August 29, 2013

HPC Wales Creates Distributed Supercomputing Network

Alex Woodie

Businesses, schools, and other organizations in Wales now have access to HPC resources as a result of a new distributed supercomputing network unveiled by HPC Wales in May 2012. The group says the network is the first of its kind in the UK, and will not only be a boon to Welsh researchers but will help with education too.

HPC Wales’ new supercomputing network is composed of a series of clusters located in Welsh universities, business, and research centers that are connected by high-speed links. The network is managed in part by Fujitsu, HPC Wales’ technology provider.

The cluster is composed of two hubs, three tier-one sites, and two tier-two sites. The Cardiff Hub located at the Advanced Research Computing at Cardiff Data Centre in Cardiff University has about 8,000 Intel Westmere and Sandy Bridge cores available for work across three primary systems, including “Goldstone,” “Granite,” and “Garnet.” The Swansea Hub, launched in April 2013, is located at the Dylan Thomas Centre in Swansea, and has about 6,000 Intel Sandy Bridge cores available for work across two systems, including “Diamond” and “Dravite.”

HPC Wales has a total of 650 Intel cores available for small and moderate jobs at the tier-one systems, which include the “Tiger Eye” system located at Bangor University; the “Opal” cluster at Aberystwyth University; and the “Iron-Ore” cluster at Glamorgan University. The tier-two resources comprise about 150 Intel cores, and include the “Jet” cluster located at Glyndwr University and the “Shaman” system at Swansea Metropolitan University. Also part of the distributed system is the Power7-based specialized analytics system running at Swansea University.

The entire setup is connected by a dedicated network delivered over the Public Sector Broadband Aggregation (PSBA). It features a 10 GBit/sec link between the two hubs and 1 GBit/sec and 100MBit/sec links connecting all other sites, according to HPC Wales.

The Welsh HPC community convened in May 2012 at the Senedd in Cardiff Bay to celebrate the launch of the network, which was made possible thanks to a £24 million investment from the Welsh Government and the Welsh European Funding Office, and a £10 million investment from the UK Department for Business, Innovation and Skills (BIS).

Among those speaking at the event was professor John Harries, chief scientific adviser for Wales. “High performance computing is one of the essential tools in the toolbox for both researchers and innovators who are aiming to compete across the UK and abroad today,” Harries said.

Dr. David Willock, a professor at Cardiff University, said the new distributed supercomputing network will make small businesses in Wales more globally competitive. “That is a world stage they [small businesses] play on, and without this resource, other people in America and Europe will have this facility in the future. Wales is ahead of the game on this, so it’s really going to help the businesses in Wales.”

The national supercomputing network will also help with training, according to David Craddock, the CEO of HPC Wales. “I’m very pleased to announce an HPC world training link with the National Science Foundation in America, which will significantly enhance our HPC training provision,” he said at the Senedd.

A few words were also said by Michael Keegan, executive director, technology product group, Fujitsu. “Not only is Wales now home to an enviable high performance computing network, but both business and researchers right across Wales are now able to access this innovative technology, securely and locally,” he said.

“The skills development program will put Wales firmly at the center of the HPC world stage,” Keegan continued. “It offers an excellent opportunity for personal and professional skills development, and increased business competitiveness underpinned by a world class workforce.”

Related Articles

The BlueBEAR Cluster Goes Online in the UK

New HPC Cluster Frees Scottish Researchers From Administrative Burden

Considerations in HPC Cluster Network Selection

Share This