Delivering a universal HPC facility

Summary
OCF and IBM worked in partnership with the Information Systems Services Department of a UK university to deliver a complete HPC solution including hardware, software, services and support. The combination of high performance, reliability, and ease of management has ensured the success of the project and the university plans to add further clustered compute facilities as new projects and demands emerge.

Client profile
Southampton University is one of the UK's top 10 research universities, supporting a population of 20,000 students and 5,000 staff.

The challenge
The Information Systems Services Department (ISS) identified a clear need to significantly increase the university's capability in research computing and to upgrade the existing Iridis compute facility. The facility needed to be versatile enough to bridge the gap between the university's existing cluster and national facilities such as HPCx.

The solution
OCF capitalised on its unique position in the UK HPC marketplace and offered specialist HPC skills and services, hardware, support and software from IBM, and support from Business Partner AMD’s market-leading Opteron CPUs.

OCF was able to assure the university of both access to the new technology and future value through a phased roll out. This was essential to steering the evolving specification of the cluster to match any new demands placed upon it.

The benefits
The initial installation of 330 Opteron cores in IBM e325 Servers, linked by a GigE network and served by 10TB disc, has grown to encompass over 1000 cores (a section of which is interconnected by Myrinet networking) over 700GB RAM, 25TB storage, and a full remote management solution.

The e325/6 has proven to be an excellent base for the cluster, demonstrating excellent performance, manageability and reliability. OCF has also been able to assist the university in installing IBM's leading cluster systems management (CSM) tools, ensuring the manageability of Iridis. High levels of reliability, backed by IBM maintenance services, is critical to the operational success of a facility the size and reach of Iridis as a large portion of the university's research community rely on it for computational support. The combination of high performance, reliability, and ease of management has ensured the success of the Iridis project.

The future
OCF continues to collaborate with the university, adding clustered compute facilities in partnership with IBM as new projects and demands emerge.

The final system comprised elements of large SMP, loose and closely coupled COTS Clusters and Storage. The Cluster currently averages in excess of 90% utilisation, proof of both the need and the solution

Visit Soton's Web-Site for more information