IP4 vs. IP6: Two very different animals
July 10, 2014
Print this page
As you read this, the last blocks of address space based on Version 4 of the Internet Protocol (IPv4) are being allocated in Latin America and Africa. In the early days of the Internet, no one had anticipated the volume of Internet-connected devices would reach the volume it has today; they certainly had not forecast the billions of sensors, readers, valves, industrial machinery, etc., that are still to come online.
When IPv4 was developed, mainframes and minicomputers ruled the computing roost, according to Phil Roberts, technology program manager with the Internet Society. The widespread success of the personal computer, let alone Internet-connected mobile devices, was not on anyone’s radar. The 4.3 billion addresses afforded under IPv4 seemed sufficient.
Fortunately, Version 6 was waiting in the wings. IPv6’s 128-bit addressing scheme, in development since the early 1990s, provides considerably more address space than IPv4’s 32-bit scheme. In fact, IPv6 allows for 340 trillion, trillion, trillion IP addresses — 48 billion, trillion, trillion for each of the seven billion individuals on earth. It’s virtually inexhaustible.
Still, there are issues. The two protocols are not compatible; transition technologies have to be used to allow the two to co-exist on the Internet. And while IPv6 was launched with a flourish in 2012, the vast majority of Internet traffic – about 97% as of March 2014 – is still travelling on IPv4.
Rewind to 1993, when work on the new protocol was beginning, says Alain Durand, now a Juniper Networks Inc. distinguished engineer, then a representative of Sun Microsystems Inc. on the advisory body creating IPv6. The public Internet was new and growing exponentially; the number of hosts was doubling every 18 months.
“The thinking was, ‘we are going to deploy this IPv6 thing very quickly, maybe it would take about 18 months, 24 months max,’” Durand says. The legacy IPv4 network would be a small, irrelevant portion of the Internet, so compatibility was not a priority.
“Obviously, we were wrong then. It took much, much longer to get deployed.”
With an estimated 20-year-window until address depletion, carriers and enterprises simply did not bother. “Going to IPv6 is not free,” says Roberts – there are considerable development and deployment costs. “In the absence of need, costs are not going to be paid.”
In Canada, uptake has been infinitely small. According to Vyncke.org, a Belgian Website that displays IPv6 traffic to search giant Google by country in near-real-time, IPv6 penetration in Belgium was about 15% in March 2014, and about 7% in the U.S. Canada’s IPv6 penetration was a miserly 0.45%.
“It is a fluke of history and geopolitics,” says James McCloskey, executive advisor with London, Ont.-based Info-Tech Research Group Ltd. Due to our proximity to the U.S. and a history of advanced technology, Canadian Internet service providers were assigned disproportionately large blocks of IPv4 address space for the number of devices they are serving. The IPv4 address depletion crisis has not hit home in Canada.
That is the good news. “The bad news is (organizations) have not voluntarily gone down that path” to IPv6 adoption, McCloskey says. Whether they need the address space or not, organizations will have to deal with IPv6 traffic from other jurisdictions.
“Enterprises need to own the technology themselves and not rely on the ISPs,” McCloskey says. “ISPs are in the business of delivering traffic.”
Transitional Technologies: Because they are such different animals, a variety of transitional technologies are used to get them to co-exist across the Internet.
Back in 1993, the only compatibility mechanism created was a dual stack – a network that could forward IPv4 and IPv6 side-by-side, says Durand. It Is like being bilingual: “I’m French, I can speak French to my family, but I can speak English to you.”
Tunneling – encapsulating IPv6 traffic within IPv4 packets, and vice versa — is another way to bridge the divide.
“Tunneling is a very powerful technology,” McCloskey says.
Tunneling technologies including 6in4 and 624 allow traffic to initiate in IPv6, transit in IPv4, and unpackage in native IPv6 at its destination. The encapsulation does add some overhead, resulting in a latency that a consumer user probably would not notice, McCloskey says. But for a grid or high-performance computing architecture, it might be too much.
More recent trends in transition technology revolve around layered network address translation (NAT), says Matthew Wilder, senior engineer for the office of the CIO at carrier Telus Corp.
“It is a way of doing it that is much like tunneling, but actually consumes a lot less resources on the CPE, on the network, by using essentially stateless address translation,” Wilder says. There is less computing required to process the packets, and failover is better because there’s less “state” to keep track of – the network treats every transaction individually without reference to session information, he adds.
One example on the mobile side is 464XLAT, which is a combination of NAT in the mobile device and NAT in the network. The handset is native IPv6. If an app needs access to an IPv4 resource, the handset maps that IPv4 address into an IPv6 packet. When the packet hits the network, it translates the packet back into IPv4 to push through the network. That allows Telus to offer purely IPv6 connectivity, Wilder says.
It is a flip of what’s carrying what from tunneling, where IPIv6 is encapsulated within IPv4 packets, says Wilder.
Infrastructure Issues: Not every piece of hardware in the stack is affected by Ipv6 compatibility issues.
“IPv6 is a Layer 3 protocol, so essentially when you hit Layer 3 in the stack, that is where (you are affected),” says Durand. A bridge or a dumb switch would not need upgrading, but an intelligent switch or router that plays in Layer 3 might. But most routing silicon has been IPv6-compatible since the mid 2000s, says Durand. (Operating systems later than 2005 support dual stack, as well.)
It is not the same story for consumer routers and SMB gateways, Durand says. “They are limited in capacity and may not be upgradable to IPv6.”
Software-defined networking – or, more specifically, network function virtualization (NfV) – to the rescue.
“If we could replace this physical box by a virtual function somewhere in the cloud, we could get some benefits,” Durand says. “It may be an easier way to deploy IPv6.”
But the infrastructure issues do not end with the hardware, says Roberts. Supporting infrastructure, like customer support and billing, is also affected.
“There is a lot of stuff that has to be touched,”he says. “The good news is a lot of the work has been done.”
It has been a long, slow trans tion to IPv6, says Rob Soderbery, senior vice president of the enterprise networking group at Cisco Systems Inc., but one that will be accelerated as exponentially more devices are connected to the Internet.
“Think every light switch, every door lock, every motion sensor, every video camera, every outlet, everything connected to the Internet,” he says. Enterprises are not going to spend the upgrade money just for a new management protocol; they are looking to take advantage of new business applications. BC Hydro has 2.4 million smart meters and counting under management. One large airport in Europe has 6 million managed endpoints.
“IPv6 becomes a driver for new business applications and then a mandated part of the system,” Soderbery says. “Now we are seeing the excitement turn to, what is the next business application, what is the provider use case that is driving the adoption and the rollout?” See also p. 38.C+
Dave Webb is a Toronto-based freelance writer. He can be reached at firstname.lastname@example.org.