Need for N-homedness (was Re: asymmetric routing)

[Michel Py]

Pim,

> Pim van Pelt wrote:
> (fear my ascii art - best viewed in xterm, not in Outlook et al :)

Actually, it displays OK in Outlook.


> TRANSIT1-(R1)----fiber-to-customer-----(R3)---\
>            |                             |     |------network at
customer
> TRANSIT2-(R2)---backup-E1-to-customer--(R4)---/

This setup is good enough for most. Unfortunately, it is not available
everywhere, and it does not provide redundancy if something happens to
R1-R2 that are located in the same building.

> I fail to see any need for more than one prefix per customer,

I assume you mean more than one provider independant (PI) prefix.

The main driving force behind that is the elimination of (PI) addresses,
One of the animals that clog the IPv4 DFZ today. It is too early to know
if there will be ANY IPv6 PI addresses at all.


> and still say that IF a customer has more than one prefix, his R3/R4
> should preform source based policy routing, sending traffic from
prefixA
> through the circuits to ISPA and traffic from prefixB to ISPB.

Definitely. However, this is not enough. There are two other problems
with
this scheme (that is likely to become part of the IPv6 multihoming
solution):
1. Address selection by the source that has to choose between ISPA ans
ISPB.
2. Make sure that the host (server) that has two addresses, uses the
correct one to return the traffic (the one that the traffic was sent
to).

Michel.

P.S. I think that this thread should be switched to the multi6 mailing
list.