Philharmonic Computing

Improved microprocessors make other technologies more valuable. Mass storage, for example, is worth more if less expensive, faster processors can execute more demanding compression algorithms. Memory is worth more if high-bandwidth processors can sharpen and color-correct a video stream, to mention just one application, as easily as they used to optimize a single image frame.

But the real explosion of processor demand doesnt come from high-end applications for chips that cost hundreds of dollars. It comes from the hordes of simple applications, formerly less expensive to do with hard-wired logic, migrating to software on newly affordable low-end general-purpose chips.

If only all those general-purpose processors could find jobs to do in their spare time. Suns Project Jxta, a hot topic at this years JavaOne conference, proposes a framework for spontaneous collaboration among processing nodes—and a peer-to-peer framework like Jxta doesnt merely partition existing large tasks. It also enables exploitation of previously invisible opportunities: for example, as suggested during Bill Joys JavaOne demo, to have all the cars near a given gas station negotiate as a group for a discounted price.

Cooperation among Jxta nodes depends, to some degree, on agreement as to independent facts. Those facts change from moment to moment, facing system designers with a familiar dilemma: Continually retrieving the same information is wasteful, but assuming that information does not change is dangerous.

Solving this "cache coherency" problem for P2P participants that come and go and that communicate by low-speed shared or wireless connections is the leap that will unite electronic soloists into a symphony of dynamic and powerful solutions.