eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
For half a century, breakthroughs in computing have come from the predictable sources of federally funded aerospace and defense. During the past three decades, though, theres been a dramatic tilt in federally funded research toward the life sciences. That means the dividends from bleeding-edge computing that filter back to enterprise IT will be coming from a different source, along with new viewpoints: much needed because the old ways of looking at new IT problems are clearly reaching their limits.
Since 1970, theres been essentially zero inflation-adjusted growth in research funds for engineering and physical sciences, even while constant-dollar life science funding has grown by more than a factor of 4. “Mechanical engineers are still doing mechanical engineering, but theyre looking at the mechanics of folding proteins,” said MIT Alumni
Association President James Lash at a breakfast meeting that I attended one recent Sunday morning.
One has to wonder if all those baby boomers, who want to feel like the Pepsi Generation forever, might be more interested in seeing their tax dollars build the Fountain of Youth than the first Mars colony. Ill send them a postcard from Olympus Mons.
Personally, Im used to the idea that IT price/performance is pushed through successive plateaus by tasks that are unique to different industries. I comprehend whats involved in modeling flows of oil and gas in underground rock formations, the way we did when I was at Exxon fine-tuning exploration strategies for the Beaufort Sea; I could get my head around the challenge of finding feasible solutions to the novel problems of building “Star Wars” space-based defenses when I moved to that world a few years later. The former type of problem has many direct analogues in all kinds of process industries; the latter type of problem is encountered every day in building everything from minuscule widgets to massive power plants. And all those large-scale projects involve budgeting, scheduling and logistics requirements that are shared by any enterprise thats too big to take just one car to an all-hands luncheon.
It may be hard, though, for application developers to see the connection between a Beowulf cluster thats processing gene sequences today and a distributed processing scheme for portfolio analysis tomorrow.
The vocabularies of the fields are so different, and the research methods of biology are so unlike the pipe fitting and metal bending of the oil patch or the airplane factory, that it will be a challenge to bridge the new gap.
The connections are already being made. Search the eWeek Web site for “bioinformatics,” and youll find references to articles on object databases, high-density server blades and parallel-computing initatives at Gate-way retail stores. Look at the roots of distributed computing ideas like JavaSpaces, and youll find researchers whose first inspiration was the need to find improved means of managing what Yales David Gelernter has called “the huge, amorphous and constantly expanding body of medical information.” Database designers are learning new respect for their limitations as they face the problems of counting genes, let alone the proteins that those genes hold the recipes for making. What they learn, they will, sooner or later, apply to enterprise needs.
One familiar pressure is the search for investment returns, and that search has a long history of learning new tricks in pursuit of that goal. Venture capitalists still have assets looking for gold at the rainbows end. A high-risk biotech venture may not come up with a cure for cancer, but it may develop intellectual property in the course of solving its IT problems that gives investors another way to cash out.
We can also expect that life science research will breed a more open-minded kind of engineer, one whos inclined to root out and understand the key behaviors embedded in a complex system, rather than to think that the system behaves only in the ways it was designed to function.
Software development on open network platforms, and distributed application management in relatively uncontrolled environments, requires that kind of thinking if were to see the kind of robust behavior that well need in future systems.