Monday 29 August 2011

Supercomputers and other large science facilities

In my recent HPCwire feature, I wrote that I occasionally say, glibly and deliberately provocatively, that if the scientific community can justify (to funders and to the public) billions of dollars, large power consumptions, lots of staff etc for domain specific major scientific intrusments like LHC, Hubble, NIF, etc, then how come we can’t make a case for a facility needing comparable resources but can do wonders for a whole range of science problems and industrial applications?

There is a partial answer to that ...

The initial answer is that there is only one LHC, NIF, Hubble, etc. - and they have lifetimes of decades. But for we want/need several leading edge supercomputers at any one time - and we want/need new ones or major upgrades every year. So the other science facilities can justify bigger investment because there are fewer of those investments.

However, that argument breaks down on examination. Yes, there is only one NIF. But there is also HiPER and LMJ. Likewise ITER is the latest in a long line of fusion experimental facilities, Hubble is one of several major optical telescope projects, and so on.

And these facilities do get major upgrades as new resolutions etc. are required by the science.

So the comparison is not really "rare vs. common" (big experimental facilities to top end supercomputers) but more like "some vs. more".

Now, remember that each of those major science facilities are supporting a single area of science, but each supercomputer can support a whole range of science areas.

A similar discussion can be had in commercial & engineering applications of high performance computing - supercomputers compared with other testing/development facilities.
I think the problem is that we still think of supercomputer as computer installations (always a hard sell for companies or governments or the public). But I have said several times, that the most powerful supercomputers are not just computers - they are major scientific/research instruments that are built using computer technology. There is a difference.

No comments: