|
I did a 2 hour interview on October 23rd with John Curran, Board Chair of ARIN the North American Regional Internet Routing Registry for the last decade.
I now understand what is at stake with IPv6. Outside of a key core group of network engineers I think darn few people do understand. And not all of them agree on how the scenario plays out though virtually all say the situation is very serious.
John believes that it is huge. It is as big as Y2K except no one knows a precise date by which everything has to be done. And because the onset of the crisis is 3 to 4 years away, there is no real incentive for people to be first movers. Why should I start spending money now in a low margin industry when, until my competitors also spend, my investment does no good? We have had years of procrastination.
But finally on May 7, 2007 a significant event occurred. ARIN sent out a letter to its members (who are primarily service providers) saying that by 2011 they expected to have no more blocks of Ipv4 numbers to hand out and strongly suggesting that they become iPv6 operational as soon as possible. If they don’t and new ipv4 blocks become unavailable they will no longer be able to add new customers. See also Geoff Huston‘s The End of the (ipv4) World is Nigher!
Therefore what seems increasingly certain is that, if hundreds of thousands of people don’t get busy making backbones and servers ready, routable ipv4 hits a wall by the 2010 - 2011 time frame. To do what needs to be done both all backbones in the core and all servers at the edge have to be rebuilt. Significant amounts of software will have to be rewritten. If the 12000 global ISP’s don’t begin to work on their backbones, and the millions of end sites don’t begin to enable their web and email servers to IPv6, then, avoiding a collision with the “wall” becomes increasingly difficult. It is inadvisable to listen to promises of a technology “fix.” There is no technology solution that can be deployed in less than four years. Randy Bush has been giving talks at NANOG and the RIR meetings [PDF].
If it Doesn’t Get Fixed
Administrative attempts to extend the life of v4 won’t work either. Here’s why. Tardy deployment of V6 combined with growing scarcity of allocateable v4 blocs is very likely to put market pressure on service providers to increase significantly the numbers of new routes they advertise to the core and ask peers to accept. The longer the v6 deployment takes the few ipv4 blocs available and the greater pressure on backbone routers. Delayed long enough and service providers either have to turn down new customers or replace backbones more and more often.
As IP v4 gets harder to get routes added to the internet core every month will increase by orders of magnitude. Even with Ipv6 allocation the need for CIDR does NOT go away.
Eventually routing pressure causes backbone fall over even with the largest services providers if they do not begin to SHED routes. Some will choose to survive by shedding routes. This route shedding makes more and more of the internet unreachable from a given service provider. Many service providers die. Those that don’t become more and more like islands. Private IP networks that reach 50 million customers can route to maybe 250 million or 500 million non customers. There are other islands out there that represent large service providers that have suffered the same fate.
So we have big ISPs that may get you to 50 million or a few hundred million but they can no longer get you EVERYWHERE - say to 2 billion people connected to the internet or another billion who would like to join the first two. How many big ISPs remain globally? 12? 40? 60? Think of these as more like balkanized islands than as the old globally interconnected internet. In this environment a social networking service like Facebook or mobile ecommerce services that are REALLY attractive because those promoting them assume we can reach the entire internet loose some of their luster. Either, because they can no longer reach the entire networked population, or because the COST of doing so has gone through the roof. The economic luster is taken off most all Web 2.0 businesses.
Use of the internet for a myriad of education government and research purposes become much less effective. They economic value of a global network is gone and once that happens putting Humpty Dumpty back together again is not likely.
If things don’t get fixed NOW, just about the only other way to forestall backbone meltdown is for service providers to say the internet is full. It is closed. We cannot accept new customers. Of course this is not likely to happen. So either everyone begins to change to v6 now, or backbones fall over, many ISPs go out of business and the internet balkanizes globally.
I’ll be publishing edited IPv6 discussions from my own list and from NANOG by next Wednesday (the December COOK Report). Look for the publication of the interview with John Curran in the January COOK Report on or about November 30.
Sponsored byWhoisXML API
Sponsored byVerisign
Sponsored byVerisign
Sponsored byCSC
Sponsored byDNIB.com
Sponsored byIPv4.Global
Sponsored byRadix
What folks are not realizing is that IPv6 is already here. Even if the ISPs don’t deploy it, transition mechanism already do IPv6 end-to-end. Is only a matter of time for the ISPs to realize that they better start offering transition services (such as 6to4 and Teredo relays) from their own networks, instead of having their users using external relays, which means extra upstream bandwidth, and longer delays.
See my latest slides at the RIPE meeting plenary.
Do the math. Aggregation is not the biggest concern. IPv4 has 2^32 addresses, about 3 billion of which are usable for (unicast) Internet. A worst-case v4 global routing table will have a few million entries at most. Which is just a single order of magnitude more than we have today. Large carriers already have 3 times more routes inside their own network than the global Internet currently has.
Getting more addresses when we’ve run out is by far the biggest issue. Turning people away just because there’s nothing more to be had is far worse than the threat of ‘cores melting down’. In order to get the next billion people online we’ll need to be able to assign them a number first.
First get the blocks of space you can route, then worry about how complex that might become - however complicated that may turn out to be. An adequate worldwide migration to v6 is about 5 years away if I’m very optimistic, 10 years being more likely. So aside from migrating to v6 we also need to worry about how we keep v4 running.