[arin-ppml] IPv4 is depleted today - unrealistic statements about IPv6 inevitability

[William Herrin]

On Tue, Sep 2, 2008 at 5:40 PM, Howard, W. Lee
<Lee.Howard at stanleyassociates.com> wrote:
> Those operators tell me there are no routers available or planned
> that can handle the projected volume of routes and updates.  One
> of the engineers for a router company you'll recognize told us about
> a year ago that Moore's Law is approaching its limits regarding
> routing scalability.
>
> http://www.arin.net/meetings/minutes/ARIN_XX/ppm1_transcript.html#anchor
> _7
> Look for "MR. LI".

Lee,

Robin and I have been working with Tony Li over on the Routing
Research Group for the past year or so. He's a smart guy but his
estimate is too low by at least an order of magnitude and possibly as
much as two.

The missing link is parallelism. Tony assumes we'll continue to
compute 100% of all routes on each uniprocessor MIPS system as Cisco
does today. On RRG, we've seen at least two viable methods of
introducing useful parallelism in a BGP-based system at a sublinear
increase in cost. They don't buy growth forever, but they buy a lot
more than Tony thinks.


> Routing does not scale to 4 billion routes.  Don't know exactly
> where hardware can't do it anymore.

Doesn't have to. As long as the /24 lower boundary on routability
holds, IPv4 will stay on a logarithmic growth curve that flattens out
somewhere around 7M or 8M routes. Given a Opteron-like dedicated
memory banks plus hypertransport architecture and some mild
improvements to the software that take advantage of multiprocessing
without substantively altering the BGP algorithm, we can build a
router that can handle 7M to 8M routes today and we'll be able to
build it cost-effectively by the time we need it. It'll be more
expensive that today's core routers, but it won't put you out of
business.

In other words, IPv6 is not the only choice. The best choice maybe,
but don't for an instant imagine that  properly motivated people can't
find a way to make the IPv4 Internet continue to grow.


> If large network operators have to [make IPv4 continue to work]
> will that always be cheaper and provide better connectivity than
> IPv6 with translation technologies (tunnels, NAT-PT, teredo, 6to4,
> whatever: http://www.getipv6.info/index.php/Planning_IPv6_Deployment)?

That approaches the real question: IPv6 migration has a cost function
and so does IPv4 growth. Where do the cost lines cross? When does it
become less expensive to deploy IPv6 and exert the push that brings
IPv4 to a close?

The plain truth is that we don't have enough data to do better than
guess at the answer to that question. But we do know that each IPv6
route costs 2 to 3 times as much to implement in the router as an IPv4
route. And we know that there is no new magic in IPv6 routing:
multihoming still requires a unique global route.

If I had to guess, I'd guess that the cost lines never cross. Even as
IPv4 gets more expensive, it remains more cost effective to squeeze a
little more out of it than to attempt an IPv6 migration. If I'm right,
it means that some successor technology that directly solves the
routing problem will leapfrog IPv6 in 10 or 20 years without IPv6 ever
having reached useful deployment or ever having paid back its early
adopters.


> Yep, we've gotten ourselves into a fine mess.

And that's the sad truth.

Regards,
Bill Herrin





-- 
William D. Herrin ................ herrin at dirtside.com bill at herrin.us
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004