[AI] Hector - the supercomputer set to go live

renuka warriar erenuka at gmail.com
Wed Jan 2 07:39:44 EST 2008


The Hindu News Update Service

News Update Service
Wednesday, January 2, 2008 : 1635 Hrs

Sci. & Tech.
Hector - the supercomputer set to go live

GUARDIAN NEWS SERVICE

By James Randerson in London

GBP113m HECToR will help British researchers simulate everything from climate change to financial markets

To the untrained eye, it looks like a sports hall half filled with neatly arranged burgundy Ikea storage units. But the series of heavy, clunking security
doors that must be navigated to get inside, and the staff's insistence on secrecy about its precise location, indicate that this is a place of national
importance.

Meet the University of Edinburgh's HECToR, Britain's newest and faster supercomputer, which scientists and engineers plan to use for modelling everything
from climate change to financial markets. The GBP113m publicly funded project, the full name of which is High-End Computing Terascale Resource, will be
opened officially later this month by the chancellor, Alistair Darling.

HECToR's computing power is astonishing. Its calculation speed is equivalent to every person on the planet performing 10,000 calculations every second -
in computing terms that is the same as 12,000 standard desktop computers operating at full tilt. HECToR's memory is also impressive. It is 3,200 times
larger than that of a top-of-the-range iPod 160GB.

According to Professor, Arthur Trew, who is in charge of the project at a site in the Scottish countryside, supercomputers such as HECToR are allowing scientists
to approach problems in a radically new way. Traditionally, science has advanced through theory and experiment - the theorists come up with testable ideas
and experimenters design experiments to see whether they are right. But for many areas of science this does not work.

In climate research, for instance, it is impossible to do the experiments. In fluid dynamics - the complex field that governs the way, for example, that
air flows across a wing - the mathematical equations governing what is going on are impossible to solve. That is where HECToR's brain comes in. Scientists
can use it to run models of the real world.

One application is in aircraft design. By better understanding the flow of air across a wing, scientists are able to design shapes that will minimise drag
and increase lift, making aircraft more efficient. The current generation of passenger jets were built before computers were fast enough to do this sort
of modelling.

Professor Jacek Gondzio at the University of Edinburgh plans to use HECToR to model financial markets. He is working on finding the safest and most profitable
investment strategies for pension funds, based on uncertain information about the future of the world economy.

"Uncertainty needs to be modelled by multiple scenarios and in order to reflect reality this automatically expands problems to large sizes," he said.

The approach means building computer models that predict the outcome of different investment strategies based on huge numbers of different scenarios for
the future of the economy. This allows Gondzio to select the best investments that will pay off most of the time.

A faster computer also means better models.

"The ability to solve larger problems means the ability for more accurate modelling," said Gondzio. "You can either do the planning for more years ahead
or you can have a more detailed description of scenarios."

Up to now, most supercomputer modelling has been done by physicists, engineers and chemists. But the improved hardware is now making biological modelling
possible.

One research group will use HECToR to understand how molecules enter cells through the cell membrane. That could lead to more effective drugs. Another group
is modelling the electrical activity of the heart to understand how best to intervene when it goes wrong.

Trew said that UK academic scientists would be given priority on HECToR, which they can access via the internet. If their research proposal is good enough
it is free to use. "Scientific excellence is the number one driver," he said.

The cost of the project has risen since HECToR was first conceived. In November, a report by the Commons public accounts committee into the costs of the
UK's 10 flagship science projects criticised their management. But HECToR emerged reasonably well.

It has cost GBP5.6m less than the GBP65m estimate to build, but its annual running costs have jumped from the GBP5.4m estimate to GBP8.2m. Trew said this
was because electricity prices in the UK had nearly doubled since the planning stages. "I think the initial estimates for power costs were unrealistically
low, but power does cost an awful lot more today than it did 



More information about the AccessIndia mailing list