Storm clouds looming for internet, experts say
IP address shortage and routing table issues raise fears
By Brad Reed, Framingham | Monday, 28 April, 2008
Experts at the recent FutureNet conference said internet architecture will face stiff challenges over the next few years that could affect the web's effectiveness.
FutureNet is an annual conference held to address communications services from the perspectives of enterprises, ISPs and vendors. Panelists and speakers at this year's conference, held in Boston earlier this month, debated just how severe these challenges would be to internet growth and what strategies ISPs and vendors should use to meet them.
One of the most immediate issues facing the internet is the impending drought of IP addresses, which some forecasters have said could occur within the next few years. Ron Bonica, a consulting engineer on Juniper Networks' routing-protocol software development team, said that while the actual exhaustion of IP addresses is probably further away than some had predicted, it is inevitable.
The reason for the coming IP address shortage is broadly recognised: IPv4, the first widely deployed version of Internet Protocol, which offers about 4.7 billion possible IP addresses, is running out of capacity to meet demand for new addresses. The situation has gotten so critical that this week Akinori Maemora, chairman of the Asia Pacific Network Information Centre (APNIC), labeled it a "crisis", while Tony Hain, IPv6 technical leader at Cisco Systems, said that "most people in the world are still in a state of denial" about it.
Bonica said there are three realistic solutions to this problem, the first of which would be to stick with IPv4. While this would create some obvious immediate problems with the impending shortage of IP addresses, Bonica said that it would also lead to the creation of an IP address trading system by which companies and individuals who own an excess number of addresses could sell them at market value. The second possibility, said Bonica, would be a rapid universal deployment of IPv6, the IETF's next-generation Internet Protocol that can support several billion more addresses than IPv4. While Bonica said this was a possibility, he also said many organisations and companies have been reluctant to make the switch, since it would require a lot of investment on the part of end-users and ISPs, and translation mechanisms to help make the switch aren't yet widely deployed.
Bonica predicted a compromise between these two solutions and a gradual progression from IPv4 to IPv6. Although he claimed to have "no crystal ball" in the matter, he said the gradual migration to the new standard made the most sense economically.
"One of these three paths will be chosen for economic reasons, and that will be the one that keeps people in business," he said.
The other big problem addressed at FutureNet was the potential strain that new IP addresses could put on routing tables, the master lists of network destinations stored in backbone routers that decide on the optimal paths between networks. During a panel session titled "The End of the Internet?" several industry experts said that the strain on routing tables was both a short- and long-term concern because routing tables are not scalable and don't adapt to exponential increases in IP addresses. As APNIC member Geoff Huston told Network World years ago, "it's not the size of the table, but the number of updates per second that kills a router stone dead."
Other attendees also commented on the problem, but when asked what long-term solutions for routing table woes might be out there, the panelists mostly balked: They said there weren't any quick fixes that would alleviate the problem and that we'd have to be content with IPv6 alleviating the problem for the time being, with the larger overall problems being considered further down the road.
"The basic, fundamental problems of scaling a network haven't been addressed in any innovative manner," said John Curran, chairman of the American Registry of Internet Numbers. "This current, rough, working model that's been held together with Band-Aids is going to go through an interesting event in a five-to-10-year time frame."
But even with these bumpy changes due to occur in how the Web works, fellow panelist Scott Bradner and others agreed that talk of the web's imminent demise is greatly exaggerated.
"For all the talk about doom and gloom, the experiment that we've been running has been working pretty well so far," Bradner said.