InformationWeek
Defining The Business Value Of Technology
|
November 20, 2000 |
 |
New Life For Peer-To-Peer Computing
continued...page 2 of 2
 |
| More on peer-to-peer: |
|
|
 Send Us Your Feedback |
Most peer-to-peer applications allow two or more people to work together, but peer to peer can also be used as a means to keep systems well-maintained and fully up-to-date.
MyCIO.com, a subsidiary of utilities vendor Network Associates Inc., introduced its own peer-to-peer technology in September that lets IT managers maintain all of their computers in a simple and rather unorthodox manner.
MyCIO's Rumor distributes updated antivirus definitions for its McAfee VirusScan utility. Rumor "asks" computers on the network if they have an updated virus definition file, much in the same way a Napster user searches all of the computers logged into Napster's servers for an MP3 file. When it finds one computer with an updated version of the antivirus definitions, it downloads and installs the definitions automatically.
"The need for this technology came out of a very natural problem, distribution of large data files with signatures over LANs with slow connections," says Victor Kouznetsov, VP of engineering and chief technology officer for myCIO, in Santa Clara, Calif. "In a Groove world, you're publishing information. In our world, it's like a network of systems that update from each other and synchronize from each other."
By operating this way, any computer can act as a distribution point, and as the updated files permeate the network, other computers looking for the update can take the new files from any system. This eliminates the need for a single server as a distribution point, which can get bogged down under the weight of hundreds of requests from clients. "That's one of the pure peer-to-peer implementations you can find today," Kouznetsov says. "Even Napster is server-based."
Rumor is designed for McAfee's antivirus and PC firewall security programs, but it's extensible to support other applications. It can be used to distribute Microsoft Windows patches, application patches, new device drivers, and other such updates, Kouznetsov says. Rumor won't download any file that doesn't have a digital signature from Network Associates, in the case of antivirus definitions, or a digital signature from any other vendor.
The third use for peer-to-peer computing is CPU sharing, or cycle sharing. This notion has been given credibility thanks to the success of SETI@Home. Distributed computing is taking a massive processing job and slicing it up to a multitude of computers, each processing a small slice and sending back the results. Rather than having one big computer run a thousand jobs, the organizers get a thousand computers to run a single job and get the same result.
There are a number of efforts, all of which use a homegrown client to fetch, process, and return the data, and a server system to distribute the data and receive the results. For businesses and researchers who want to run their own distributed computing projects and put idle PCs to work on a major processing job, Mithral's Cosm software developers' kit lets programmers quickly create their own version of a SETI@Home program.
Mithral created the Cosm software developers' kit to quickly build peer-to-peer and distributed applications, especially the latter. The easiest application to build is something such as Folding@home, Beberg says, "because it's drop and load." Folding@Home simulates how genomes, the genetic material of an organism, form proteins; understanding how proteins are formed will let drug companies create medicines that fight diseases, since all diseases are made up of proteins. "Drop code in a couple of places and it's done," Beberg says. "Not much coding is required."
For assistant professor Vijay Pande, a researcher in Stanford University's chemistry department, the Cosm kit let him quickly build a screen-saver program to simulate research. Pande's group is examining how genomes fold into proteins, which are the biological building blocks for actions in an organism. They make up all functions in the human body, including antibodies, which recognize and combat disease. Likewise, viruses are made up of proteins. Proteins are made from genomes, which self-assemble in a process called folding. By learning how a genome self-assembles into a protein, researchers hope to find ways to combat diseases.
But genomes self-assemble extremely fast, often in as little as 10,000 nanoseconds. On a 400-MHz Pentium II, it would take a full day to simulate one nanosecond of folding. Multiply that by 10,000 and you get an idea of the scope.
That's why Pande's group wrote Folding@Home (www.stanford.edu/ group/pandegroup/Cosm), a screen-saver utility that simulates genomes folding into proteins. He signed up 10,000 users to run the program, which simulates protein folds until it finds a valid step in the self-assembly process.
Using the Cosm software developers' kit, Pande's group was able to write the screen saver in two weeks. All it needed was the folding simulation, validation code, and some OpenGL code to draw the animation on screen. The Cosm developers' kit had everything else, such as distributing new data and handling incoming results from users.
Just to show how easy the Cosm developers' kit is to use, Folding@ Home was written by a graduate student in biophysics who had little background in computer science and programming. "We have a pretty strong computer background here, but not in network computing," Pande says. "The libraries were very straightforward to build an application. It made our life a lot easier."
So far, Pande's group has been able to assemble smaller proteins and has already learned how smaller proteins self-assemble. From there, his group can extrapolate how big proteins can self-assemble. Pande wouldn't discuss specifics of the research, because that research is under peer review. But the work of the people running the Folding @Home screen saver has shaved months off running simulations on workstations at Stanford, he says.
This type of effort is better for researchers than for IT, says Alliance Consulting's Wollman. "Most of the computing resources [that large companies] need they already have," he says. "Research might need it, but I don't see it so much in corporate America. Most companies we deal with didn't have those options before, so they invested in the hardware. In the future they might use it, though."
return to page 1
Illustration by Claudia Newell
Photo of Kouznetsov by Shane Young
Back to This Week's Issue
Send Us Your Feedback
Top of the Page
