Inside Sun Labs: Exploring the Future of Computing

Sun Microsystems Laboratories (Sun Labs) is an engine of invention and exploration that has helped keep Sun at the forefront of network computing for more than 15 years. A cornerstone of Sun's multibillion-dollar R&D investment, Sun Labs has developed many of the technologies that have shaped Sun's product portfolio and reputation for thought leadership — including Java technology, UltraSPARC processors, Sun Cluster high-availability software, elliptic curve cryptography (ECC), Netra carrier-grade servers — the list goes on.

But Sun Labs focuses on exploring the future of computing, not looking back at past accomplishments. To wrap up the old year and ring in the next, Sun Inner Circle decided to peer into the future of computing by examining some of the current projects under way at Sun Labs. Sun Inner Circle invited Dr. Robert Sproull, who took over as head of the labs in March 2006, to give us some insight into what's happening at Sun Labs today — and a sneak preview of the new ideas and directions the labs may explore in 2007.

Dr. Sproull is quite simply one of the pre-eminent computer scientists in the world. His father was an early director of the U.S. government's Advanced Research Projects Agency (ARPA), which was instrumental in the development of the Internet, and his son proved to be "no slouch himself," in the words of former Sun Labs Director Jim Mitchell. Dr. Sproull has worked closely with Sun Fellow Ivan Sutherland — considered by many to be the creator of computer graphics — in asynchronous logic-based integrated circuit design, proximity interconnect, and other innovations that show considerable promise.

'This Is Not Ivory Tower Research'
Dr. Sproull was happy to comply with our request for a look inside the Sun Labs project portfolio, but wanted to make sure we understood the big picture first: the overarching purpose of Sun Labs and its fit within Sun as a whole.

"Sun Labs fills the key role of transforming fundamental research into tangible technologies, prototypes, and proofs-of-concept that can be transferred to product divisions," he said. "And what really sets Sun Labs apart as an applied research organization is our close connection with the rest of the company. We consider Sun's engineering organizations to be our primary customers. Those are the folks who take our prototypes and turn them into something real."

The ultimate goal for a Sun Labs research project is not a dissertation but a product, according to Dr. Sproull. "This is not ivory tower research," he said. "We're willing to take on big ideas and tough challenges, we're willing to take big risks, but success for us is technology transfer. We want to deliver technology that is relevant to our product groups so that they can create products that directly benefit Sun's customers."

Proximity I/O: Massive Bandwidth with Lower Power Consumption
One of the most intriguing and potentially important areas of focus for Sun Labs is called Proximity I/O, a method for chip-to-chip communication. Proximity I/O enables high-speed communication between neighboring chips without the need for wires of any kind. In fact, Proximity I/O has proven capable of increasing bandwidth by up to a hundred-fold while decreasing power consumption and improving reliability, manufacturability, and cost. This technology helps keep the performance of ever-faster chips from being curtailed by limitations of getting data on and off a chip.

Sun Labs has been investigating various aspects of Proximity I/O for several years with great success. The challenge that Sun Labs now faces is to figure out how to expand chip-to-chip communication to board-to-board and rack-to-rack signaling. In order to achieve this level of wireless communication, Proximity I/O must bridge to optical networking technologies that might link boards of chips with other boards of chips.

Using Proximity I/O, according to Dr. Sproull, it will one day be possible to create extremely high-speed communications among very large systems. Sun Labs intends to remain at the forefront of Proximity I/O research, and will continue enhancing the robustness of the technology so that it will soon be capable of solving the immense bandwidth challenges of large commercial and governmental customers.

Project DReaM: Taming DRM Chaos with an Open Solution
Almost everyone who owns a digital device has had a problem with digital rights management (DRM). You can't transfer the songs you purchased for some music players to a different brand player. You downloaded a movie legally, but you can't watch it a friend's house instead of your house. The battery dies on your music player, so you have a choice between spending $100 to replace the battery or starting over with your music collection.

Sun's approach to DRM, like Sun's approach to so many other technologies, has been to create an open, standards-based architecture and to share it — royalty free — with the development community so that the community can innovate and add value without technological encumbrances or prohibitive licensing costs.

Central to the DReaM solution are four key concepts:

• Approaching DRM from a network identity perspective. Rather than the traditional device-centric approach, the identity-enabled approach aims to connect the content with the individual, not the device.
• Providing an open, standards-based framework for interoperable DRM implementations. DReaM can work with virtually any type of content needed, any type of device, and any operating system. It also works with multiple file formats and codecs.
• Providing security without obscurity. Current DRM solutions are based on a false foundation of supposedly uncrackable cryptography, but DRM solutions have been cracked and will continue to be. DReaM promotes the view that open system architectures can be better evaluated and corrected ("review & repair" versus "hope & pray") to provide the greatest protection possible.
• Avoiding onerous licensing fees. Project DReaM is designed to be royalty free to encourage development and future innovation.

