PLANETARY MUSINGS

Entries in "Computing"

Research computing - High performance?

One of the things that I am interested in, primarily because it helps me do my research better, is high-performance computing. I have only been here at Case a few years, but issues of research computing have been around here a while I gather... The primary issue is often: is there enough? It isn't always clear what enough means. What is enough for my research group might be overkill for another researcher, or not even scratch the surface of the needs of yet another.

High-performance may mean fast CPU's to me, but super-fast networks, incredibly high-speed access to enormous datasets, or screaming 3-D visualization tools to others. So what? Well, a lot of great research comes from being at the edge of what is possible... but computational resources are expensive - especially at the edge and primarily because they become obsolete so quickly. It is not like buying lab supplies that could work for a decade or more if properly maintained, etc - optimistically, 3 years and a computer is toast. So, is there potentially an economy of scale that could allow more researchers to be closer to the cutting-edge and really expand research opportunities?

Research units working together to share computational resources could be a good thing. But what would make it good for researchers with few (no) ties other than that they use computers throwing in their resources together?

I think there are three aspects to an answer to this question:

(1) Can we do better, more cutting-edge research?

(2) Can the workloads, related to computational infrastructure, of individual researchers or research groups be reduced?

(3) Is it financially sustainable?

All important questions, but the one that interests me most right now is (1). What research could you pursue (that you can't do now) if there were significantly more resources available to you on campus than a single department (or researcher) might garner by themselves? Nice to think far forward like that.

The ACRC is thinking about this question too... anyone else?