EPCC today and tomorrow
As well as expanding into Europe and writing the roadmap for the future of high-performance computing, the EPCC also continues to focus on industry projects, including everything from simulating climate change to how to tow a fishing net to maximise the catch. Over the last eight years, EPCC has spent almost £150 million on computing resources, but as Professor Trew said at the recent celebrations, “EPCC is not machines but its people.”…
Article by Peter Barr
As well as expanding into Europe and writing the roadmap for the future of high-performance computing, the EPCC also continues to focus on industry projects, including everything from simulating climate change to how to tow a fishing net to maximise the catch. Over the last eight years, EPCC has spent almost £150 million on computing resources, but as Professor Trew said at the recent celebrations, “EPCC is not machines but its people.”
Looking to the future, EPCC recognises that budgets may come under pressure, but “parallelism is now endemic,” said Trew, and a golden age may lie ahead for high-performance computing.
Deputy director Alison Kennedy stressed the importance of Europe in EPCC's future plans, and said that it must “prove itself in Europe at the same time as delivering results at home.” As a sign of its engagement with colleagues in Europe, EPCC has hosted more than 800 research visits in recent years under the HPC-EUROPA Programme which links across both universities and application areas – for example, automatic inspection technology, renewables, health and well-being. “We look at all these problems from an international perspective,” Kennedy added, with collaboration focusing on expertise and services as well as on machine time. She also suggested that EPCC will have an increasingly key role to play in the future in the development of a European Data Infrastructure to cope with the data explosion, as computer simulation and data become more closely linked.
Twenty years ago, scientists dreamed of the kind of machines now available to the researchers at EPCC. For example, HECToR (High End Computing Terascale Resource) runs a million times faster than the “supercomputers” of only two decades ago. EPCC recently upgraded its facilities, and HECToR now has more than 44,000 Opteron processing cores and runs at more than 350 Tflop/s (Tflop/s = 1012 floating point calculations per second). In the 1980s, the quest to build a Gflop/s (109 floating point calculations per second) computer was considered wishful thinking, but now the target is the Pflop/s computer, capable of 1015 floating point calculations per second. EPCC has also had the equipment to make sure it has kept ahead of Moore's Law for the last 20 years, more than doubling its performance every two years, and consistently ranked in the world's top 20 HPC facilities.
In addition to its raw computing power, EPCC also strongly promotes its portfolio of services, including its helpdesk, training and consultancy services, as well as its involvement in collaborative research – for example, IBM and Cray.
EPCC has also not forgotten its role as a teaching resource, and introduced one of the world's first MSc degrees in HPC ten years ago. Next year, it plans a big increase in student numbers. “Universities should not be diverted from their core purpose,” said Parsons. “It would be a big mistake to focus on short-term gain.”
The challenge for the future, in the eyes of the EPCC, is not just to accelerate the power and intelligence of supercomputers but also to convince industry and government that it is worth it to invest in high-performance computing – like scientists in the USA, China, India and Japan, who are already investing billions of dollars in HPC. In Japan, for example, Fujitsu is involved in a $1.2 billion government project to develop a supercomputer capable of running at 10 Pflop/s, using a technology called “Tofu” – so maybe the link between supercomputers and mushrooms is not so strange, after all.....