Computing Colossus

As increasingly cutting-edge research takes place in our universities, there’s more of a need to have the IT power to match. And it’s not just in traditional fields of research either. Once only the preserve of physical sciences, now the humanities have just as much stake in grabbing their chunk of memory. So how do you balance such competing demands, and deliver the right strategy for these mammoth machines? David Harrison, Assistant Director of Strategy and Enablement Information Services at Cardiff, takes up the story:

“We would like to think that the university is more aligned with the ways students communicate nowadays on social networks. It’s the way they think and act, and we need to be more receptive to that way of working. In a drive to improve business efficiency, we’ve engaged in a partnership with IBM to deliver something we call the Modern IT Working Environment. The first version will deliver a set of communication and collaboration tools that will lead to an ambitious business process modelling implementation.”

It sounds as though Cardiff is indeed redefining the script, placing IT firmly at the core of tomorrow’s strategy. At the heart of this is the realisation that it’s not good enough to have students experiencing poorer download speeds or information access depending on where they sit on campus. True service orientation avoids this:

“We are constantly working on how systems talk to each other on an enterprise service bus, as part of a service-orientated architecture strategy. This makes for better information flows and less replication and errors, as well as saving time. For students and staff, we want to make things seamless and less reliant on paper.”

Cutting down the need for paper is a laudable goal, but it also appears that such IT systems require some serious wattage to keep things running. Professor Martyn Guest is Director of Advanced Research Computing at Cardiff, and he’s the man who makes things tick at the frontline.

“In terms of Bull [the current system], and our overall efforts in high-powered computing, the research agenda over the last four years has shown that costs have gone down, that there have been major injections of funding, and several hundred-fold increases in capacity. This means it makes sense to look at these kind of systems.

“You need to remember that you’re talking figures in the region of £2.9 million in terms of investment in supercomputing. You’re talking three-year procurement phases and these machines are big. There’s lots of real estate needed to house them and 150/200kw of power is necessary to run them, which requires water cooling and major chillers. Initially, we had conversations with seven or eight potential suppliers but, in terms of the technology itself, eventually it became clear that Bull was the right option.”

Such scary figures point to serious headaches in terms of the infrastructure that’s needed to take on a project of this kind. And how long might it eventually take to payback? Luckily, it appears that this new breed of supercomputer has a nifty way to share its appeal:

“This type of computing relies on clusters – lots of individual PCs linked together through an interconnect. You need bespoke software to run these effectively and this enables the co-operation of some 2,000 PCs altogether.

“The old days of a single mainframe are long gone – our machine is around 2,000 times faster than the standard office desktop model. You’ve the option to run 2,000 small jobs or one big one. We can potentially impact across any area of research from the physical to the social sciences and we should be going live officially on 2 June.”