Sun Labs opened Project DReaM as part of its Open Media Commons initiative in August 2005. Since early 2006, Sun has taken steps to contribute DReaM technology to new and expanding standardization efforts around DRM.

The ultimate goal is to help unlock the full potential of rights management and use it not only as a security mechanism, but to expand access to content and open new revenue opportunities. For example, enterprises could use DReaM to control access to sensitive documents or authorize privileges such as entry to buildings or facilities.

Healthcare organizations could use DReaM to make it faster and easier for authorized people, such as emergency room workers, doctors, and nurses to see all necessary patient information (MRI scan videos, X-ray images, patient interviews, test results, treatment history) without compromising privacy. Network service providers could leverage DReaM to offer better service packages at lower rates to customers.

Used well, digital rights management can offer the public more rights than they have today.

Project Darkstar: Massively Multiplayer Online Games Made Simple
The $28 billion video game market is now the second largest media market in the world — beating out music, books, and every other form of entertainment except films. But for developers, writing online games that must scale to dozens, hundreds, or thousands of players — and that must accommodate a diverse assortment of client devices — is anything but fun and games. It is an expensive, time-consuming, risky, and frequently frustrating venture.

Sun Labs aims to change that. Project Darkstar, unveiled publicly at the 2006 Game Developer Conference in San Jose, California, harnesses Sun's expertise in development tools, Java technology, and massively scalable back-end infrastructure to simplify the process of developing games that can be deployed on a massive scale to players using virtually any client device.

In essence, Project Darkstar gives game developers access to powerful technologies that will do a lot of the heavy lifting for them. There's no need to build a game server from scratch and then figure out how to scale it for massive multiplayer scenarios. With Darkstar development tools and the Sun Game Server — a platform that enables companies to host their games and add server resources incrementally — game developers are free to focus on their game, not the infrastructure.

A freely downloadable SDK for writing client- and server-side code is available from Sun Labs at www.projectdarkstar.com.

Project Sun SPOT: Inspiring a New Generation of Consumer Devices
Sun has consistently predicted that the Internet will one day link more than computers; it will extend to virtually everything of value. This vision of an "Internet of Things" includes cars, bicycles, refrigerators, astronauts, toys, even trees — collecting and delivering diverse data to equally diverse devices. (Yes, trees — embedded with sensors that detect climate changes and animal movements).

The "Internet of Things" will generate a massive amount of data to process, manage, and store. Over the past 24 years Sun has built its business by creating the back-end infrastructure to handle vast volumes of data. And now, with Project Sun SPOT (Small Programmable Object Technology), Sun provides developers a Java platform for inspiration and innovation on the device side.

Project Sun SPOT began at Sun Labs in 2003 as an exploration of wireless transducer technologies. Since then, the project team has been investigating new ways to make small devices smarter, more secure, and more capable. For example, the "Squawk VM" (a small J2ME virtual machine) provides the ability to run wireless transducer applications "on the metal," saving overhead and improving performance; "SPOTworld" simplifies development by providing a single tool for programming, configuring, managing and monitoring Sun SPOT devices; and the use of elliptic curve cryptography (ECC) dramatically improves device security.

Most importantly, Sun Labs has built a platform to inspire people to build cool new types of devices and systems with Java technology in sensors. There are already some amazing applications built with Sun SPOT technology — using Sun SPOT devices to monitor package shipments in order to prevent criminal activity, using Sun SPOTs as sensors that can alert you if you've left the iron on at home, building Sun SPOTs into rockets so that they can monitor and graph the progress of a launch as it happens, and more. For a list of applications, see www.sunspotworld.com).

Sun Labs believes Sun SPOT technology will be important in a broad range of emerging application areas, particularly in education, industrial research, and government and military applications. By enabling whole new classes of devices to connect and share on the network, Sun SPOT technology brings the vision of the "Internet of Things" much closer to reality.

CAP: Taking the Guesswork out of Performance Prediction
Needless to say, it's difficult to predict the performance of a chip before it is built. It can be even harder to predict the performance of a system, because memory, I/O controllers, and networking must be factored in. And it doesn't stop there because customers typically care about application performance, not just system performance.

At Sun Labs, the Computer Analysis and Performance (CAP) project has developed technologies and methodologies for making predictions about chip, system, and application performance with more confidence than ever before. The CAP team is composed of collaborative groups of experts from within Sun and from partner organizations such as universities. It brings a quantitative approach to the exploration of new trends, technologies, and system trajectories (i.e. proximity communication, chip multiprocessing, and integrated hardware and software solutions).

Using methodologies such as workload characterization, trace simulation, RAS modeling, and integrated hardware/software performance modeling, the team finds new ways to optimize computer systems for the attributes customers value most: price/performance, quality of service, growth capacity, availability, reliability, and so on.

Recently, the CAP team predicted application performance on the UltraSPARC T1 processor within 5 percent of actual levels — long before the chip design was finalized. Moreover, the CAP predictions and models were ready in time to influence final chip design decisions.