If you think peer-to-peer networking has come and gone, think again.
While peer-to-peer networking may have stumbled in the past, its implementation has now been developed and its infrastructure strengthened. Peer-to-peer computing today is about file sharing and distributed processing, combined with increased speed and availability - all while reducing bandwidth bottlenecks.
A peer-to-peer network exists when client PCs communicate directly with each other, rather than communicating with a server. These client PCs, or "peers," share their resources, including processing power or storage space, with other PCs on the network.
When most individuals think of peer-to-peer computing, software like Napster and Gnutella no doubt come to mind. While these MP3 file-sharing sites have encouraged the resurgence of this network-computing model, they are certainly not the only participants.
And while serious moral and legal implications surround the Napster controversy, there is no doubt that Napster has shown peer-to-peer computing can work on a global scale. If nothing else, Napster's 21 million users have proven that the theory behind peer-to-peer computing is valid.
Now take it a step further and consider the implications for distributed processing across a peer-to-peer network. While distributed processing has been around for some time, the infrastructure of the Intranet and a few radical implementations create mind-boggling applications. Take for example the Search for Extraterrestrial Intelligence at Home (SETI@Home) Project at the University of California at Berkeley.
The project's mission is to process radio transmissions received from the Arecibo radio telescope in Puerto Rico for signs of extraterrestrial life. To achieve this goal, volunteers download an application file that uses their individual computers to process small chunks of data. Every few days, the program downloads data from the telescope, and while the computers are not in use by the users, it processes the data for patterns. Upon completion of a package of data, the computers return the results of the analysis to the server.
With over two million registered volunteers downloading and processing data, the SETI@Home project has been considered the fastest computer in the world. In fact, the project has already performed the single largest cumulative computation to date.
Now imagine that enormous distributed processing power coded to run on a peer-to-peer computing model. The result - a massively scalable, distributed, redundant, virtual supercomputer.
Beyond the theory, several companies have begun to put the model into practice. New ventures such as Centrata propose to create a system that will allow users to "sell" a portion of their hard drive space and processing time. Companies in need of processing power or storage space won't necessarily need to own the CPUs or buy the redundant disk arrays. Instead, a company could buy storage space from Centrata, which it licenses from users of computers located anywhere on our global Internet. Data would be encrypted and stored on multiple machines to provide redundancy.
Distributed processing using peer-to-peer could be just as revolutionary. Like data storage, processing time could be purchased from Centrata. So whether a company is looking to offload database queries or perform intensely time-consuming algorithms, sharing the MIPS could be a revolutionary solution.
Other peer-to-peer start-ups like AppleSoup will offer file-sharing services for a fee. Unlike Napster's free model, AppleSoup recommends solutions that provide security and copyright features to facilitate secure transfer of business files.
But don't think that peer-to-peer innovations are limited to Internet start-ups or surround legally questionable activities. Industry giants, including Intel, are at the forefront of this revolution in computing. In fact, a track was dedicated to peer-to-peer computing at the Intel Developers Forum in late August. At the forum, Pat Gelsinger, vice president and chief technology officer of Intel's Architecture Group, declared Intel's support for peer-to-peer computing and announced the formation of an industry working group to promote it. Members of the group, which will hold its first meeting on Sept. 26 in Santa Clara, include Intel, IBM, Hewlett-Packard and several start-up companies involved in peer-to-peer computing, including Groove Networks and Entropia. (To learn more and view the Webcast, click here.)
The power of peer-to-peer computing may reinvent the very essence of computing on the Internet. Everything from data storage and processing to business transactions and information exchange across the Internet could be redefined. Peer-to-peer computing may transform our entire computing experience with levels of scalability, usability and availability unknown today.
RELATED LINKS
@currid.com. Intel backs peer-to-peer computing
IDG News Service, 08/24/00. Currid & Company's P2P computing resources Archive of Network World on The Technology Executive newsletters
