If there were an award for the most disruptive technology that's never been used, IPv6 would be the hands-down winner. Originally conceived of a decade ago, IPv6 - the successor to version 4 of the Internet Protocol (IP) technology that underpins the Internet and a growing number of corporate networks - has languished in laboratories and product sheets as vendors, customers and even standards bodies find increasingly sophisticated ways to avoid having to use it.
On the surface, IPv6 is a significant improvement over the previous technology: its most commonly mentioned benefit is its mind-boggling address space, which at 128 bits (compared with the 32 bits of IPv4) offers 2128 possible addresses. This is a number so big - 3.4 x 1038 - that it would allow each of the 6 billion people on earth to have 56.7 billion billion billion IP addresses, or support 79 billion billion billion sub-IP addresses for each currently available IPv4 address.
That's more than enough to last a while, even with a planned addressing scheme that will assign ranges of addresses to individual service providers and may include some geographical correlation. This should make IPv6 more logically coherent than IPv4, where two similar addresses may point to devices on opposite ends of the earth.
Consistency in addressing isn't the only feature being built into IPv6. Recognising that the chance to upgrade the world's fundamental language for data exchange doesn't come too often, engineers have been packing IPv6 full of everything you could ever want in a protocol: built-in security in the form of IPSec (secure IP); quality of service through mooted use of MPLS (multi-protocol layer switching); and a host of other features to make IP packets travel more efficiently than ever before.
A NATural evolution
Despite its seeming promise, however, IPv6 has been a technical and commercial lemon, mainly because the many participants in the Internet recognised early on that it would be easier to work around IPv4's limitations than to execute a forklift upgrade on the billion or so networked devices currently using IP. To do so would be akin to having Parliament declare the new national language of Australia is Esperanto - it would eventually work, but would require one very long learning curve in which nobody got much of anything done.
Aiming to stave off wholesale technological collapse, technology vendors have instead backed NAT (network address translation), a chameleon-like technology that lets one gateway router have just a single Internet IP address but manage thousands of connections on the private side of the gateway. NAT has quickly become commonplace around the Net, extending the useable life of an IPv4 namespace that many observers had said would be oversaturated five years ago.
Yet NAT has its own limitations. Once they were esoteric problems, but now they are becoming more common as business applications like voice over IP (VoIP) and videoconferencing become more common. The connection-oriented IP sessions that make such technologies possible don't warm to the deceptions posed by NAT, instead demanding that end points have real, verifiable IP addresses. The relative paucity of applications like VoIP until recently meant this didn't pose a serious problem, but the limitations of NAT are becoming increasingly apparent as the Internet continues its unchecked growth.
"We still see a number of limitations with NAT," says Simon Newstead, Asia-Pacific Regional Product Manager with Juniper Networks, which recently won a bid to build IPv6 and MPLS into a network upgrade by Taiwanese carrier Chunghwa Telecom HiNet. "It's been something that's been forced on everyone, and it's far from perfect. IPv6 reduces the requirement for NAT, and growth in the end-user consumer devices supporting IPv6 - and their needing to be always on - will drive the growth of the carrier IPv6 networks."
Since commercial customers have little need for IPv6 at this point, carriers have the luxury of determining their own timeframe for rolling out the technology. Worldwide, most have tried it in one way or another: peak body the IPv6 Forum (www.ipv6forum.com) currently lists 65 IPv6 testbeds and commercial services in 24 countries.
Japanese carrier NTT/Verio, for one, launched its IPv6 Global Community (www.v6.ntt.net/globe/index_e.html) in 2001 and now offers IPv6 gateway, tunnelling and ADSL using IPv6. US carrier Sprint has been testing IPv6 since 1997 and currently has around 250 customers - mostly academic and research institutions - using its freely provided IPv6 testbed to get acclimated with the protocol's capabilities. "We're ready to roll it out, but it's a function of when customers really want it," says Sprint Principal Network Architect Rob Rockell.
In Australia, IPv6 remains all but invisible save far-flung efforts such as AARnet's investigations into the technology, which are being headed by a six-member IPv6 Working Group. Commercially, the protocol has been a no-show since carrier UeComm partnered with Ericsson to launch the country's first IPv6 network in mid 2000, claiming that "most experts agree that IP addresses will run out some time in the next 1-2 years."
