The University of East Anglia is continuing its quest to bring HPC to greater numbers of ‘non-traditional’ research users by investing in a powerful new High Performance Computing (HPC) system to sit at the heart of its HPC service.
The system marks the University’s second new HPC system in four years and, like its predecessor, can be easily scaled and expanded in the coming months through a framework agreement to match rapidly increasing demand for compute power.
Chris Collins, Head of Research and Computing Support at the University of East Anglia said: “The University has long recognised the benefits of HPC to multiple disciplines of research. We received funding in 2010, which was provided in a conscious action to get more non-traditional HPC users on board. We scaled that system continually because of demand and now, four years later, we’re replacing it with a more powerful system. We’re keen to add to our existing 500 research users.”
The new system, designed and integrated by HPC, big data and analytics integrator OCF, comprises Fujitsu Primergy servers with IBM’s Platform 4.2 cluster management system and integrates with the University’s existing IBM SAN storage system with IBM Spectrum Scale [formerly GPFS]. Some of the newer hardware from the University’s old cluster will be migrated over to the new HPC environment by OCF during the next few weeks, with nearly 6,600 cores available to researchers at the University by early 2016.
It is important for us to respond to changing demands for HPC within the University
The University boasts HPC users from traditional disciplines including climate change, bioinformatics, computational chemistry, environmental sciences and computing. Plus, users from non-traditional disciplines too, such as the faculty of medicine, business schools and economics department.
A research group from the School of Computing is currently piloting the new HPC system, using its compute cores to investigate a computational comparison of varying algorithms. Benefitting from the newer architecture, the researchers have run around half a million jobs in the first few weeks.
The Fujitsu cluster will be housed across the University’s two data centres. The migration to the new cluster will also involve moving around 800 different applications that cater for the diverse usage of the HPC system across the University’s departments. Mostly Open Source, it also includes some central licenses for commercial software, different bioinformatics tools and other applications such as MATLAB, R and Python.
Chris Collins added: “It is important for us to respond to changing demands for HPC within the University. OCF’s people, expertise and knowledge in both HPC and higher education, accessible through our framework agreement, gives us the opportunity to continually expand our service with the best technology, in the most cost effective way. We are also benefitting from OCF’s world-leading partnerships within the HPC industry, which has given us the opportunity to widen our own understanding of potential opportunities in HPC and make use of the best technologies.
Regular University funding over the past four years has meant that we can be forward thinking and can plan our resource growth more effectively. This worked with our previous cluster in 2011 and now with our second, we can work with OCF to purchase and install new hardware, so we can scale the HPC service when required. Plus, rather than using – for example, InfiniBand – across all of the compute nodes, we can identify specific areas like GPU or parallel utilisation that require further upgrades, so that our investment is more targeted and cost effective on an on-going basis.”
The University plans to expand the HPC system and purchase further hardware before July this year.