I studied Physics and Computer Science (Informatics) at the University of Hamburg, Germany before completing the Ph.D. program at CMU's CS department in 1989. At the Applied Physics Department of the University of Hamburg, I worked on digital image processing systems in a research group focused on optical information processing, holography, and related areas. This led to work on multiprocessing systems for digital image processing, which was the basis for my Diplomarbeit in Informatics. For my Physics Diplom (roughly equivalent to a MS degree) I applied digital image processing methods to electron microscopic pictures of gold crystal formation on alkali-halide crystal surfaces.
As in Hamburg, my life as a graduate student at CMU's CS department was divided
between my main research interest on multiprocessor systems, their interconnect
structures, and work on a multitude of very diverse projects. This included
the Deep
Thought Chess project, indoor helicopters
,
VLSI and CAD tool development, computer arithmetic, statistical modeling of
application program behavior, and many little fun projects.
After a brief, but enlightening (and profitable) vacation on Wall Street as a "rocket scientist" I started a project at Sun Microsystems on a seamless interconnect system to link workstations so that they can be used and administered as a single, potentially distributed multiprocessor systems (S3.mp). While a lot of interesting technology was developed for this project (like integrated, high speed serial interconnect systems, better communication protocols, clock synchronization and distribution techniques), S3.mp ultimately never became a commercial product. I still think that the idea of building large MP systems out of cheap and cost-effective workstations and/or PC's is sound, but now I also appreciate the economic argument against this approach: why endanger the high-margin, profitable server market by allowing the easy aggregation of low-cost, low-marging and low-profit PC's.
In the fall of 1996, I joined Digital's (now Compaq's) Western Research Laboratory, where I worked on several scalable MP server projects. The most significant of these projects is Piranha, which integrates 8 simple Alpha microprocessors on one chip along with memory controllers, I/O and a scalable interconnect system. Compa WRL is also a great place to incubate less mainstream projects, like ONSIS.