Posts Tagged ‘ipv6’

Cloud Distribution Forecast: Unsettled

December 20, 2011

The issue of network latency and loss – and how we meet the challenge of developing cloud-ready applications on best-effort networks –  is a topic that has gained a bit more exposure recently with the Alcatel and Cisco announcements, coinciding with a paper from Joe Weinman.

In addition to the demands of realtime cloud applications, another reason that the network is returning to the foreground of discussion is IPv4 exhaustion. I’ve spent a fair amount of time looking at this particular issue in 2011, including the likely impact,  in terms of costs and performance, of the various duct-tape IPv6 transition methods that are going to be implemented globally.  Here is a copy of the slides from my presentation’ at Cloudcamp London (5th October 2011).


The internet has become our defacto distribution network for utility computing (otherwise known as cloud computing). The global response to IPv4 exhaustion has, unsurprisingly, not been timely adoption of IPv6; as such we now face an immediate future internet where the incumbent players dictate the terms and conditions for entry. This is likely to have a negative impact on innovation in systems that are built on cloud.


Bad news: The IPv4 barrel is almost empty. Good news: The network will be relevant again.

December 6, 2010

“He who talks more is sooner exhausted” — Lao Tsu

Buying Internet access is a simple decision for most people in the developed world.

We take it for granted that we will be able to reach everyone, and everyone will be able to reach us, so the only things to consider are “how much bandwidth do I get?” and “how much does it cost?”.

However, if the Internet started to split into two logical networks, you would want a network provider that connected you to both parts. If you wanted to reach everybody, the quality of the Internet access, IPv4/IPv6 integration, and the quality of support you received would be key differentiating factors.

The ubiquity of everyday, humdrum Internet access has brought about fierce price wars and consolidation in the IP transit market over the last decade. Right now Internet access is a digital commodity, but the impending IPv6 shakeup will change that .

The physical Internet will be expected to accommodate a new logical layer for which few have invested. Meanwhile a new way of working is going to be imposed on people who are starting from a knowledge and operational skill level of zero.

The consequence of this disruption is that something that we have come to consider simply as a cost of doing business will once again have strategic relevance. Internet access that could previously be relied upon to be a ubiquitous commodity, will become differentiated by variation in performance, user experience and support. We will need to ask new questions: Can I reach everyone I need to? Can they reach me? What is the user experience like? Will this service even work?

With one year to go before even the RIR supply of IPs is exhausted, it is now inevitable that the Internet will fragment and develop a split brain. When the RIRs themselves run dry, and the trade in IPv4 space really gets going, prices will escalate rapidly. Eventually it stops being commercially viable to buy IPv4 address space in order to maintain network growth.

There is simply no way the entire IPv4 Internet will be running dual stack by the time we see IPv6-only users, content and services start sprouting up. It’s also inconceivable that there will be any large-scale protocol NAT that can deliver a decent user experience. So this means, regardless of how much IPv4 you have bought or hoarded, if you don’t migrate to dual stack anyway you can expect to be branded a legacy network by competitors.

As I explained previously, there will be a severe skills shortage in IPv6. Making sense of the technical transition and all the implications is exceedingly difficult; it involves taking a step back to get an objective overview of the Internet as a whole.

Disruptive effects can take various forms – market forces, technology and regulation. The beauty of this technical one is that we’ve known about it since forever. Everybody has had a chance to make sure they’ll not be left standing without a chair when the music stops.

So congratulations to those operators who have the nous to emerge as leading, capable providers in the confusing multi-protocol landscape that will emerge in the next 2-3 years. You just found your product differentiator again.

IPv6, The Final Frontier: The real problem is not your network. It’s your people.

August 9, 2010

Salar de Uyuni

You must have seen the memo by now. Yes,  that’s it – we’re running out of numbers.

If you’re in the Internet connectivity business (and you’ve had your head in the sand) you’ve got a big problem. The exhaustion of IPv4 is clearly a business continuity issue, but not only for the infrastructural reasons you’re thinking of.

As I write this only 5.5% of IPv4 address space remains unallocated. The sequence of events that will occur as we approach the zero hour and beyond is now fairly well understood. We will see several stages of depletion, by the end of which vintage address space will become a highly valued commodity on the black market, possibly resulting in a Mad Max type scenario with everyone  scavenging for IPs in a post-apocalyptic wasteland.

Enter IPv6. The last logical addressing scheme we’ll ever need; the saviour of the Internet, that has been patiently sitting around for two decades, waiting for someone to show it some love.

The discussion thus far has been focused on three areas:

  1. Whether IPv6 will ever ‘take off’.  It will.  If you don’t believe this then what you’re actually saying is that the Internet will just stop growing, in which case you should start hanging out with these guys.
  2. Countdown updates, accompanied by articles on how silly everyone is for not taking action sooner.  We do of course all understand the lack of business case. “Markets trumps technology. And politics trumps markets” as Craig Labovitz said in this post .
  3. The best way to crowbar IPv6 into our IPv4 world and make it play nicely. No, translation really isn’t going to solve this. Get to used to dual-stack. If two parties who speak different languages want to converse properly, one of those parties needs to become bi-lingual.

Talking about the technical deployment, or lack thereof, is all well and good, but everyone seems to be ignoring a key issue here.  I’m talking about the changes required in another core asset, the one you can’t upgrade so easily:


The world has been using IPv4 for three decades. Those dotted decimal numbers are all the IT industry have ever known. Network engineers work on instincts developed through experience; they know how to provision IPv4 services quickly and what to look for when solving a fault. The concepts and mechanics of IPv4 are ingrained into their minds and many of these habits will not carry over to IPv6.

Do you think providers can train their entire support staff up to the IPv4-equivalent standard overnight? Of course not. The ability of these teams to function efficiently is built on real-world experience on real-world operational networks. The problem is that hardly anyone has a production IPv6 network yet. Everyone is waiting for the party to get started.

There is a big skills gap that will become apparent once ISPs start selling and supporting IPv6 services. A number of competent support engineers I know have tried reading the book we’ve all seen lying around. They didn’t get very far.  My own eyes glazed over during chapter two.  It’s a difficult, dry subject for most people to digest easily, perhaps with the exception of a few highly driven senior techies – the ones who are waiting to be given the go-ahead on IPv6 rollout.

This isn’t your everyday network upgrade. It’s not just an isolated extension of things people already understand. This is an entirely new logical overlay and it affects the whole network and all the systems on it. Imagine the effect of having all support issues on new services, however trivial, immediately escalated to the top of the support chain. I don’t think those senior engineers will be doing much else.

ISPs are rapidly approaching the day when their IPv4 networks start to become ‘legacy’. The evangelists are telling them that if they haven’t started technical deployment then it’s already too late.  Maybe it’s time we added another item  to the to-do list: “address skills gap”.