Beefing up the world's IP address book
Known largely for its hefty IP addressing capability, Internet Protocol version 6 (IPv6) is already making waves in Asia as the region's well of IPv4 addresses runs dry, but those ripples probably won't be felt in North America for several more years.
At the same time, Internet equipment vendors have already introduced products that support IPv6 or have placed those enhancements squarely on their roadmaps to take advantage of the situation in the Asia-Pacific region. Still, making the change to IPv6 is much easier said than done, and, domestically, it remains a relative long-term strategy.
An IPv6 address is 128 bits. An IPv4 address, in comparison, is 32 bits. Some have argued that a larger address could make it a difficult proposition to manage deeper routing tables, but Internet equipment vendors have countered that the scalability of their routers will erase the issue.
On an application level, IPv6 can carry voice and video alongside data. With enhancements, IPv4 can do that, too, but the process "gets pretty old," says IBM Technical Evangelist Laura Jeanne Knapp, who presented an IPv6 tutorial last month during the SCTE Conference on Emerging Technologies.
Today, Network Address Translation (NAT) lessens the strain on the pool of available IPv4 addresses by spawning virtual domains.
"That's what MSOs all do," explains Tim Wu, technical marketing director for Riverstone Networks. "As long as traffic goes through their router, they can get it to the virtual IP address. If not for NAT, we would've exceeded IPv4 [capacity]. NAT's not necessarily the right thing, but it's been good enough."
Knapp argues that the prevalence of NAT is hindering IPv6 deployment. "But we're starting to see NAT erode as a primary v6 inhibitor," because it doesn't offer much help with multimedia applications, she adds.
Calling IPv6 an evolutionary step in the Internet environment, Knapp notes that IPv6's beefier addresses only scratch the surface compared to its other capabilities.
The advanced protocol also features standard encapsulation, quality of service, multicast support, built-in authentication and security, and the ability to coexist with legacy IPv4.
On an application level, IPv6 eliminates v4's Type of Service (TOS) bit and replaces it with a Flow Label, which identifies special QoS requirements for time-sensitive applications like streaming video.
Despite those advancements, the IPv6 movement hasn't progressed as rapidly as some had originally thought. "It was supposed to be everywhere four years ago," says Wu. That may change soon, thanks to developments in the Asia-Pacific region.Asia feels the IP squeeze
IPv6 is not just a technical issue, but a geo-political one as well, due to the current imbalance of IPv4 addresses that have been allocated to nations around the globe.
Most of the best blocks of IPv4 addresses were awarded to the United States in spades. That left other regions in the world, especially Asia-Pacific, with a dearth of top-notch IP addresses, which stirred up trouble when Internet usage started to explode there.
The running joke is that the Massachusetts Institute of Technology has more IP addresses than all of China. Wu calls that discrepancy an "accident of history." Plus, the problems stretch beyond just IP addresses to include address "quality."
Prior to the Internet's version of the Big Bang, U.S. companies, universities and organizations were awarded the best addresses in the IPv4 lot, Class A, which provides 16 million sequential IP addresses. Class B and Class C, which grants a mere 4,096 sequential addresses, were basically left for everybody else.
With IPv6 as a global standard, it's expected that the Internet Assigned Number Authority (IANA) and its regional authorities will hand out those addresses much more judiciously this time around. That will come in handy in Asia, where Internet phones with always-on IP addresses and multiplayer gaming already are making a deep impact.
While multiple client-server gaming networks work well in a NAT environment because users are required to log on each time they access them, a shift to peer-to-peer gaming requires more investment in the infrastructure and separate IP addresses for the players, says Cisco Systems Inc. Distinguished Engineer Stephen Deering.
Globally speaking, Deering sees the IPv6 pull coming from three directions: regional (particularly in Japan and Korea, where IPv6 deployments are already underway); political motivation (IPv6 has more international context than IPv4 does); and commercial motivation (the adoption of IPv6 could open up the market to more equipment vendors).Domestic drivers
Internet observers believe advanced wireless applications and IP-based telephony will initially drive a migration to IPv6 in North America, with home networking applications coming along further down the road.
IPv6's first true beachhead in the U.S. could come in the form of VoIP, Deering predicts. "Trying to build a large-scale IP telephony service without unique IP addresses is very challenging," Deering says. "You can do it through NAT, but it's more complicated and fragile."
Intelligent cell phones based on the advanced 3G platform will also make an impact. "3G is a tidal wave everyone can see out there," Riverstone's Wu says. Knapp agrees that 3G will represent the first driver of IPv6 for the business sector, but says nothing for home use has emerged to goose IPv6 in North America.
That day is coming, she adds, predicting that consumers, not the commercial sector, will form IPv6's big initial push. On that, she points to how cell phone usage rocketed only after consumers really took to the technology.
Another potential IPv6 driver is the so-called "smart" networked home, where not just PCs, but home monitoring equipment, refrigerators and other household appliances all need their own IP addresses, Knapp says.
Adam Stein, director of corporate marketing at Juniper Networks, predicts that tiered data services will be among the first applications to push IPv6 among U.S. cable operators.
