What are ISPs going to do for deploying IPv6? (Was: v6 bgp peer costs?)

Tue Jul 27 18:20:21 UTC 2010

On 2010-07-27 20:03, Jared Mauch wrote:
[..]
> I'm honestly interested in what the US based DSL (incumbent) providers
> are doing for IPv6 (eg: att/bls/sbc/uverse, qwest, vz dsl).
> Most of the "ethernet" (including PON) equipment is more likely
> to do IPv6 correctly, but I'm not sure that the PPPo* DSL equipment
> is going to be quite as happy with it.
> 

I actually have only one answer that makes sense for this: 6rd

Or to state it in a more complete way:
 - native IPv4 + IPv6 where possible
 - native IPv4 + 6rd everywhere else

Any other method has deployment-wise too much overhead or not enough
control and 6rd is rapidly getting implemented in a lot of hardware.
Actually today I noticed that even iproute (aka 'ip' on Linux) has 6rd
support built-in, dunno when that happened, but that is nice to see.

Yes, you lose a few bits, yes you have to come up with a way of mapping
the IPv4 space in your IPv6 address plan, but that is not soo difficult
and actually will make network admins happy as the bits are easy to
identify.

Of course, if one has CPE at the enduser which can do PPPv6 then PPPv6
definitely is also a proper "native" alike deployment scenario.

> This should be interesting.  I also look forward to seeing what
> devices start to keel over by software vs hardware switched IPv6
> paths as traffic increases.

That one will be very interesting indeed. In the case of 6rd though one
can just add more hardware to the pile and anycast it to make it scale
as far as one wants. And yes, I indeed say to just add Linux/BSD boxes
for handling this, 1U boxes are cheap, OS is easy to install as you do
with all those webservers/storage/mail etc you already have anyway, thus
it is just another box to add to the auto-deploy setup.

One has to remember though that 6rd is a TRANSITION mechanism that
should fade away on the long run. For the next year or two though most
likely one can get away with tunneled connectivity, after that, when
major sites will be enabling IPv6 and thus content and traffic shifts
from IPv4 to IPv6 the folks who are already trying to get native to
their customers today and have hardware IPv6 enabled in their cores will
definitely have a monetary advantage.

One important thing folks should not forget though is to make sure that
they can handle abuse and statistics properly. Take that into account
from the start and you'll save yourself a lot of hassle.

As for non-managed/non-owned paths, ISPs can then always still opt for a
tunnel-broker like solution, be that PPTP based, TIC(AYIYA/hb) or TSP based.

Like always, what shoe best fits your foot. But do think about what you
want to be running on till the next upgrade cycle of your hardware comes
around ;)

Greets,
 Jeroen