IP addresses are running out, but as Douglas Adams might have said: don't panic. We explain the IPv6 solution and what it means for you.
Allocation of the last extant blocks of IP addresses in February was a wakeup call for those who thought the net could keep growing as it is, forever. When peak body the Asia-Pacific Network Information Centre (APNIC) gives regional internet service providers (ISPs) access to its final ‘/8 block’ of 16,777,216 addresses later this year, Australian ISPs will, like their peers around the world, face a critical transition point in the internet’s evolution.
The solution for this problem — IPv6 — was first formalised in 1996, but patchy implementation and successful workarounds have kept it on the backburner. It’s no longer possible to ignore IPv6, however — especially since the two technologies can’t communicate without the use of a purpose-built gateway.
Tomorrow's a very important day: the "test flight" for the future of the web.
This will present issues for ISPs that need to ensure their customers stay connected as the world’s networks shift towards IPv6 in the coming years. Yet not all of them are ready to go live: “It’s a business failure rather than a technical issue,” says Mark Newton, network engineer with Internode, which has a number of eager customers running on IPv6 today. “Many network builders haven’t been inspired to spend money on switching to IPv6 while IPv4 has still been available, and there have been more pressing things to dedicate resources to.”
In the short term, existing ISPs will support both IPv4 and IPv6 ‘stacks’ — the collection of software protocols that manage the addressing and delivery of IP data — in parallel. This will ensure existing configurations continue to function while letting customers move to the new technology as network operators and available equipment allow. It will also ensure compatibility for operating systems and applications, which have been written to IPv4 standards for two decades and are in many cases unprepared for its successor.
A new approach to IP
Just as the explosion of fax machines, pagers and other devices drove Australian communications regulators to rework the nationwide phone numbering scheme in the late 1990s, the transition from IPv4 to IPv6 has come about from an explosion in demand. The internet is no longer just about home PCs: everything from mobile phones to cars, traffic lights to railway signals, trains to airplanes and building alarms to surveillance cameras are now getting online — and each needs an IP address.
IPv4 was never designed for this. It offers 32-bit addresses — which allow 232, or around 4.3 billion, possible addresses represented by four eight-bit numbers (with a value of 0 to 255) separated by full stops. IPv6, by contrast, offers 2128 –340 billion billion billion billion — different addresses, in eight groups of four hexadecimal digits separated by colons.
This may seem like overkill, but nobody wants to have to rework the internet again any time soon. Yet IPv6’s larger address space isn’t about sheer numbers: its designers took the opportunity to fundamentally redesign the protocol to work more efficiently, and to support a broader range of features.
For example, IPv6 mandates the use of IPSec, a security protocol that was a bolt-on to IPv4; allows use of unique 48-bit and 64-bit Media Access Control (MAC) — built into every piece of networkable hardware made — to simplify addressing; supports multicast IP, a method for broadcasting data to many endpoints at once that’s optional in IPv4; enables address autoconfiguration that eases the allocation of addresses to devices when they come online; does away with time-consuming activities like calculating checksums to ensure data integrity; offers options to designate quality of service levels and mobility enhancements; and more.
In other words: where an IPv4 packet is primarily concerned with routing the data from one place to another, an IPv6 packet offers more space to detail how that data is to be delivered and secured. IPv6 addresses are split into two discrete elements: a 64-bit subnetwork prefix that indicates which network the device lives on, and a 64-bit identifier that identifies the specific interface (read: device) to which the address points. This lets applications manage networks and devices independently without the complex IPv4 system of ‘masks’.
To consider the difference, remember how mobile phone numbers worked before network portability was introduced. You could tell which network a person was using based on their mobile phone number prefix: 0408, for example, was a Telstra phone while 0403 was Optus and 0414, Vodafone. Now numbers can be ported between carriers so the prefix is no longer a guarantee; mobile networks have to perform extra steps to figure out which carrier the number belongs to, and route the call accordingly.
IPv4 has traditionally taken a similar approach, especially with the proliferation of NAT (network address translation), a technique designed to work around the looming exhaustion of IPv4 addresses. You use NAT all the time without even realising it: your broadband modem maintains one IP address for the rest of the internet and separate local IP addresses (usually starting with 192) for the various devices connected to your network. NAT uses IP ‘port’ numbers to tag traffic coming from a networked computer or device as it goes onto the internet. Data returned from the remote host also carries that port number, which the NAT recognises and sends back to the correct device on the local network using its 192-prefixed IP address.
Given the massive number of addresses in IPv6, NAT will no longer be necessary: individual devices can not only have their own internet-addressable IP address, but can be grouped by network. This makes IPv6 an inherently ‘flatter’ network topography that will be faster to run and easier to manage.
In environments where IPv6 is not yet ubiquitous, ‘tunneling’ techniques encapsulate IPv6 data within IPv4 packets using techniques like 6to4, a standardised method that flags packets with IP port 41 and embeds information in both standards within the data stream. 6to4’s use of port 41 means it can be blocked by many NAT systems, a deficiency addressed by tunnelling techniques like Teredo and ISATAP.
Managing the transition
It will be many years before IPv4 can be relegated to the dustbin of history, but most important is that ISPs and equipment makers upgrade their networks to ensure that IPv6 packets can be carried end to end as necessary. Many of Australia’s major ISPs — iiNet and Internode are often held as the most progressive — are advanced in their transition planning.
Just how far we’ve come — and how much is yet to be done — will become evident on June 8 (tomorrow), which the global Internet Society has christened World IPv6 Day
. On that day, global internet properties and vendors around the world — including Google, Facebook, Yahoo!, Akamai, Limelight Networks, Cisco, Voxel, Juniper Networks, Huawei, Microsoft Bing, Mozilla, Fortinet, Internet2, CANARIE, and many others — will run in dual-stack configurations for 24 hours, stress-testing them to see how well the internet’s infrastructure can cope with the new IP.
If all goes well, you won’t notice any difference in the transition to IPv6; by the time it’s needed, your ISP will have any issues sorted. Yet there’s every chance of hiccups during the biggest-yet overhaul of the internet’s workings. “We’ve really only aimed at the geek end of the marketplace so far,” says Newton.
“It has been fairly painless for most people, but we tend to be skewed towards people that can work out problems themselves. We’re about to get started with the next wave of user that might need a bit more hand-holding. Everyone’s going to have to get a new modem at some point – but I don’t see that as a huge problem because the NBN is coming, and everyone’s going to have to get a new modem anyway.”