Today's internet is fraught with disconnects. A frustrating experience in Germany made me wish there was something better than client-server relationships. And there is: it's called peer-to-peer computing, and it was the original reason the Internet was started. In the late '60s, when scientists wanted to share research results with their peers, they made hard network connections to and among themselves. The result was that important information was passed among a group of colleagues working on the same or similar problem.
At the recent Intel Developer Forum, Pat Gelsinger, vice president and chief technology officer of the Intel Architecture Group, said that Intel Corp. believes peer-to-peer computing is the next revolution in the Internet. For me, it can't happen fast enough.
Gelsinger detailed an internal media training application that was too expensive to implement using client/server technology but that might be appropriate using peer-to-peer computing. The results, he said, have been good. Intel launched international training courses online comprising 60 modules of 10 to 20 Mbytes each. More than 2,000 participants performed many file transfers and completed training sessions across 50 sites at Intel without disconnects. Intel employees experienced a 5x to 6x improvement in the file access time when more than 80 percent of the accesses were delivered in a peer-to-peer manner as opposed to accessing files from the original source. The company aims to turn the program into a production application in this quarter, making it available to all 85,000 Intel employees. It also plans to work with other information technology companies and ask them to take the code and the courses, and begin to deploy them as well. To that end, the Peer-to-Peer Working Group has been formed for the industry to "get religion" about what was once the essence of communication among machines.
A peer-to-peer architecture that enables more-efficient computations and communications in this 24/7 world should be welcome.
It certainly would have helped the DATE conference editorial crew when we recently tried to update a conference site while at the venue, using server/client relationships between our laptops in Munich and our data centers in New York and San Diego. A peer-to-peer architecture would have definitely helped editors like me to have reliable connections with the home office at 9 a.m. Munich time. Then we would not have had to sit there in midafternoon playing the "disconnect" game. Take it from me, Germany is a hard place to be when you're relying on connections.
Nicolas Mokhoff is Editor, Special Issues, for EE Times based in Manhasset, N.Y.
RSS
RSS| 
Digital| 
Mobile
