Joe
--
Joseph Mack PhD, Senior Systems Engineer, Lockheed Martin
contractor to the National Environmental Supercomputer Center,
mailto:mack.joseph@xxxxxxx ph# 919-541-0007, RTP, NC, USA
|
