Is v6 as important as v4? Of course not [was: IPv6 internet broken, cogent/telia/hurricane not peering]
mleber at he.net
Wed Oct 14 09:21:52 UTC 2009
Patrick W. Gilmore wrote:
> For the v6 'Net to be used, customers - you know the people who pay for
> those router things and that fiber stuff and all our salaries and such -
> need to feel some comfort around it actually working. This did not help
> that comfort level. And I believe it is valid to ask about it.
That is entirely correct and I'm glad you asked that question! ;)
Let me explain:
(Lots of truisms here, bear with me!)
IPv6 is newer than IPv4.
As IPv6 is newer than IPv4, the equipment to support IPv6 natively is
newer than legacy equipment already deployed that only supports IPv4.
As the equipment that supports native IPv6 is newer, there are fewer
core networks that run native IPv6.
As these new IPv6 networks are deployed they are growing and developing.
(Like neurons forming connections, the IPv6 network is.)
Deployment of IPv6 in the core has been growing year to year, with that
growth accelerating. In fact, I'd tell trend watchers of business
econometrics the accelerating growth curve both represents something
important happening right now and something that is likely to have real
world implications for Internet infrastructure companies in the future:
(Short url: http://tiny.cc/An6fl )
If you are in the connectivity business, you can add a caption to this
graph of your choosing:
"Ignore at your own peril."
Or (I like this one):
"I see opportunity."
> However, the question still stands about the stability, and therefor,
> utility of the v6 'Net. Is it still some bastard child, some beta test,
> some side project?
As you know, the IPv4 Internet of today is a product of the hard work of
people of yore (ok well, more seriously, a large number of the people on
this list and at networks around the world).
The nature of things is that the coherent shared illusion of a single
Internet routing table is the result of a rough consensus produced by
years and years and years of accumulated business relationships and
network engineer routing policy configurations.
IPv6 is going through that phase right now, at accelerated pace.
Perhaps geometric growth is not good enough for you as a business
person. Perhaps where we are on the curve is not good enough for you
yet. Perhaps you'd like to retire before working with another protocol.
I hereby apologize to you on behalf of IPv6 that it has not had the same
three decades of deployment and experimentation as IPv4. ;)
IPv6 is not going to spring into existence as a fully complete global
network to replace IPv4 on a specific flag day (December 21st 2012?).
IPv6 will grow in deployment at the same time the Internet continues to
work, at what appears to be on a geometric growth curve, due to some
reasons a business economist can write a paper about. Network effect?
Risk avoidance due to IPv4 run out? Risk avoidance due to technology
shift? Yukon gold rush? The after the fact result of careful planning
by thoughtful people started years earlier? Or perhaps, the projected
functional economic value of IP addresses?
> Or is it ready to have _revenue_producing_ traffic
> put on it?
IPv6 is production for some value of the word production. We see
traffic around 1.5 Gbps, peaks at 2 Gbps and growing...
Perhaps this says something about the amount of traffic that will be
seen when it gets used widely.
1000 times as much? (Our guess) What's your guess?
Warning! If you pick a low number you are saying that IPv6 is in
widespread production use right now. :-P
> In summary, we have the standard Chicken & Egg problem. No one cares
> about v6,
speak for yourself (introduce into evidence exhibit 1: the graph linked
to above, exhibit 2: we note how part of the original poster's problem
got fixed that day).
> so no one puts anything important on v6,
speak for yourself (reference real traffic above).
Once upon a time, something called IPv4 was invented, and some people
created hardware for it, wrote software for it, tried it out, wrote some
papers, wrote some RFCs (after writing working code, the way it should
be done LOL), and then experimented some more. There were lots of
problems that got solved, things that worked in real life in spite of
theoretical problems, and bugs that got fixed. Some companies got
created... blah blah blah.
> Sad times for the future of the Internet if we all need to use v6
> Real Soon Now.
Or, expect real freaking huge opportunity and dislocation ahead.
Of course, this dislocation may only affect some specific players and
companies and industries. For the regular user it could just happen
transparently that by the time they get their next computer with
Microsoft Windows 9 or Ubuntu Quick Quagga... it just works.
Imagine, what would it be like if all the core network operators had to
figure out who get Internet connectivity from again. Imagine, if they
had to setup all their peering and transit sessions again because of
their existing telecommunications vendors only some of them decided to
try out this new Internet thing. (Deja vu yet?)
Well you don't have to imagine! That is what is happening right now!
We are talking about a major sea change here. How long might the
central wave of this sea change take to pass? 5 years?
To come up with a wildy guestimated date, you might average the time to:
* Replace 60 percent of home computers (count existing Windows Vista and
Macintosh OS X computers towards this total, since they already just
magically work with IPv6 if you have IPv6 on your LAN).
* Replace 60 percent of residential CPE. (Base it on customer churn,
CPE failure, and technology upgrades.)
* Replace 60 percent of core routers.
Why use 60 percent as a milestone?
Because it represents more than half.
Why does more than half matter?
Because we could use it to make a nice graph to project cross over for
the prevalence of IPv6 connectivity vs IPv4. In most companies I've
been at, this sort of graph is used to predict when to stop spending
additional money on the item that will not be relevant to the company's
future sources of revenue. In otherwords, 60 percent is the point of no
return because at that point capital allocation budgets are co-opted.
BTW, if you plan on getting started after that ship has sailed, well...
> I asked for HE's view on that.
My pleasure! :)
More information about the NANOG