The internet has become our defacto distribution network for utility computing (otherwise known as cloud computing). The global response to IPv4 exhaustion has not been adoption of IPv6; as such we face a future internet where the incumbent players dictate the terms and conditions for entry. This will have a negative impact on innovation in systems that are build on cloud.
4. Our global cloud distribution network http://www.zageex.ro/wp-content/uploads/2009/09/internet-24591-288x300.jpg
5. The numbers game: IPv4 exhaustion Why is this such a complex issue to solve? http://www.potaroo.net/ispcol/2009-05/ipv4model.html
6. Problem #1 "The Internet is not a thing "* * quote stolen from Martin Geddes http://blog.ecomm.ec/2010/07/martin-geddes-interview-cloud.html It’s a collection of agreements to exchange traffic
11. 1000 customers behind one public IP? Hmmm... http://www.delegator.com/wp-content/uploads/2011/06/Geo-Targeting.jpg Geo-targeting? (Ad revenue models?) Abuse detection? Lawful intercept?
12. Rising costs ? IPv4 Black market: New entrants "pay to play" with IPv4. http://blogs.creditcards.com/assets_c/2011/02/price-hike-tag-thumb-250x290-1029.jpg
14. Failure of the free market? "...in a networked environment that stalls on IPv6 the resultant NAT and ALG-ridden IPv4 environment is one where the current incumbents will hold all the addresses and any further competitive entry into the Internet by new actors, at both the levels of carriage and content services, would be effectively limited to the terms and conditions imposed by the incumbents. " Geoff Huston, APNIC
The topic of this talk is cloud distribution. And I hate to be the bearer of bad news, but the forecast is unsettled weather. There is trouble on the horizon.
Cloud is the delivery of computing as a service rather than a product. We all know this. It represents a completely new paradigm for IT. etc etc That is why we are sitting here.
However, to actually deliver computing as a utility service you need a distribution network. This is an essential component.
Now, we have a popular global cloud distribution network. It's called the internet.
Back in the 1970s 4.3 billion addresses sounded like a lot. When you fast forward to 2011 is really doesn't sound so great any more. Because we only have 30% of the world's population online and we've already run out of IPs. Whoops! So we have a problem to be fixed on our cloud distribution network. Using a new set of numbers.
There are 3 big problems with the internet when it comes to upgrades It would be ideal if we could manage the internet holistically like a hornby train set. A single autonomous system. Just one beardy network admin to call all the shots. But it doesn't work like that. Because of course the internet is "not a thing" like the national grid is a thing. The internet is simply a collection of agreements to exchange traffic.
Once apon a time the internet was rare. Access to the internet was a competitive advantage and a decent level of access comanded a high price. it was cool and new and fashionable and attracted a lot of investment. But today of course IPv4 is a ubiquitous digital commodity. It's a boring cost of doing business. And it *just works*. Everywhere. This means that it's invisible and very low down the funding agenda.
Problem #3 Ok - game theory. So we know that the internet is not a thing. It's actually a collection of agreements between 36,000 autonomous systems, with millions of customers between them. Now each autonomous system and each end customer is playing the IPv4 exhaustion strategy game. The best outcome for all would be some sort of collaboritive effort to implement IPv6. But in reality each player acts independantly and acts in their own interests. The perceived payoffs of implementing IPv6 before it reaches critical mass are not good. Delaying tactics appear to have a better payoff. So we begin a slow-motion train wreck of IPv4 exhaustion.
Ok, here is the top 5 rundown for "impacts of IPv4 exhaustion on cloud"
Carrier grade NAT will have some pretty interesting implications for content and application providers who depend on advertising revenue, which is itself dependent on geo-targeting using IP addresses.
The costs of building, operating and expanding networks are going up as of now. The v4 black market exists today. There is at least one v4 trading exchange. Any new entrants will have to buy their way into IPv4 at the market rate. While the costs and level of brokenness are increasing, we can expect CDNs to step in to "solve" our problems. I've already seen an Akamai presentation which is focused on this.
So much effort goes into squeezing milliseconds of latency out of the application architecture. And now we're set to blow it all to bits with large scale NAT which won't scale and will introduce session limits, latency and bottlenecks. Anyone in cloud gaming should start planning their exit strategy now. Furthermore, with double NAT some things just won't work full stop.
This is a quote from APNIC's Geoff Huston, who is one of the few people writing insightful stuff on this topic. It echo's the sentiments of the IEEE, which warn of the harm to the digital economy from failing to implement IPv6. Essentially our global apathy to this issue is going to land us in a situation where the incumbents hold all the cards. They've got all the IPs and any new protocols will have to be approved by them. And the stakes are higher now, because the internet is about more than blink tags and digging workman animated gifs. It's now the distribution network for utility computing.
And we reach number one in the top 5 impacts. This is a big one for cloud because it's all about limiting innovation. The commoditisation of infrastructure should allow a faster rate of innovation in higher order systems. But we are looking at a situation where the free market takes a serious hit. Remember that the incumbents in today's internet will hold all the cards. We live in a client-server world right now. But surely we don't want to be locked into the present. Want to introduce a new protocol? Fancy developing peer-to-peer cloud? Internet of things? Forget it, without an open end-to-end principle.