Avert Imminent Internet Crash

Scott O. Bradner thinks the Internet is going to collapse.

Bradner, a senior technical advisor to the university who built Harvard's first computer network, thinks that unless the computing community takes action immediately, the Internet will break down within two years.

The problem, Bradner explained in a talk earlier this month at the Next Generation Networks conference in Washington D.C., is that the size of the road map that hubs at the very center of the Internet have to store in their memory is growing exponentially.

Six years ago, there were 10,000 local networks making up the Internet that the hubs used to forward traffic from one local network to another had to know about. About a year ago there were 60,000 and today, there are about 90,000.

Bradner thinks that in the next few years, that number will expand exponentially--until it reaches a size that the hubs just won't be able to handle.

To avert the crisis, he says, the Internet needs new and improved algorithms for hubs to use to move data to their destinations. But developing those algorithms, he says, will require government sponsorship.

The Internet's Center

The Internet is an enormous web of interconnected computers, each part of a smaller network and without a single center.

Instead, there are many decentralized hubs called routers that direct data from one local network on the Internet to another.

Each piece of data sent through the Internet has a number to specify one of four billion destinations. Routers refer to a road map--called a "next hop table" in computer terminology--to point the data in the right direction.

When one Harvard students sends another an e-mail message, Harvard's routers know to move it from one Harvard computer to another. This data never leaves Harvard's section of the Internet.

But if a student wants to send an e-mail message outside of Harvard, the university's routers cannot handle this request alone. They send it to a router outside the university to see if that machine knows where to send it.

Today, the next hop tables in the memory of central routers have 90,000 entries. This may sound like a small number with tens of millions of computers on the Internet. But creating a next hop table with this many entries is a computationally intensive process. And each time a new listing is added to the central next hop tables, the entire table has to be recalculated.

The time and computing power it takes to recalculate the central next hop tables after each additional new entry increases as the tables get larger.

The problem, Bradner says, is that the Internet is growing in size exponentially. He argues that in about two years it will take so long to recompute the Internet's central next hop tables that it will be impossible to expand them.

"You need to increase computing power more than is practical and faster than you did earlier on," he says. "At some point you will just have to stop adding new networks."

This will cause a major crisis, Bradner argues, because demand for expansion of the network will continue. The Internet, he says, will have effectively stopped working.

"Under current routing technology, there is no reason to think the Internet will keep working for more than a couple more years because the routers will collapse," he says.

Colleagues Respond

Some of Bradner's colleagues in the world of Internet infrastructure say his view is overly pessimistic.

Fred Baker, president of the Internet Engineering Task Force, a non-governmental organization that works to develop the architecture of the Internet, thinks that Bradner is overly worried.

"Yes, there are issues in Internet routing," he says. "There is not evidence that the Internet is in the process of collapse."

Baker says that everyone agrees that the growth of the Internet's central routing tables is exponential. The question is whether router hardware and routing protocols can be improved enough so that it can deal with this growth.

Baker thinks it can. He argues that as computer memory becomes exponentially cheaper with time, the amount of memory available to computer designers will match the need for more powerful routers.

He also points out that statistical data about the Internet's expansion over the past few years suggests that central routing systems have been able to cope with the exponential expansion of the Internet in the past.

"How many [companies] have announced that they have no plan to build a next generation router?" he says. "If it is impossible or not feasible to build such, then what underlies the business plans of the several startups in the...space?"

David Newman, president of Network Test Inc, which tests router technology, says that Bradner points out a worrisome trend in Internet infrastructure, but it's not yet time to press the panic button.

"I'm a little more sanguine than Scott is," he said. "But there's more than a hint of truth in what he says."

He worries about the impact that five- or ten- fold expansion of the Internet will have on hop tables.

"It's not unreasonable to say that we'll get to much higher numbers--500,000, a million, in the not too distant future," he says.

One senior computer executive points out that some of Bradner's noisemaking about this problem can be attributed to his position as a technical advisory board member at Packet Design, Inc., which is working on the router problem.

"Scott has a financial interest in this," says the executive. "I don't think anything he's saying is factually incorrect. He's smart enough to see that if he does come up with a good idea [he can capitalize on it]."

But Bradner responds that he did not know that the company was working in this area when he prepared his talk.

"I made the statement before I knew they were working on the problem," he said. "Serendipitously, later on I found at the Packet Design is working in this area. I certainly do not think that is a bad idea."

Solving the Crunch

Bradner thinks that developing new router hardware is not going to solve the Internet's problems. An entirely new way of sending data through the center of the Internet, or routing protocol, is necessary.

But no such organization schemes have been developed or are in the pipeline, he says.

"The Internet community doesn't have any projects underway that can guarantee with any degree of [certainty] that the Internet will be running in two years," Bradner says.

He says this is because there is no way to make money making routing protocols and that it is also impossible to get everyone who runs the routers at the center of the Internet to adopt a new, better protocol once it is developed.

"The return on investment is low," Bradner says. "There's no business model."

Bradner argues that because of the extraordinary importance of the router problem, and because private industry has done little to address it, the government should intervene to fund university research in routing design.

"Universities can do something which multiple vendors can adopt easily," he said.

But others argue that private enterprise can solve the problem by developing new router hardware and new routing protocols.

"Private industry has put a lot of money into research for next generation equipment," says Pepe Garcia, from the Routing Systems division of industry leader Cisco Inc. "With next generation routers, routing protocols, and more and more memory, I do not believe the Internet will be in crisis."

Bradner responds that if new routing protocols are developed, the Internet can be saved. He just doesn't think that private industry can create them.

"He thinks that Cisco can do it and I don't think they can," Bradner says of Garcia. "I question that Cisco and Nortel and the rest of them can produce the next generation routing protocols. I think there needs to be government assistance....I would love to be proved wrong."