Richard Newton is dean of the College of Engineering at the University of California-Berkeley. A native of Australia, Newton joined the faculty of UC-Berkeley in 1979 after receiving his Phd there in 1978. Newton is uniquely qualified to discuss the evolution of computing. As a researcher and teacher, he has advanced the areas of design technology, electronic system architecture, and integrated circuit design.
But he's not your run-of-the-mill academic. Newton also moonlights as an entrepreneur and venture capitalist. Since 1997, he has been a member of the Technical Advisory Board of Microsoft Research Laboratories. He has helped found a number of electronic design companies, including SDA Systems (now Cadence Design Systems), Crossbow, and Synopsys. From 1998 to 2002, he was a venture partner with Mayfield Fund, a respected Silicon Valley VC firm.
BusinessWeek Computer Editor Spencer E. Ante spoke with Newton from Munich, Germany, about the Next Big Thing in the technology industry: utility, or on-demand, computing. Following are edited excerpts of their conversation:
Q: So what do you make of computing on demand?
A: Some of our faculty members are working on e-business on demand. In the future, anything I want to do can be outsourced. There's no doubt that the long-term future of computing, from basic communications to storage to computing, will be provided and used largely as a collection of services. In fact, many companies are moving strongly in this direction already.
Q: What are the benefits of on-demand computing?
A: It creates very large economies of scale. And this is critical to maintaining the efficiency of corporations.
Q: What is the nirvana of on-demand computing? Where does it all lead?
A: Developers will build applications that completely transform our lives. I will be able to take e-services off the shelf and stitch them together. For example, I want an e-mail client that filters out spam and offers to pay you money for time. Wouldn't that be great?
In business it gets you out of the game of building the 747 in your house. You don't buy a wing and buy an engine and put it together. It turns all of that computing into a set of standards, which means I don't have to worry about all the possible configurations.
Q: How important a development is this?
A: If it can be done, it's as important as the development of the standard computer was in its day, back in Cambridge in late 1940s.
Q: As you know, computer researchers have been working on this problem for 40 years. For example, some of this has roots back in the 1960s with time-sharing of computing resources. How hard will it be to create the new on-demand technologies?
A: There are huge, huge technical challenges. But beyond that the main issues I see are standards and trust. Reliability, predictability, availability, and privacy. Those are the key challenges. As utility computing is deployed and we all become dependent on it as we are with electricity and clean water, the services provided must be reliable and predictable. Even seconds of downtime, or the loss of unauthorized sharing of information, will cause businesses and consumers to retrench and invest in proprietary off-line systems and services.
Q: Why are standards and trust so important?
A: If the industry is going to scale, we need some analogous standards, from infrastructure all the way up to services. We don't have standards in storage and privacy. We have them in TCP/IP [the basic network protocol of the Internet Age]. There are also protocols we have to standardize. For example, how do you implement micropayments? It is absolutely critical in these areas.
If we don't have standards, that's the biggest risk to fragmenting and limiting the reach of this technology. Trust is earned. Once you blow it, it takes a huge effort to regain it. It's like visiting a Web site. If you have a bad experience, you don't go back. If I was a company, I wouldn't go out and say I'm going to provide everything you need. Start small. And go from there.
Q: How long is this going to take?
A: It will take a while for the full ramifications for this model to be deployed. This is 10- to 15-year process.