R&D Gems

FRAMINGHAM (01/31/2000) - You can almost hear the paradigms shifting way up in those ivory towers. At the University of Virginia, they're inventing a "worldwide virtual computer." At the University of California, it's a "planet-scale, self-organizing" system. And at Carnegie Mellon University, they call it an "invisible halo of computing."

While researchers at each of these universities are pursuing their visions in very different ways, at a fundamental level, they all are dreaming the same dream for the 21st century. They say that computers will disappear yet be everywhere, that virtually every person and thing will have digital connections to every other person and thing and that the pain and risks of computer use will greatly diminish. They say the impact on computer managers and users will be profound.

The vision stretches far into the future by information technology standards - 10 years at the University of California at Berkeley - but some capabilities are scheduled for prototyping in the next year or so. And the University of Virginia has already found real-world users for Legion, its virtual computer.

"This research is moving us in the right direction," says Leonard Kleinrock, a computer scientist at the University of California at Los Angeles (UCLA) and one of the fathers of the Internet. "We are going to have distributed intelligence, distributed knowledge. Internet services will be everywhere, always available, always on, but most of all, invisible, just like electricity is."

"We need vast amounts of computer power, and there are problems we won't even touch unless we know the computer power is there," says Michael Crowley, a scientist at The Scripps Research Institute in La Jolla, Calif. For example, a job that models protein-folding can run for 12 hours on a supercomputer, so Scripps asks Legion to roam the country sniffing out spare computer cycles.

"We just say, ‘Legion, run it,' and it finds machines that are open, finds the correct executable, gets all the input files over there, runs the job and brings the output back," Crowley says.

Legion (http://legion.virginia.edu) is a highly flexible, wide-area operating system designed to build a virtual computer from millions of distributed hosts and trillions of objects - while presenting the image of a single computer to the user.

Originally developed for U.S. government scientists, it is now finding use in private labs and will eventually move to mainstream commercial use, says Legion architect Andrew Grimshaw, director of the Institute for Parallel Computation at the University of Virginia in Charlottesville.

Legion applies many of the object-based interoperability principles in the Common Object Request Broker Architecture. It finds and schedules resources and handles security issues among disparate operating systems and objects written in different languages. That frees users from the need to negotiate with outside systems and administrators.

Grimshaw says Legion is more robust than today's commercial systems because it's based on a "good, clean conceptual model" that considered basic services such as security from the outset, not as an afterthought.

"The concept of Legion is something I've been calling for for years," says Kenneth Neves, director of computer science research at The Boeing Co. in Seattle, which is evaluating Legion. "If you're a company like Boeing and you're going to invest millions of dollars to develop something like a new fluid dynamics program, you'd like to have some consistency in the model of the program, and we are not getting that from the individual vendors."

Legion gets its consistency from interoperability standards that mask the incompatibilities among lower-level components.

Neves says Legion possibly could run Boeing's highly complex and distributed product data management and manufacturing resource control systems. "The technology came out of the scientific computing world, but it will probably have its greatest impact in other areas," he says.

For example, Neves says, Legion could prove ideal for the large, distributed transaction processing applications employed by big financial institutions.

"With its virtual file system, one could write unlimited-size database applications."

In its Endeavour project, the University of California at Berkeley (UC Berkeley) has a vision that's broader than Legion's, but the projects share some goals - principally that computing resources ought to be invisible, easily managed and invoked, and infinitely scalable.

Endeavour (http://endeavour.cs.berkeley.edu) envisions an "information utility" that adapts on the fly to the supply and demand for computing resources anywhere in the world. "Fluid software" will choose where to execute and where to access storage. It will requisition resources as needed and negotiate agreements to ensure "always-available" processing.

Endeavour is really a collection of projects. One, called OceanStore, will provide "nomadic data access" - serverless, homeless and freely flowing through the network - that will make the storage available to a user essentially infinite.

For a small fee, users will subscribe to a storage utility that will let them store data "in the net." The utility will be provided by multiple, independent service providers, which will be responsible for providing backup, encryption and disaster recovery.

That will lead many companies to outsource data storage, says OceanStore project manager John Kubiatowicz. "You'll have the ability to travel anywhere in the world, and your stuff is just there," he says. The first prototype is due to be unveiled next year, he says.

A second Endeavour project is inventing microelectro-mechanical systems (MEMS), tiny devices that can combine sensors, communications devices, processors and power supplies into packages the size of this letter o. MEMS could revolutionize supply-chain management, for example. An item such as a refrigerator with an embedded MEMS device could generate a lifetime stream of information, transmitted to the manufacturer or owner. With a MEMS chip inserted early in the manufacturing process, the refrigerator could direct its own factory routing, machine controls, warehousing and maintenance.

In another Endeavour activity, researchers are learning how to capture an organization's tacit, or implicit, knowledge. "We are at the point now where we can embed intelligence in our systems so they can observe how they are being used and adapt themselves to that use," says Randy Katz, Endeavour project leader. For example, an application might infer and report indirect relationships among people because they behave similarly. The application might say, "Your team seems to be organized like this other team, and you are creating documents with similar sounding names. Do you know that this other team exists?" UC Berkeley is working with IBM to put these kinds of capabilities into Lotus Notes, Katz says.

UCLA's Kleinrock agrees that such capabilities are coming but says they raise privacy concerns, because people say they "don't want to be tracked all the time."

Carnegie Mellon University's Aura project (www. cs.cmu.edu/aura) has many of the same goals as Endeavour, but it focuses especially on the human-computer interface. Its developers say the most precious resource in a computer system is no longer processor speed, memory, disk storage or bandwidth - it's user attention. Systems today constantly distract users, making them less effective.

"Computers are consuming much more of our time than in the past, especially as we move to heterogeneous systems and networks," says Daniel Siewiorek, director of the Human-Computer Interaction Institute at Carnegie Mellon in Pittsburgh.

"The distractions are due to incompatibilities in data representations, applications and platforms."

Aura's goal is to give users an "invisible halo of computing" that persists regardless of location. "Systems should adjust to the context you are in," Siewiorek says. For example, a budget spreadsheet might display individuals' salaries in the privacy of a user's office but automatically suppress them when the spreadsheet is displayed on an airplane.

The halo will make for "smart offices," Siewiorek says. "As you enter the room, your [body-worn computer] interacts with the room's aura, and it knows who you are and your preferences. So I walk into a colleague's office to show him a slide presentation, and by the time I walk in, I should not only be authenticated, but the presentation I had open on my desk should be open on the computer there. And if I'm a Mac user coming into a PC office, that system should know that and do what needs to be done."

The idea is to allow users to just specify tasks - "make a budget," for example - rather than the underlying application programs, files and information formats.

Kleinrock says Aura gets at the problem of "feature shock," which overwhelms end users. He says much work has been done to develop application programming interfaces at the IP layer. "Now, there's another place to focus - at the top, at the interface with the user. We need a more common API to the human."

Join the newsletter!


Sign up to gain exclusive access to email subscriptions, event invitations, competitions, giveaways, and much more.

Membership is free, and your security and privacy remain protected. View our privacy policy before signing up.

Error: Please check your email address.

More about Boeing AustraliaCarnegie Mellon University AustraliaFinancial InstitutionsIBM AustraliaMellonUCLA

Show Comments