February 19, 2009

A Stimulus Package for HPC

Chris Willard

A long-standing concern in HPC is the graying of the industry. That is, new people are not coming in at the skilled positions as older ones retire. (In the sports world this is how championship teams become also-rans in the matter of a few years). Although this issue has been noted and discussed with growing alarm over the decade, few solutions have been proffered. In this regard, I have for the last six to eight months been conducting an informal, unscientific, on-again-off-again (wibbly-wobbly, timey-wimy) survey* of end users as to what can be done to help secure the future of HPC. Responses reveal a great deal of concern and real passion about the future of the field, and suggestions invariably center on moving HPC into the educational mainstream. These recommendations include:

  1. Implement HPC specific programs at the undergraduate level. Programs could go as far as offering degrees in High Performance Computing.


  2. Include HPC methods as part of the standard curriculum in engineering and science programs. Engineers should not have to wait until their first job before they are introduced to HPC. This may mean that someone has to teach some fairly dull programming languages again (if you can’t program it in Fortran, it’s not worth coding – very ancient cybernetic proverb). It may also mean that ISVs need to work with academia to make their software available at reasonable prices (there maybe regulatory issues here as well that need to be addressed).


  3. Increase and intensify the teaching of parallel programming and methods. Simply put: there are not enough people with parallel programming skills in the world today. Someone has to teach them.


  4. Provide funding for these programs (fund it and they will come). This is where government comes in — it is the only social institution that can cause quick, broad-based implementations of the above programs and maintain support until the programs are able to demonstrate their long-term value (i.e., students from engineering schools with strong HPC focus are sought after in the market).


  5. Bribe students. Unfortunately HPC is not currently a high-status or cool career choice among those entering college or looking to go on to graduate school. To bring more people into the field, significant financial incentives need to be offered in terms of financial assistance either through loans or preferably outright grants for tuition, etc. I believe that this type of offer should be made to science and engineering students in general.


  6. Continue support for HPC centers. If we are asking people to make a “bet your career decision” on HPC, then we need to demonstrate a long-term commitment to the technology. It is heartening to note that such funding is appearing in the US Senate stimulus bill. 

Certainly most if not all of these ideas are being implemented at various levels across the world today, and there are people actively trying to advance them (for example, new directions in computational science education, the University of Canterbury, and the Ohio Supercomputing Center). However, it was clear from the concern and passion with which respondents presented their ideas that not enough is being done. Perhaps this is the time for the community to make a concerted effort to promote, support and accelerate comprehensive programs for bringing HPC into the mainstream of education.

*To participate in the poll, please either add a comment, or send email to chris@taborresearch.com. Respondents will receive my sincere thanks.