PlanetLab has been described as a microcosm of the new Internet. Hosted by Princeton University, it boasts the collaborative efforts of more than 70 organizations, including Hewlett-Packard, Intel, MIT, Stanford University, and Cambridge University. Unlike previous "new" networks, it exists as linked nodes overlaying the current Internet backbone.
The Internet, created primarily by and for researchers as a tool and a technological experiment, was rendered useless for researchers once the rest of the world caught on. Researchers responded by launching next-generation projects such as Internet2, NGI, and others, producing new technologies and commercial apps that, by and large, Joe Consumer could enjoy on the so-called public Internet. But as those first next-generation projects became popular (with scientists, universities, and K-12 classroom users), they, too, became crowded.
According to Larry Peterson, a professor of computer science at Princeton University, "Internet2 is primarily about providing capacity for high-bandwidth applications…but that has not really been fulfilled." Peterson, a founder of PlanetLab and its director, said the new network is complementary to Internet2.
Pardon the Disruption
Rick McGeer, HP CITRIS scientific liaison, said PlanetLab is an environment "designed for disruptive technology." Disruptive is good. Disruptive technology paves the way for truly new, different, and eventually stable technology. But, it is disruptive. It wouldn’t be possible in the general Internet community.
"A lot of what we do triggers intrusion detector alarms," Peterson said. "We’ve gotten calls [from ISP network administrators and IT organization security managers] because they didn’t expect a certain behavior." So far, each time PlanetLab has explained the aberrant network behavior to the "alarmed" organization, the administrators have agreed that PlanetLab should continue the operation.
Triggering alarms isn’t a goal in itself, but Peterson noted that the behavior that triggers an alarm also presents a challenge that is a good learning opportunity for researchers.
About 100 projects and experiments are currently running on PlanetLab. Truly killer apps include Internet "weather reports" that spot DNS attacks and allow watchers to reroute traffic around the Internet’s default pattern, experiments with reliable distributed video, and other projects that Peterson said involve not just communication, but also processing and storage. PlanetLab also provides a platform for experimenting with grid services. Of these, Peterson said, there isn’t an obvious "killer app." However, "running different services all over the world—the management of that is the killer app."
To reach their initial goal of building a 1,000-node network, PlanetLab founders approached network administrators with a bold pitch. He expected many administrators to deny the request, but instead he found most readily accepted the idea. "This is something the network and systems community has wanted. It is a system that gives the system-building community room to breathe and play," he said.
Peterson is looking forward to a more widely deployed lab, with "more people taking advantage" of PlanetLab’s resources. But, if more people take advantage of it, won’t the researchers risk being run off this new playground they’ve built?
Peterson admitted that’s a possibility. "But I’m not worried about it," he said. "If we run into that, it means we’ve done the right thing."
advertisement
