Imperial College London and the University of Cambridge have merged their High Performance Computing (HPC) facilities to create one massive cloud supercomputing resource, known as CORE.
Together, the two facilities amount to 300 teraFLOPS of sustained double-precision computing delivered by over 22,000 Intel processor cores, attached to more than 3 petabytes of high performance file system.
In total, CORE users have access to around £8 million worth of capital infrastructure, including some of the single largest systems in the UK, such as the fastest Intel-based HPC system in the UK, which has over 9,000 cores and is rated 93rd in the Top500 list.
“A lot of people that are new to HPC start off by running applications on a workstation. When they fill that workstation they buy another one and so on until they're told they can't plug in any more workstations, and that's when they come and talk to us,” said Dr. Paul Calleja, director of HPC at the University of Cambridge.
“It's relatively easy to move from a workstation to an Intel cluster because the core of the machine is the same – it's an Intel processor. So this allows people to scale up very easily from the four workstations under their desk to the cluster.”
Other systems that make up CORE include the largest shared memory system in the UK, with over 16TB of RAM, and one of the UK’s largest NVDIA GPU clusters. Several of the systems are partly owned by the DiRAC consortium, led by the Science and Technology Facilities Council.
The HPC cloud is available to both industry and academia. Calleja explained that CORE is a cloud in the sense that resources can be accessed remotely over the internet, but should not be compared to Amazon or other virtual machine clouds, because a high level of support is needed.
For partners who would prefer to deploy their own in-house systems, CORE also provides consultancy services covering procurement and design, project management, system integration, performance analysis and optimisation.
CORE is part of the UK government's e-Infrastructure expansion programme, led by the Department for Business, Innovation and Skills (BIS). The programme is aimed at delivering business-ready HPC and big data solutions to both industry and academia.
The two universities are now working to achieve uniformity of access and support infrastructure, and to integrate their talent pool of systems staff.
“CORE can take this increased talent pool, create a range of hardware and consultancy services and actively push those services out to industry,” said Calleja. “This will help to increase competitiveness, drive growth and facilitate technology transfer.”
Although the collaboration was only officially announced this week, the universities have been working together in stealth mode for many months to deploy CORE across a wide range of industrial sectors, together with over 26 partners including Rolls-Royce, Caterham F1 Team, and Audio Analytic.
“The flexible scale out resources provided by CORE combined with excellent technical support has provided us with a step change in compute capability increasing the scope of what we can achieve and reducing time to market,” said Chris Mitchell, CEO and founder of Audio Analytic.
CORE operates on a pay-per-use model, providing both SMEs and larger corporations with on-demand access to large-scale processing and data storage infrastructure. The universities are working with both Dell and Xyratex to develop Big Data storage solutions, including a Hadoop On Demand system.
Organisations can either run their own workflow or use solution packages, which are currently available in four verticals: engineering, life sciences, materials modelling and digital media. CORE is also offering a free starter package, which includes 10,000 core hours free and one day of free support.