Time's up for putting off IPv6 decisions
With the Internet of Things, we're going to be needing a lot more IP addresses. That's exactly what IPv6 has in mind.
Computerworld | May 15, 2015 8:56 AM PT
As companies become used to the Internet of Things, they are assigning IP addresses to everything from printers and watches to refrigerators and garbage cans. If we're all going to be moving that much closer to recreating a Jetsons episode, we're going to be needing a lot more IP addresses. That's exactly what IPv6 has in mind.
IPv4 is expected to run out of new available addresses in a matter of months, forcing a truly near-term move to IPv6. How bad is the shortage? The Wall Street Journal offered some sobering stats. Back in 1981, IPv4 launched with 4.3 billion addresses. Today, barely 3.4 million are left. IPv6 is launching with 340 undecillion addresses. Yes, that's a real number. It's equivalent to one trillion times one trillion, which is roughly the length of my to-do list when I return from vacation.
On the one hand, that makes it sound like moving to IPv6 should be a no-brainer. Well, it is, but exactly how to do it — without abandoning the overwhelming majority of global systems that are still using IPv4 — is a lot trickier.
Also, despite the age difference between IPv4 and IPv6, companies are unlikely to see any meaningful performance improvements when making the switch, so this really is all about capacity.
"Early research at Akamai has shown sometimes IPv4 paths exhibit lower latency and sometimes IPv6 paths do. This suggests IPv4 and IPv6 perform similarly, but also vary similarly over time," said Arthur Berger, a principal research scientist at Akamai.
There are various ways of making the transition gradual, including middlebox translation, which is a good — but not universally available — option. "It's not supported in every network," said David Belson, a senior director for industry and data intelligence at Akamai. "Duct tape and chewing gum solutions are going to eventually wear out."
Dual-stack models supporting both v4 and v6 are going to likely be the more popular option.
When to start supporting both? Getting v4 addresses now is quite expensive, since the best source right now is to buy such addresses at auction from companies that already have more than they need — or that are going bye-bye. In retail, it was simultaneously sad and amusing that when the assets of the once-mighty Circuit City and Borders were sold at auction, the most valuable item they offered were their IP addresses.
Hence, companies need to make the move to v6 right away, admittedly without ditching v4. As for when to make the complete switchover, that could easily take years and quite possibly decades, as the last holdouts for v4 will likely be stubborn and fairly numerous. Here are some credible peeks at where IPv6 acceptance stands today, both by country and by telco. In short, this will take a while.
"There are networks that will run IPv4 forever," said Phil Roberts, the technology program manager at the Internet Society, where he specializes in IPv6 global issues. "The first thing you want to do is get your public-facing stuff on IPv6."
That's good advice, but in an Internet world, how is "public-facing" defined? Even the most internal systems — say, for example, an expense report app — are going to be accessible through an intranet for employees in locations across the planet. And inventory and ordering systems will have to be available for suppliers and resellers via a hopefully secure extranet. (Roberts clarified that he meant unrestricted content.)
Recrimination is rarely helpful, but seriously, isn't the running out of IPv4 addresses the most predictable IT problem since Y2K? Given that fact, why the delay? This is the latest example of the consequence of having enterprise CIOs as the highest-paid temp workers in the industry. The enterprise CIOs who stay in those roles more than six years is the extreme minority. This encourages avoidance when it comes to expensive and time-consuming projects that won't deliver near-term profit payback and also won't deliver near-term pain.
If there was ever a project tailor-made for "I'll let the next CIO worry about it," it's IPv6.