Back to January 2003 CRN Table of Contents

[Published originally in the January 2003 edition of Computing Research News, Vol. 15/No. 1, pp. 1, 9.]

By Greg Papadopoulos

In technology, success comes from equal parts innovation and pragmatism.

Jim Mitchell, the director of Sun Microsystems Laboratories, puts it this way: "This is not university research. We are here to make things that will someday make money."

Creating useful technologies--marketable technologies--is what keeps us going. We're idealistic enough to want to make a difference in peoples lives and practical enough to realize that selling useful products enables us to keep doing what we love.

Since Sun Labs was founded in 1990, a number of our rule-changing innovations have made it into the mainstream--from platform-independent Java technologies to our ground-breaking Sun Ray and Sun Cluster products. A very important piece of our business model is not just to predict, but also to help stimulate uses of the network that ultimately create more demand.

As the network expands, the fundamental questions that drive our thinking include: "What will the impact be on the infrastructure?" and "What pieces of the network infrastructure are missing?"

Secrets of the Supernet

Certainly one of the critical pieces is security--and one of the key concepts currently being developed at Sun Labs is a new method of communications tunneling. We call it Supernetworking. What it does is add a new layer of abstraction to a layered model of computer networking, making it easy to encrypt both the transmission and storage of data.

Communications tunneling is already used in today's virtual private networks, but mainly on a network-to-network basis. Other uses are possible, but remain costly and complicated. The beauty of Supernetworking lies in how easy it is to manage.

I won't go into the technical details; suffice it to say, the Supernet layer sits directly above the network layer and includes its own addressing structure and security services. This makes it possible to create multiple trust domains within any Supernet with ease. Supernets of any size can be created or disbanded with a few simple commands. Even individual participants can be added or removed without having to redo the whole setup--which fits nicely with the way most organizations work.

Ace in the Hole

Another of the pressure points we're addressing in the labs is a fundamental shift in software, from shrink-wrapped product to Net-based service. What that does to developers is radically alter the old ratio: It used to be one user, one machine, one software license. The new ratio is one instance of the application serving millions of users over the Net.

To help developers tackle that challenge, we're looking into new ways to create distributed application with a project called Ace. Ace is unique because it provides a natural way for developers to describe the "intent" of the application rather than manually writing the code that implements that intent.

Developers use it to create a high-level specification that provides enough information so that Ace can automatically generate the implementation code. In other words, Ace completely separates the implementation details of a distributed application from its specification.

The implications, for both developers and business executives, are huge:

• Ace reduces the level of expertise required of application programmers. They no longer need to learn technologies such as the Java 2 Platform, CORBA, .Net, database APIs, application servers, Web servers, or Web page programming.
• The Ace specification is so concise and straightforward that developers can complete applications ten times faster than with manual programming techniques.
• Applications can be designed, prototyped, and modified "on the fly."
• Management can easily understand and contribute to the application specification, ensuring that business goals and software capabilities are tightly aligned.

Beat the Clock

In the software example above, it is easy to see how underlying concepts tend to shift in technology, sometimes dramatically. Our thinking around hardware is shifting as well. In a radical departure from traditional microprocessor designs, in which a clock governs all activities, we've been experimenting with asynchronous designs.

The reason: today's individual logic transistors--the on-off switches that move data--can run as much as 20 times faster than the system clocks that regulate them.

Our proof-of-concept prototype, the FLEETzero chip, takes advantage of a sort of reversal of fortunes that has taken place over the years. Synchronous circuits take an operations-centric view that made sense in 1940s and '50s, when logic cost far more than communication. But now, with each circuit sporting many millions of transistors, the reverse is true. As the team noted in a recent paper: "The task of getting two numbers to an adder takes more chip area, consumes more energy, and takes longer than doing the addition."

Controlled by logic rather than a clock, the FLEETzero chip moves data faster than ever before--and reduces power consumption. (After all, it doesn't have a clock that keeps ticking whether needed or not.)

The Next Big Thing

What we hear consistently from customers is that system utilizations are poor, complexity and management costs are outrageous, and the 80/20 split in IT budgets (80 percent for service and support, 20 percent for hardware and software) is a source of real frustration.

That's where N1 comes in. Our most far-reaching project, N1 is a new architecture that treats the whole network as a single computer. N1 will provide the means to "virtualize" the elements of the network--the servers, the storage, even the cabling--so that it can all be managed more easily and cost effectively.

An open architecture, N1 is designed to create a single pool of resources that can be dynamically allocated to meet the needs of a whole list of services. Yousef Khalidi, the principal architect behind N1, describes it this way: "Whenever demand for a service goes up or down, the N1 architecture adjusts to it automatically."

This ability to match resources to services on the fly is one of the most exciting aspects of the N1 architecture. It makes it feasible to take distributed applications from concept to wide-scale deployment with new-found speed: You just carve out the resources you need and light up the application. It should be no more difficult than running an application on a single computer today; it's just that, in this case, the network is the computer.

It's the People

From James Gosling to Whitfield Diffie, Ivan Sutherland to Guy Steele, Robert Sproull to all the other great minds, it's the people of Sun Labs--in California, Massachusetts, and France--that produce not just big ideas but big results.

I think Director Jim Mitchell said it best: "Our researchers generate great ideas and innovative technologies, but it doesn't stop there. To ensure effective transplant, our researchers accompany the technologies to the product divisions where they work to develop strong, competitive products."

That said, I should add that we don't believe in going it alone. We make it a point to collaborate with other researchers--in a wide range of industries, with major universities, with standards bodies, and on open-source projects--because innovation can happen anywhere, and frequently does.

We really have only seen a single percent, at best, of what the network revolution will be.

Greg Papadopoulos is Senior Vice President and Chief Technology Officer of Sun Microsystems, Inc.