Why big iron matters

In one of those surprise moves that shouldn't really be a surprise any more, Saudi university KAUST yesterday announced plans that will catapult it into the highest levels of supercomputing research, and ownership, in the world

Tags: InfrastructureInternational Business Machines Corporation
  • E-Mail
By  Mark Sutton Published  September 23, 2008

In one of those surprise moves that shouldn't really be a surprise any more, Saudi university KAUST yesterday announced plans that will catapult it into the highest levels of supercomputing research, and ownership, in the world.

The new university, which isn't even due to accept students until September next year, is working with IBM on a system based on the Bluegene/P solution which will match the top ten supercomputers for processing power.

For my money, it's a great move for KAUST and for Saudi, and could have a significant implications for IT R&D in the region.

It's not that there aren't supercomputers in the Middle East already - Saudi Aramco has several, but commercial computing resources are generally treated as an expensive tool with a job to do, and aren't shared among students and researchers, so they don't add greatly to the sum of local expertise and knowledge.

It's also not the only academic supercomputer in the region - the UAE University has a grid set up, arguably a more innovative and experimental deployment than an IBM solution, but at 8 teraflops versus 222 teraflops, its not in the same ballpark as what KAUST has planned.

Even making the Top500 lists isn't even that important - being in the top five or top ten worldwide will certainly put Saudi on the map, but this isn't some superlative Dubai announcement about the biggest, fastest, most modern etc. The university is being quite modest about where the system will actually place in the Top500 listings, and there's some debate over the usefulness of the current Linpack benchmarks anyway.

Where I think KAUST will really count is in two areas.

First is the fact that their set up isn't just the central computer, they've bought the whole farm - storage, network, plus a state of the art data centre to put it in. By working with IBM on the application layer as well, KAUST should have a powerful tool for scientific research.

With a target of just 2,500 students at the university, that's a hell of a lot of state-of-the-art computing power to go around. By offering access to these sorts of tools, if it's coupled with the right academic environment, the university could attract scientists and researchers across many different disciplines and from many different countries, with the benefits that would mean in developing a centre of excellence.

The second reason is the partnership with a leader in HPC like IBM, and the university's plans and commitment to HPC research. Just by having the hardware, KAUST is likely to encourage and attract students and researchers in the field, but by creating a centre with the tools to research HPC and international links to other institutions they can really play a major part in developing the science.

Not only that, but the deal covers the commercial benefits too - IBM and KAUST will share any intellectual property generated by the centre. In a recent Economist Intelligence Unit study, Saudi ranked very low on IT competitiveness - mostly due to a lack of local IP generation and a lack of IT infrastructure. Developing new innovations in Saudi, that will have a financial benefit for the university, will help to remedy the lack of technology that is created in the Middle East.

If broader development of IT talent needs greater PC and broadband penetration, then supercomputing marks the other, thin end of the wedge - it may be a narrow focus that might not mean a lot to most people, but this investment could be an important step in the Middle East taking an major role in driving the research necessary to build web-scale computing.

Add a Comment

Your display name This field is mandatory

Your e-mail address This field is mandatory (Your e-mail address won't be published)

Security code