Supercomputing in the Public Interest
Jack 'daWabbit' Imsdahl
jack.imsdahl@oncomputers.info

A while back, I mentioned to Deepak (and shortly thereafter spoke of it on the show) that I knew someone who needed some serious computing power. The man in question was at the time an instructor at the University of Texas at Arlington and someone I knew from my Linux User's Group. He wanted to model automotive accidents in order to develop better child safety seats but was unable to get the necessary computing power. The University's computers simply weren't available to him on the necessary scale, due to demand from users with higher priority in the pecking order.

Deepak mentioned this man's dilemma to someone at AMD and they graciously offered to help. My friend had solved his own problems by that time, though. He moved to another institution of higher learning with a guarantee of access to the necessary computing power and a chance at tenure, to boot. Problem solved.

I've started to wonder how many such projects in the public interest are starving for computing power, though. It's not like they all need access to a supercomputer, either. My friend certainly didn't. Such modeling as he wished to do and at the level he needed to do it is commonly done (by auto makers, etc.) on small clusters of multi-processor workstations. Even with these [comparatively] modest requirements, he could not satisfy his need at the university. They were simply booked solid and he didn't command enough authority to requisition access.

Of course; in many instances, the problem is software availability first, computing power second. Without the right software, you can't get far, even if you have a few Crays out in your garage. It's easy to waste resources or even corrupt the results if the software isn't kosher. In my day, I have made very, very simple computer models of a few things (mostly to save myself hours on the hand held calculator, written in QBasic, which will tell you the low level of work I'm talking about here). I can say with authority that such software, even on my trivial level, is not particularly easy to write and test. I have no idea if my friend had the software he needed, but he seemed to feel he did. I don't recall him saying whether he'd written it, had it written or obtained it from some other source.

What is needed is some philanthropically inclined individual or group who can and will set up such a computer, or cluster of computers, and the necessary ancillary facilities containing skilled programmers and technicians. I'll bet potential projects would come out of the woodwork once the existence of such a facility is known. And I'll wager some hardware company (like AMD, who has already demonstrated their willingness to help on worthy projects such as my friend's) will pitch in with parts at cost or less. {job troll} I'd be willing to work for such a project cheaply, if need be. {/job troll}

I envision the deal like this;
The institution should be allied with some University or college so computer science students can have access to it for learning purposes and to help those needing the machines to get or write good software and run their projects well. Even a community college would do for this, provided they have faculty and students of a level sufficient to benefit from and contribute to the project.

The machine itself (and it's attendant facility) should be located in some economically disadvantaged area so that at least the jobs for maintenance and facility support go where people really need them. Such a location offers other opportunities to serve, too. For instance; One could add a teaching facility to give those among us less fortunate than we are a place to obtain at least basic computer skills, thus allowing them to get better jobs and so live better lives. You already have the necessary people and with each one doing a small amount of the teaching as a requirement of their being there, a real difference can be made in people's lives. As any such facility will need a rather fat pipe to the Internet, the small amount of bandwidth needed to teach folks how to use that medium for their betterment can be bled off without compromising or delaying anything.

The question arises of which projects are worthy? That's actually easier to answer than it sounds. To be considered, a project has to benefit a defineable segment of the public, large or small, such as children injured in auto accidents or sufferers of some disease. Perhaps things like automotive traffic analysis for local and regional governments which might not otherwise be able to afford it could be included. The results of the research and the source code for the software used to achieve them are to be made public on web sites hosted and administered by the facility. (The software requirement does not of course apply if a proprietary program is used. Generally, though, software for research is written for the purpose and not likely to be found on sale anywhere.) The results of any research done at our hypothetical facility may indeed be patented or otherwise protected as intellectual property. However; they must be made available to others for licensing under “reasonable and non-disciminatory” terms (usually abbreviated as “RAND”). These licensing and royalty terms must be agreed to by the governing board in concert with the researcher and the researcher must be contractually bound to them before any project is accepted so there'll be no surprises later.

It's not all that hard to weed out the anti-gravity proponents and those wishing to reconstruct ancient Atlantis and so on. The governing board would not have to put too much of their time into sorting those out.

This is not an easy or quick project to execute. It would require a good deal of money, the hiring of dedicated, skilled and savvy staff (programmers, technicians and managers) property to house the facility, building the machine and getting it up to speed. Still, it seems to me to be well worth someone's public-spirited investment.

Should you read this and know of someone who might need or use the services of such a machine and institution, please put them in contact with me at the address above. I'd like to gather a bit of ammunition with which to proposition those in a position to make something like this happen. I think it is needed and I'd like to have some examples with which to prove it.

Jack