Wednesday, February 02, 2005

E. Brewer on IT for developing regions

Eric Brewer of UC Berkeley gave our department's invited colloquium talk today. (Well, yesterday.) Brewer's talks are always fascinating, and totally different in subject matter from a typical computer science talk. For example, you learn about the Grameen Bank, which has given out almost $4 billion in loans to Bangladesh's poor over the past couple of decades --- in roughly hundred-dollar increments --- and managed to turn a steady profit while lifting nearly half of its customers out of poverty.

Or you learn that if you take an off-the-shelf $20 wireless network card, completely re-hack the operating system's network stack to eliminate all the features designed for shared-channel communication (round-trips, exponential backoff, etc.), mount the antenna at the top of a big fixed pole, and point two of these devices at each other, you can get 5 Mbits/sec. transmission on the order of 25 kilometers (and probably 1 Mbit/sec at double that distance). This could enable relatively cheap network communication for remote areas where it's not economically feasible to lay good wires. Which is a big deal in India, since apparently most of the nation's population, even in rural regions, lives within 50-100 km of a fiber optic line. You might be able to facilitate things like efficient communication of market prices for farm goods, or give rural people access to medical diagnostic services via remote tele-medicine. (These are not pie-in-the-sky applications --- they answer needs felt by actual Indian rural communities, and Brewer's group and others are building systems that will really do these things. They aren't charity either --- Brewer places a great deal of emphasis on "sustainability", i.e. whether the service can support itself economicaly after the initial capital infusion.)

Or you learn that if you plug First World computing hardware into a developing world power grid, your hardware will go kaput in relatively short order, because the grid generally doesn't put out clean, reliable 220V AC (or 120V, or any fixed voltage, actually).

Of course, it's humbling to contrast the urgency and potential impact of the work Brewer advocates doing --- millions of lives saved or lifted out of poverty --- versus, say, the formal semantics of programming languages, which is what I'm working on right now. (Considering this issue in detail, obviously, leads one down a deep hole of depressing moral implications, so I'm not going to write more about this at the moment, though this is not to say that I'm not thinking about it.)

The colloquium talk has not (yet?) been archived, but Brewer also gave an extended hour-plus version of this talk last quarter via tele-lecture, for Ed Lazowska and Steve Maurer's course on IT and public policy, and so it's been preserved for posterity in the online lecture archive (at the bottom, 12/09/04). You will, alas, need to install Microsoft's experimental WebViewer software in order to view the simultaneous slides/lectures, though you could also download the video and HTML, and page the slides yourself while watching the video. If you're in a hurry, Brewer also has a 25 min. talk video on the TIER project (the umbrella for his group's efforts w.r.t. developing world technologies) on his home page.

No comments:

Post a Comment