At 12:00 26/08/00 -0400, Jay Borkenhagen wrote:
>pfs> Yes, the customer has to do this. It's like a normal multihoming,
>pfs> just using a private AS.
>
>Umm, is that really the advice we should be giving? To my mind, if a
>network is connected to multiple service providers, a public AS number
>is needed and it should be connected to all the multiple providers.
>Private AS numbers have their place, but not for multi-provider
>connectivity.
Taking this to the logical extreme, if every network connected to the
Internet today wants to multihome we'd run out of ASes even sooner than the
current exponential growth of ASN assignments would suggest. Maybe we
should do this to hasten the day of 32-bit AS space... :)
>As per rfc1930:
>
> 7. One prefix, one origin AS
>
> Generally, a prefix can should belong to only one AS. This is a
> direct consequence of the fact that at each point in the Internet
> there can be exactly one routing policy for traffic destined to
> each prefix. [...]
Okay, and that doesn't have to change when using a private AS to multihome.
If you get say one /24 from one upstream and one /24 from the other
upstream, your address space will be announced as part of each upstreams
address block. If you using something like BGP conditional advertisement,
you are only breaking this recommendation (should = recommendation, not a
rule) when there is a link failure.
If you have your own registry sourced address space, then you should use a
public ASN. But very few end-networks these days get PI space, as far as I
can tell. They get address from their upstreams, their SPs. So using a
private AS as I described above will work perfectly well.
>To simultaneously satisfy the two goals that (1) each prefix should
>originate in a unique AS number, and (2) private AS numbers must be
>seen in the core of the Internet, a unique public AS number must be
>used.
It's important to remember that the rfc says "should", not "must". The only
place where I've seen inconsistent origin ASes being a problem is trying to
describe policy in a routing registry. But have we seen any operational
problems caused by inconsistent origin ASes in the Internet? At the moment
there are 617 such prefixes, and picking some at random I can reach them all...
Okay, this has all been a bit off topic, my apologies to the list for that.
But I think multihoming in today's Internet is a lot more than simply
spraying /24s everywhere and hoping you'll get some connectivity. (Route
flap dampening, "Net Police",.... 50000 /24s and growing rapidly,...)
cheers,
philip
--
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:15 EDT