UeComm initially planned to install more than 70 Ericsson AXI 462 switches in five cities by early 2001, but has since scrapped the service after it was "deemed not to be commercially viable," according to spokeswoman Katrina Walker. "We are maintaining a watching brief on its development," she says, adding that company engineers believe 2006-2007 may be a more realistic time frame.
Crossing the IPv6 archipelago
In the intervening years, IPv6 will likely follow a slow ascension into the infrastructure of the Internet. IP-connected products of all sorts are beginning to get IP stacks that support both IPv4 and IPv6, promising a two-headed existence that should facilitate the move to the new standard without forcing customers to think much about it.
Supporting IPv6 within IP stacks at network endpoints is just one part of the game, however. Since any changes to IP must be supported by every Internet router to ensure those changes will persist from origin to destination, current efforts at introducing IPv6 have been focused on how to successfully push IPv6 packets across the IPv4-based Internet.
In most cases, the answer has been tunnelling, an approach that's come into widespread use with rapid take-up of Internet virtual private networks. To move IPv6 packets over the Internet at large, they're hiding those packets inside a conventional IPv4 packet, which is transported as normal and then unpacked once it reaches an IPv6-capable gateway that's able to rebuild the packet in its original form.
This approach has been widely used, providing a workable way of linking what NCR Australia Senior Consultant David Gabo calls "IPv6 islands" - the universities and consortia currently using IPv6. Over time, as IPv6 capabilities become more commonly found in other parts of the Internet at large, those tunnels will become shorter and shorter. "The islands are going to begin to encroach onto the core of the Internet," says Gabo, who sees emerging wireless networks and their myriad connected devices as the first application that could commercially justify use of IPv6.
Dual-standard equipment, with tunnelling used to span areas that haven't made the upgrade, will remain the norm until commercial imperatives kick in. Yet while this solution works, it necessarily limits the utility of IPv6 because the strength of its advanced features - QoS and security most notably - is diluted across the IPv4-based Net. The endgame, of course, is end-to-end IPv6, something that's currently only possible on closed IPv6-only networks, which are being linked by dedicated exchanges such as NY6IX and NTAP in the US, AMS-IX in the Netherlands, and NSPIXP-6 in Japan.
Bulking up to handle IPv6
Ultimately, how the carriers bring IPv6 to customers will be irrelevant. More important will be discussions around the business benefits the technology might provide and what's involved in going live with the technology. By the time customers can build a business case around IPv6, odds are that they'll already have all sorts of products - ranging from operating systems to network switches, from computing appliances to packet-based 3G mobile networks - capable of using the technology.
Initially, support for the new standard was typically implemented through a software change to switches' controlling software, but it was not specially accommodated for in hardware. Cisco Systems formally incorporated IPv6 capabilities into its core IOS operating system early in 2002 but will soon offer hardware-assisted IPv6 support across its product range as IPv6-targeted ASICs (application specific integrated circuits) are developed.
Although IPv6 can be implemented at the software level, large-scale application of the technology requires hardware support, since the standard's address size and technical requirements - for example, that all packets be handled using IPSec - imposes a considerable processing burden on the routing process.
"Building hardware that works with a larger address space requires more effort," says Michael Boland, Distinguished Systems Engineer with Cisco Systems Australia. "If you want to filter packets with large address types, custom ASICs are required. It's not trivial, but it's doable. We spent a lot of time getting our software right for the IPv6 supporting service layers, then we'll inject the hardware as the market requires."
The need for hardware to support IPv6 has become a common feature across the industry. Extreme Networks - which has supplied IPv6 gear to primarily academic customers including Japan's Kobe and Kanagawa Universities and Shanghai Jiaotong University in China - has met this need with an add-on hardware blade that provides wire-speed IPv6 packet handling. But IPv6 will remain an optional feature; Extreme has no plans to support it in hardware for its chassis.
At this pace, odds are that IPv6 will become relevant to most companies during their next network upgrade. "In the US, whenever we're responding to tenders it's always a check off item on the list," says Charles Cote, ANZ Sales Director with Extreme Networks. Customers want to know it's on our road map."
IPv6, then, will permeate the corporate world by osmosis. And once it's in their equipment, availability of IPv6 throughout the networking equipment space will drive many companies to at least consider migration strategies. Now that naysayers claiming an impending IPv4 overpopulation have been silenced, time is a luxury available to everyone.
