CAMBRIDGE, MASS. (06/26/2000) - Six companies last week pledged more than US$25 million to a Massachusetts Institute of Technology project aimed at reinventing computing for the Internet.
Project Oxygen was launched last fall to collect and advance an array of technologies for mobile devices and network-based applications. The goal, members say, is to make computers easier to use, less expensive and smarter.
Oxygen is a project by the MIT Laboratory for Computer (LCS) Science, with additional researchers drawn from MIT's Artificial Intelligence Laboratory.
Altogether, 250 researchers are involved in the five-year, $50 million project.
Joining the Oxygen Alliance last week were: Acer Group, a leading PC and components builder; Delta Electronics, a manufacturer of components and power supplies; Hewlett-Packard Co.; NTT Group (formerly Nippon Telegraph and Telephone, Japan's biggest telephone company); Nokia Corp., the Finnish-based cell phone and wireless net builder; and Philips Research, an arm of Philips Electronics, the Dutch consumer electronics giant.
The project's advocates envision a world where Oxygen provides the essential technology underpinnings to deliver to a given user the exact information or capabilities the user needs at a particular moment, in a particular location.
Mobile computing devices with voice and advanced visual user interfaces.
Stationary devices, such as computers, displays and sensors, those embedded in offices, homes, public spaces, vehicles and the like.
Secure nets that can support a range of communications protocols, and can form and dissolve as needed.
This will take more than improved speech processing and slick 3-D user interfaces. Oxygen's researchers will have to tackle new designs for servers and storage arrays, and bring systematic organization to the Internet.
"Oxygen is not about an ultimate [end-user] device," says Anant Agarwal, an MIT professor of computer science and engineering. "It's about creating a coherent system for deploying these devices."
Just how that will be done, he acknowledges, remains to be seen. Oxygen nets could have devices with microprocessors and software stacks to support multiple protocols. An alternative would be for Oxygen clients to run just one universal protocol to the servers, and for the servers to run several protocols as needed to interoperate. Or an Oxygen net, as the Web does today with HTTP, could have one protocol for the client devices to connect to servers which then communicate among themselves with different protocols.
Creating that coherent network will mean new computer and network designs, says Gary Herman, director of HP's Internet and mobile systems lab. Servers designed for online transaction processing workloads are not suited for the different workloads of the Web, he says.
Even client technologies, such as speech recognition, will have to be rethought for the kind of nets Oxygen requires. "Today, speech recognition is typically PC-based," Herman says. "How do you scale that for a hundred thousand users?
How do you make speech recognition efficient?"
As the number of network nodes or computers and appliances surges, there's a significant increased cost in systems and network administration and deployment, says Hari Balakrishnan, an assistant professor at LCS. Oxygen intends to create nodes that can automatically configure themselves with other nodes, without human intervention.
He also predicts network services will be much more highly distributed with Oxygen technologies, just as Web caching is becoming a distributed service spread over numerous servers.
Part of this trend will be increased intelligence in the network, requiring changes to the underlying TCP/IP protocol foundation. Oxygen is working with the Internet Engineering Task Force on some protocol proposals, Balakrishnan says.
Project Oxygen: www.oxygen.lcs.mit.edu.