[6bone] Re: routing concern

[Rik van Riel]

On Thu, 1 Aug 2002 itojun@iijlab.net wrote:

> 	the problem is that there's no distinction between sTLA cloud and pTLA
> 	cloud - they are overwrapped, and most of sTLA sites do hold pTLA
> 	address with them too.
>
> 	if we could split these two clouds into two (interconnected via few
> 	routers), it would be very nice, but i don't think it is going to be
> 	possible.  so we need to harden 6bone (= pTLA cloud).

One thing possible to make 6bone traffic more reliable is
BGP confederations. Inside the COMPENDIUM pTLA we have
multiple ipv6 islands in a BGP confederation, advertising
the same BGP AS number to external peers and non-confed
islands in the pTLA.

If routing to one of the central ipv6 islands is lost,
the packets will be routed via one of the others.

Of course, performance is bad in a pTLA as geographically
dispersed as COMPENDIUM, but ipv6 has proven to be much
more reliable for me than ipv4 has ever been.

This idea could be used by organisations who are close to
each other geographically and by hop count ... something
vaguely like multihoming, without violating RFC 2772.

Having 4 well-connected (to each other) sites participate
in the same pTLA and same BGP confederation could make the
6bone traffic more reliable for both the sites themselves
and the peers they transit for.

Performance can be regained later when the sites move to
native ipv6 links.

regards,

Rik
-- 
Bravely reimplemented by the knights who say "NIH".

http://www.surriel.com/		http://distro.conectiva.com/