SGI recently completed a project for The Oxford Supercomputing Centre, who provide the HPC services for Oxford University. DCM found out how the project was managed and implemented

Universities are increasingly faced with the challenge of managing growing quantities of research data. In addition, research institutions may now also be involved in commercial services, further increasing the amount of data generated, placing additional burdens on IT systems.

One example of an organisation that has been required to increase computing power to address demand for HPC applications is the Oxford Supercomputing Centre (OSC). The OSC provides high performance computing (HPC) services to researchers at the University of Oxford. The supercomputing centre assists with visualisation across a wide range of multidisciplinary collaborations and serves academic research in astronomy, chemistry, biology and, increasingly, bio-informatics and humanities. To support this, and its recent expansion into providing HPC services to commercial users - who lease the computational capabilities as well as utilise in-house expertise - the OSC needed to upgrade its existing hardware to gain additional performance and memory capabilities.

The centre required a particularly flexible solution, able to rapidly process vast amounts of data, and provide high performance shared memory for running demanding technical applications such as Gaussian.

The OSC initially reviewed solutions from a number of vendors before selecting SGI to upgrade its existing hardware configuration. With an SGI Altix 4700 and an Altix ICE, the OSC now has a balance of distributed and shared memory platforms that lend themselves to many different high-end applications.

“When working with the broad range of applications, data and computational requirements for university research, it is essential that our equipment can meet our exacting demands. In addition, the centre has equally rigorous requirements for reliability and energy efficiency,” said Jon Lockley, manager of the OSC.

The SGI Altix 4700 provides over one Teraflop per rack and is optimised for data intensive workloads. By incorporating shared memory, it simplifies software development, workload management and system administration. The Altix 4700 platform also features modular blades with interchangeable compute, memory and I/O components, as well as special purpose blades for ‘plug and solve’ to simplify configuration and maintenance.

It was also important for the OSC to implement an easily scalable solution enabling the institution to keep pace with evolving technology needs of the user community.

Altix ICE can be configured with up to 512 processor cores per rack —providing up to six Teraflops of computing power. Featuring cable-free blade enclosures, integrated switches and a high-performance interconnect architecture; it can scale to thousands of nodes, but has optimised power and cooling overheads, thus keeping operating costs down.

“With the SGI Altix, the OSC can tackle complex research projects that were not previously possible,” continued Lockley. “SGI’s designed-to-order model means that the SGI systems we rely on exactly meet our own, very specific needs.”

Since the opening of the OSC, SGI Altix has enabled researchers to make significant advances in academic fields and has assisted in forwarding a wide range of studies. Widely diverging research projects have included the study of mantle flow under mid-ocean ridges to modelling and assessing complex social trend data.

The OSC is now also fully equipped to deliver HPC services over the cloud. By capitalising on the operating efficiencies that have resulted from this large scale service deployment, the university is offering competitive rates for its compute-on-demand services.

“We have experienced great progress in our research capabilities since implementing the new HPC systems. Now, as we begin to undertake commercial projects, we are enabling local businesses to capitalise on the university’s resources and expertise,” continued Lockley. “The OSC is now fast becoming a leading facility for supercomputing for both academic and commercial purposes.”