from PRINCETON UNIVERSITY
Evolving the Internet: Experimental network allows
testing of new global services
PRINCETON, N.J. -- Princeton University is helping to develop an experimental global network of computers that is expected to become a testing ground for a future generation of the Internet.
The network, called PlanetLab, is designed to allow researchers to develop and test powerful new types of software that are not confined to a single computer but run on many computers at once, treating the global network, in a sense, as one large, widely distributed computer.
Once fully developed, such a network could yield many benefits, from faster downloads to more powerful search engines. A person watching an online video, for example, might receive it from many computers that work together to avoid congested parts of the Internet. Software that scans the entire Internet for malicious behavior could catch problems before they could be detected by a single computer at one particular site.
"If I can observe the behavior of the Internet from multiple vantage points, I can see what the traffic looks like, where the losses are and where the congestion is," said Larry Peterson, chair of Princeton's computer science department and one of the founders of PlanetLab.
In its initial stages, PlanetLab is expected to be most useful as a vehicle -- open to any researcher in industry or academia -- for testing globally distributed applications. Previously, researchers relied on network simulation, or a cluster of computers in a single room or building. With PlanetLab, which is built onto the Internet as an "overlay" network, researchers have a real-life testing ground that would be impossible for any single institution or company to create, said Peterson. "The Internet is much flakier than a controlled machine room," he said. "And it's just not practical for any one researcher to have hundreds of machines spread all over the world."
The PlanetLab project began in March 2002 when researchers from several institutions met to discuss the idea. The founding group included David Culler of Intel and the University of California-Berkeley, Tom Anderson of the University of Washington and Peterson. Intel, the world's largest computer chip maker and a manufacturer of networking products, provided seed grants of equipment to set up an initial network of 100 computers, which are hosted by 60 international institutions that quickly joined the project. The technology firms HP and Google also are joining the project and will commit resources to it.
Princeton University recently has committed to hosting a formal consortium of PlanetLab users and developers. The University will provide administrative and technical support as the system grows toward a goal of 1,000 nodes worldwide.
"PlanetLab is unlike any other collaborative research effort I've been involved in," said Peterson. "It has an energy much like existed in the earliest days of the Internet. Researchers are using PlanetLab as a platform for advancing their individual research agendas, but at the same time, they are looking for opportunities to contribute to PlanetLab's core infrastructure. If I had to put my finger on the key idea in the architecture, it's that we have designed PlanetLab to provide a level playing field for innovation, and the research community has responded in force."
"Intel is excited to be working with Princeton University and a global collaboration of corporate and academic researchers to develop novel planetary-scale Internet services," said David Tennenhouse, vice president of the Corporate Technology Group and director of research at Intel. "PlanetLab researchers are going to unlock a new era of innovation on the Internet, bringing new services within milliseconds of their users around the world, and affecting the design of smart servers, embedded storage and network processors. We are simultaneously creating the engines of future Internet growth and learning about the new products that will be required to make them tick."
"HP Labs is extremely excited about joining PlanetLab," said Patrick Scaglia, vice president and director of the Internet and Computing Platforms Research Center at HP Labs. "It's a perfect example of collaborative research. Neither we, nor Intel, nor Berkeley, nor Princeton, could create alone the next generation Internet that will be born in PlanetLab."
PlanetLab researchers believe the project will help the Internet evolve beyond a simple structure for transmitting data from place to place and toward a system that can manipulate the data before and after it travels. It is a first step toward an Internet that has processing power built into the infrastructure of the network. Because it is so decentralized, the Internet cannot change overnight. But if PlanetLab demonstrates value, companies that sell Internet hardware might begin to work aspects of the system into their products, said Peterson.
PlanetLab is known as an overlay system because it builds on the basic infrastructure of the Internet. The Internet itself began as an overlay system on top of the phone system and still depends in large part on wires and cables that were first designed for telephone communications.
"PlanetLab is plainly an idea whose time has come," said Vinton Cerf, one of the founders of the Internet and now senior vice president for architecture and technology at MCI. "While it is too early to tell what is likely to be discovered, it seems inescapable that this platform will furnish opportunities for innovative experiments."
Peterson said PlanetLab already has become a valuable teaching tool for both graduate and undergraduate students who study networks. Courses at several universities have been built around the project, and Peterson plans to teach one this fall. Other Princeton researchers involved in the project are assistant professors Vivek Pai and Randy Wang.
"I'm very excited by the PlanetLab initiative," said Maria Klawe, dean of Princeton School of Engineering and Applied Science. "It is a unique collaboration between academia and industry that has already produced a powerful tool for both teaching and research. I am particularly proud of Princeton's role and Larry Peterson's team, which has been responsible for defining the architecture and building the key pieces of technology. Princeton is now stepping forward to create an academia-industry consortium to take PlanetLab to a significantly larger scale with enhanced core tools and technologies. This exemplifies the kind of leadership that our School of Engineering and Applied Science aspires to achieve."