Internet pioneer Vint Cerf on grid computing

Von Ian Foster

In my first Computerworld column I touched briefly on the Globus Consortium"s belief that open standard approaches (like the Globus Toolkit) are what will ultimately usher grid into mainstream enterprise use.

This month, I sought the wisdom of one of the pioneers of IP networking, and perhaps the most influential figure in open standards and interoperability discussions -- Vint Cerf. Currently senior vice president of technology strategy at MCI Inc., Cerf championed the TCP/IP protocol and played a vital role in steering the commercial growth of the Internet.

Here"s what this icon of networking had to say about the importance of open standards in the growth of the Internet.

Computerworld (CW): Talk about the role of standards in the creation of the Internet.

Vint Cerf (VC):The Internet created a layer of standardization that allowed a lot of different computers with different operating systems on different packet-switched nets to interoperate with each other. The wonderful thing about standards is that they allow different parties or systems to interoperate, even if they didn"t originally come to some agreement about how they would exchange information.

The grid notion has a lot of the same characteristics. There is a layer of standardization that is above the layer of the traditional Internet protocols, but like those protocols, it creates a kind of virtual commonality between all of the participating computing systems. And that"s very exciting, because in a peculiar way, this standardization creates a common clay out of which you can fashion almost anything.

In some ways, it has a characteristic not unlike electricity, where the electrons are pretty much application-insensitive, but will drive to any lights willing to accept them. The grid computing environment and the Web services environment creates this potential. I think we are far from fully realizing the full value of the grid concept.

CW: I believe that grid is still in the early stages of bridging the gap between research and enterprise. How do technologists differ politically from the research realm to the commercial realm?

In the hands of government research, the possibility of exploring something that on the face of it, might not yet have economic viability, is extremely important. The Internet certainly wasn"t an economically viable thing until routers were available and Cisco and other companies were building them. So the economics depended very heavily on investment from industry seeing an opportunity to build and sell something.

The promise of grid, being able to virtualize the computing, storage and communication resources -- and to turn all of the resources on the Internet into a sort of virtual, gigantic multiprocessor -- is still a very visionary view. But what is practical to do today, that I"m very excited about, is to take clusters of computers and turn them into virtual resources -- so that it no longer matters which specific resources are used to carry out a particular computation or provide a particular service. It means that business continuity has a hope of being implemented simply by moving from one virtualization source to another. So if you"re doing networkwide backup and things like that, the virtual resources look no different from one resource to another.

CW: What should the grid community be thinking about with the carriers and network providers?

VC: We are, of course, very interested in grid, and [MCI CEO] Michael Capellas is very articulate on this. MCI has two things to offer here. We obviously have the backbone capabilities and computing facilities to house computers that can be configured to behave in the ways the grid specifies. That means we can offer to our customers business continuity possibilities, as well as the possibility to expand the computational resources available.

Let"s suppose you"re a company that has a quarterly, significant increase in computing requirements, because you"re trying to complete your quarterly reports. And you don"t want to buy and own and operate computing resources to meet the peak requirements, because you only need them 10% of the time. If everyone uses grid-like standards to implement their computational requirements, MCI -- along with others -- could offer dedicated facilities for a temporary period of time. It"s the equivalent of time sharing, but you have to use hardware and software that"s conformant with grid standards. The good part is that by adhering to these standards, any company that needs the rapid expansion of computing or storage resource can get it.

CW: Do you see MCI starting to provide access to large computer farms?

VC: Are you getting significant demand from people to do that? We"re getting a lot of inquiries about what"s possible. Grid"s still in its early stages, as the Internet was. Not many people have reconfigured their IT environment to take advantage of grid.

The more we see people migrate towards this alternative way of implementing their IT environment, the more potential we see there for us. And the potential is there for us in multiple dimensions, not just because you have to use the network to move data back and forth. The real excitement is that once you create the standard grid interface, then you can start implementing applications that you can make available to customers through the grid.

CW: Talk a little bit about the importance of the standards bodies.

VC: You do need to have a critical mass of people who are all focused on the same thing. I hope the grid standards guys manage to overcome a certain degree of engineering debate that we experienced within the TCP/IP community. The grid guys really do need to jump out of the fray and ask themselves, "What is it collectively that we"re trying to accomplish?" What is the old saying in the American Revolution -- "If we don"t all hang together, we"ll all hang separately ..."

The only other thing I would say is that we need some really basic research into how to make best use of grid ideas. We need to figure out what algorithms will work best in the distributed environment and how we can quantify and measure that. If you"re trying to create a market, you want to know which applications will be the most effective. Ultimately, I hope we will see more cooperation between the people who are interested in this project, although there"s nothing necessarily wrong with competition. Sometimes it sharpens the results.

(Grid pioneer Ian Foster is a board member at the Globus Consortium, a vendor-neutral, nonprofit organization promoting the open-source Globus Toolkit in the enterprise. He can be reached at