The Irish Centre for High-End Computing recently celebrated a decade in existence, making now a perfect time to take stock.
The Irish Centre for High-End Computing (ICHEC) has undergone some changes over the years, evolving into a viable research entity at a time when global investments into supercomputing have never been higher.
Arms race
Annually, an arms race between economic powerhouses reveals new winners. Last summer, China emerged as the owner of the world’s most powerful computer.
Although there’s always a leader, this development was new as, unlike China’s previous entries in the list, the 93-petaflop machine was entirely created in China, without the need for any western assistance.
Called TaihuLight, it’s currently up and running at China’s National Supercomputing Centre, being used for research work across climatology, life sciences and data analytics.
Not to be outdone, the US is looking at the long term, aiming for a machine that will make TaihuLight seem like a Game Boy.
US president Barack Obama last year signed a plan to build the world’s first exascale computer – meaning 1,000 petaflops. Costing $200m to develop, the 15-year plan is fairly wide-ranging, with the US Department of Defense, Department of Energy, and National Science Foundation involved.
Irish effort
In Ireland, things operate on a smaller scale. ICHEC manages the Irish supercomputer Fionn, not as powerful as many international contemporaries, though effective nonetheless.
At a recent celebration of ICHEC’s existence, morale was flowing as the future of the industry was discussed. For anyone who thinks tomorrow’s challenges will be different from today’s however, history paints a different picture.
“I’ve been in this industry for 25 years and I’ve learned that the paradigms all tend to stay the same,” said Prof JC Desplat, director of the ICHEC. We’re told that things will get simpler, that compilers and architecture are so advanced, before another technological shift makes it all seem antiquated.
However, now there is significant investment into supercomputing, with everybody getting on board.
“We have an ambition to do more,” said Roberto Viola, director general of DG Connect at the European Commission, when in Dublin recently to discuss employment, innovation and supercomputing.
“The private sector, academia and [the EU] need to work together across multiple topics.”
State aid
It’s when bodies such as the EU, or the US in particular, get behind supercomputing initiatives that advances truly kick off.
Obama’s pledge for a 15-year lead into the supercomputer of tomorrow is one such example, or the genuine advances made in health on the back of greater access to data and data management tools.
Calling the amount of data stored in the EU as “tiny” in comparison to what’s used, and what’s stored elsewhere, Viola thinks the OECD stats “are impressive”.
“We’re doing the impossible,” he said, of maintaining some form of fundamental rights in a data-currency world, all the while “investing in supercomputers” as much as possible.
And in Ireland, it’s starting to bear fruit.
“In many cases when we work with industry, we try to help them to shorten the time for general availability,” said Desplat. “So say it’s for new technology; we do internal testing, we may help them in certain aspects of product element. We tend to be involved at alpha stage or pre-alpha stage.”
Big footprint
That involvement has come across 1,200 projects in the decade since ICHEC came into existence. Computing assistance for international projects can only be achieved if you’re hooked into a continental framework, of which ICHEC is.
PRACE, the Partnership for Advanced Computing in Europe, welcomed the Irish body into the fold eight years ago and, since then, things have sped up dramatically.
Desplat’s organisation has brought Irish universities to the fore, secured €2.7m worth of computing time and seen the likes of Met Éireann, Tullow Oil and the Biomedical Diagnostics Institute secure its services.
Indeed, in the latter case, it touches on an area which Desplat finds “truly inspiring”: high-performance computing for the prevision of medicine.
This is something on which Viola agrees, claiming the future of medicine “needs big data and computing power”. However, supercomputing needs the public to understand all the good it does.
The public needs to know
Desplat says the use of high-performance computers is “becoming a necessity”, but how will that translate in terms of increasing life expectancy?
“The public needs to understand and value the initiatives,” he said, noting that ambivalence is not good enough. The public needs to know what massive computer power can do.
“Imagine the concentrated power of more than 1m laptops working to screen a tumour sample from a patient against thousands of drugs and millions of drug combinations,” asked Warren Kibbe, director of the US National Cancer Institute, recently.
“At the end of this screening process, this mega-computer would help to identify a specific treatment with the greatest potential to combat that patient’s cancer.”
If the public knew of initiatives like that, then support would inevitably grow.