Sun Microsystems Inc. co-founder John Gage holds the type of title chief researcher that allows him to explore advanced computing and networking concepts on a cross-corporate, cross-academic basis one day, while advocating social-policy goals the next. In the latter role, Gage plays positivist foil to the dark musings of fellow co-founder Bill Joy, Sun's chief scientist, who fears the implications of runaway technology. Gage comes across as a no-nonsense humanitarian, engaged in efforts for education, poverty relief and digital literacy in the developing world. He is gregarious enough to make friends wherever he goes, but sober enough to know he must be a realistic ambassador for his company, at a time when Sun is involved in the first survival struggle in its 20-year history. EE Times' Loring Wirbel sat down with Gage before the executive's speech at the National Association of Broadcasters (NAB) conference.
EE Times: What has captured your interest within Sun's R&D teams?
John Gage: There's a lot of interesting implications to bringing back [Sun co-founder] Andy Bechtolsheim to work on Opteron architectures [as chief architect and vice president for network systems]. Andy's not only responsible for the new Opteron-based server line, but he's been looking at multiprocessing Opteron architectures for video streaming, combining the servers with forty-eight 500-Gbyte drives in a single system, for multiple terabytes on the desktop. There's more here than meets the eye. Andy and Dick Sillman are working on a very clever streaming device and on a redundancy mechanism much better than RAID arrays. In theory, their work could give you permanent, perfect storage.
We've also got some interesting work coming on the processor side. Niagara [the 32-thread, eight-core Ultrasparc engine due for 2006 release in servers] represents the ultimate Ultrasparc gain realized by concepts such as multithreading. We think we will be able to show some significant advantages in overall heat dissipation in large systems, vs. server clusters based on something like [Intel's] Itanium.
Beyond that, we took a clean-slate approach to come up with Rock, which will abandon deep pipelines, out-of-order execution, anything else that makes a CPU subsystem needlessly complex. Rock is the design effort within Sun to go back to Sparc simplicity.
EET: Everyone at NAB this year was talking content management. How do you view the emerging content-description efforts?
Gage: We're very interested in some of the industry projects, like WGBH-Boston's metadata project, developing a standard way for describing what's held in a digital-media repository. I'm still waiting to see more efforts to tie this kind of asset management into the communications network and into what the hardware vendors are doing to define the metadata at the moment of image creation.