"MSOs have to prepare to offer all of these services at the same time without bringing the CMTS (cable modem termination system) performance down," Stein says. "Today, the vast majority of legacy CMTSs out there are CPU based. If you try to put more than two services on them, they move to a very slow crawl."
To fix that potential problem, Juniper bought Pacific Broadband Communications for about $200 million, plus access to PBC's next-generation CMTS chassis, the G10, which is designed to boost performance by handling packet processing on the ASIC (Application Specific Integrated Circuit) instead of the CPU, which, Stein says, isn't optimized for that.IPv6 migration to take work and time
Plenty of work must be completed before IPv6 becomes widely available. For starters, routers require upgrades, protocol stacks need to be updated and software must be rewritten to translate IPv4 data for the IPv6 world, and vice versa. And, until IPv6 becomes ubiquitous, Web sites will likely have to support both versions. IPv6 "is a major protocol change... It's not like moving from v3 to v4," Knapp warns.
"It's not a small operation," Wu concedes. Although the Ethernet side of the router is untouched in an IPv6 environment, "routing algorithms don't change, but every chip that handles IP packets has to recognize v6 packets, which is a top-to-bottom change," he adds.
That translates to a heap of equipment. For a full, ideal change out, IPv6 would have to be active in about 1 million intelligent Layer 3 devices and a core network made up of about 100,000 devices, Wu estimates. Lopped on top of that are the 106 million active users in the Internet universe, according to Nielsen/NetRatings' latest figures.
As for a migration among MSOs, a commitment from a major operator will have to occur before IPv6 begins to snowball, Wu predicts. But, before that happens, operators will probably want assurances that making the shift represents a bona-fide revenue driver. A migration to IPv6 "will be costly to invest in, and it will come with a long-term profit model," Wu says.
Today, there's no real value proposition to IPv6 beyond the extra addresses, notes Zeus Kerravala, research director for the Yankee Group.
Because initial IPv6 initiatives will originate from government mandates in countries like China and Japan, "I don't necessarily think it will happen overnight, but we'll see a few v6 islands [in 2002]," he says.
By 2004 to 2005, Europe will get IPv6 underway, and the U.S. will begin to show some interest, Kerravala predicts. In the 2006-2009 timeframe, the U.S. will begin to see strong consumer adoption, with commercial adoption another two or three years behind. It will be at least 2010 before many businesses retire IPv4 applications in favor of IPv6, Kerravala estimates.
Although companies such as Japan's Nippon Telephone & Telegraph are moving their backbones to IPv6 now, the technology is already out there today.
The 6bone is a public IPv6 network that enables testing for the technology's early adopters. The 6bone features IPv4 compatible addresses that can communicate with v4 devices. Internet2, meanwhile, is a closed, university IPv6 network designed to make advances and changes in the technology for the commercial sector.
IPv6 is also supported in popular PC operating systems, including Linux and Microsoft Corp.'s WindowsXP, which has "an exceptional v6 implementation," Knapp says.IPv6 support
While IPv4 will continue to play a prominent role for the foreseeable future, Internet routing vendors are preparing for the eventual shift to v6. Already, many have added IPv6 support to their software and hardware.
Riverstone Networks, for example, expects to offer IPv6 support in hardware and software in 2002, Wu says.
Cisco Systems, meanwhile, started offering IPv6 support in software last May, and is making preparations to phase in support for the hardware side. "We've demonstrated hardware forwarding on the 12000 series routers in Japan," says Deering. "We have several projects underway."
Deering says it "would be nice" if Cisco had IPv6 hardware available today, although the company isn't particularly worried that its own progress has fallen behind general adoption of the technology.
"No one will write IPv6 stuff without the platform," he says. In Korea and Japan "they're doing deployment before the applications. They don't need hardware forwarding at this point because there's negligible v6 running over the network."
That hasn't stopped Juniper from enhancing its software and hardware for a v6 environment. "We do a quite a bit in hardware today," Stein says.
Last November, Juniper launched production-ready IPv6 products for its M-series Internet access and core router platforms and interfaces based on its Internet Processor II ASIC and the latest release of its JUNOS Internet software.
Supporting v6 in software alone creates immense platform degradation, Stein argues, noting that Juniper will leverage PBC's ASIC products and expertise to add software triggers that perform hardware functions.
Juniper has also won some initial support for its v6 products. For instance, France Télécom Research and Development is using Juniper products to manage a very high bandwidth platform (VTHD) for research institutions, universities and engineering schools.
Of course, IPv6 support among vendors is not limited to Cisco, Juniper and Riverstone. Last October, Foundry Networks said it is integrating IPv6 into its line of Layer 2–Layer 7 switches and Internet routing products, and 3Com Corp. unit CommWorks Corp. has added IPv6 support to its Total Control 100 gigabit routers.
But it's not yet time to start waving good-bye to good 'ol IPv4.
"We tend not to call it a migration," Deering says, "because two versions will coexist for a very long time. It's difficult to move the world. IPv4 will linger around."
Linger, or maybe remain a perpetual presence. "You'll never see v4 go away," predicts Kerravala.