"Peer-to-peer is what we think is possibly the next computing frontier," Pat Gelsinger, chief technology officer of the Intel Architecture Group, said today in a speech at the Intel Developer Forum here. In peer-to-peer networking, individual computers talk to one other, allowing computing tasks to be managed among those computers instead of or in addition to using a centralized network.
The chip giant estimates that there is two-and-a-half times the computing power in individual computers at a typical large business than is available from its servers. By harnessing the power of each computer to power the network and distribute computing tasks, Intel executives said companies can save billions of dollars.
Intel knows from whence it speaks, having used peer-to-peer networking to manage the complicated task of designing chips since 1990. Gelsinger said the company has saved half a billion dollars during the past decade by shipping simulations from active computers to faraway systems that had been underused.
Gelsinger said that before Intel started the program, it was "buying mainframes like candy" even though it was only using about 30 percent of the computing power of its workstations and 50 percent of its server power. Now, about 70 percent of its workstations' muscle is being tapped and servers are running at about 80 percent.
To help get other businesses thinking along these lines, Intel today announced the formation of a peer-to-peer working group with 19 initial members, including IBM and Hewlett-Packard.
A number of businesses are already using the technology. One type helps companies exchange data both internally and externally. A second type pays home users to harness their computing power. According to Intel, about 75 companies exist whose business centers on peer-to-peer computing.
Gelsinger also identified a number of business areas created by peer-to-peer networking. Among the promising arenas, Gelsinger said, are universal file sharing a la Gnutella and distributed computing, like that of SETI@home.
SETI@home is an early example of the peer-to-peer revolution to come, Gelsinger said. The service now has more than 1 million home computers processing signals from space to find whether intelligent life exists beyond Earth.
Strictly speaking, SETI@home and many other applications are not peer-to-peer but rather are examples of distributed computing where a task is split among many computers but managed by a central server.
Gelsinger was careful to note that it is the technology behind Napster that it is endorsing, not the music-swapping service itself.
"We are neutral on Napster," said Gelsinger, adding that Intel supports intellectual property rights but has no opinion about whether Napster violates those rights.
Gelsinger said Intel is not worried that making better use of individual computers will lead to a drop in sales of its processors. If peer-to-peer takes off, Gelsinger said, it might lead to a fewer number of servers being sold in individual cases, but it will lead to whole new applications that will increase the overall computing demand substantially.
Others working in the industry agreed, saying that peer-to-peer technology will require more processing power, sometimes called millions of instructions per second (MIPS).
"I think it will in general create a demand for more MIPS," SETI@home founder David Anderson said during a panel discussion this morning.
That statement elicited a loud "yes" from Gelsinger, who was sitting in the audience.
Panel moderator Cheryl Currid, an analyst at business and technology researcher Currid & Co., added that companies might be more willing to pay for faster processors if they knew they would be fully utilized.
Not everyone was convinced. At a press conference this week, a Belgian technology reporter suggested that any corporate computing manager who approved using peer-to-peer networking should be fired because of the security risks involved.
Currid said that corporate IT managers will be leery of the technology, as they were in the early days of networks, but she added that the opportunity is too great to ignore.
"Any IT manager who fails to look at peer-to-peer should be fired," Currid said.
Security will be a key to peer-to-peer's success, said panelist Andrew Grimshaw, of Applied MetaComputing, a member of the new peer-to-peer working group.
"People aren't just going to put their $30 billion system on the Net and say, 'Here, have fun,'" Grimshaw said.
Dan Beldy, a venture capitalist at Napster-backer Hummer Winblad Venture Partners, said companies will rapidly look to adopt the technology because of the need it serves and the potential windfall to those who succeed.
"It will happen quickly," he said, "because there is a lot of money out there and people are interested."
Intel plans to start funding companies that are working on peer-to-peer, although it has not yet set a dollar target or established a special fund as it has with other projects.
"You can certainly expect us to play an aggressive role with Intel Capital," Gelsinger told reporters. Many of the investments will be small, however, because a lot of the companies in this area are just getting started.