The long-awaited UK national supercomputer has finally been unveiled. The computer, dubbed HECToR, is based in Edinburgh and will be open to academic and industrial users to perform computationally intensive tasks that would otherwise be impossible using conventional systems.
HECToR, which has taken more than a year to install, can perform computations at a staggering rate of 63 million million operations per second (the equivalent of more than 10,000 laptops), leading to a ranking of 17 in the November 2007 Top500 list of the fastest computers in the world.
The computer is run by the Edinburgh Parallel Computing Centre (EPCC), an institute of the University of Edinburgh, which had maintained the UK’s previous national supercomputer, HPCx and already contained the infrastructure to accommodate the new system.
‘In terms of installation, powering the machine and cooling the heat generated by the computer proved to be the major civil engineering challenge,’ said David Henty, the group manager at EPCC.
The funding council behind HECToR chose Cray to provide the hardware after extensive testing of the various competing architectures. Henty told scientific-computing.com that although IBM’s BlueGene system has the world’s highest peak performance for the jobs used to compile the Top500 list, Cray’s system proved to be more efficient at the kinds of tasks that would be important for scientific research in the UK.
Academic users can gain access to HECToR through an application process similar to the method used to award funding grants to research groups. The institute has also appointed a commercial director to locate industrial users that may benefit from HECToR’s enormous processing power.
With dual-core processors present in everyday laptops, Henty believes that researchers are now more aware of the benefits of parallel computing than ever before – a fact which should benefit HECToR’s popularity as a viable resource.
‘In the early days, 10-15 years ago, using parallel processes seemed fairly esoteric. Now it’s more mainstream and it’s more common in industry in particular,’ he said. ‘The challenge is that these machines are getting bigger and bigger, but parallel processing is becoming a more standard technique.’
HECToR will be used for traditional applications of high-performance computing like molecular modelling and climate modelling, but the centre also plans to encourage users which would not have previously considered parallel computing to use HECToR, for tasks such as financial modelling.