Customer Solution Case Study
/ University Develops Non-Invasive Treatment for Cancer with High-Performance Computing
“Simulation of the 3-D oesophagus transient problem for 45 minutes real-time patient therapy now takes less than five minutes. This gives clinicians an almost immediate prognosis.”
Prof. Kelvin Donne, Dean, Faculty of Applied Design and Engineering, Swansea Metropolitan University
Swansea Metropolitan University (SMU) has gained international fame for excellence in research into cancer by its Applied Design and Engineering Faculty. Scientists needed a fast high-performance computing (HPC) tool to support clinicians treating patients with oesophageal cancer. They upgraded to Windows Server® 2008 with the Microsoft Message Passing Interface (MS MPI) for parallel testing on a single workstation to support the non-invasive treatment.
This case study is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY.Document published November 2009
Business Needs
Applied research with HPC tools at SMU in Wales has produced a computational model for laser-based cancer therapy that can help optimise the treatment protocol. The university has established a reputation for educational excellence over a period of 155 years. SMU has around 6,500 students, of which more than 1,500 are postgraduates. In 2009, the university achieved a 36 per cent increase in its research funding—the largest of any traditional university in Wales.
Prof. Kelvin Donne, Dean, Faculty of Applied Design and Engineering, SMU, says: “Although our research budget is always stretched, funding was secured from the Welsh Assembly and the European Union for our pioneering research into cancer treatment. We worked on modelling the photochemical and photothermal effects in light-based therapy for oesophageal cancer. This is non-invasive treatment—diagnosis and cure using laser tools without conventional surgery.”
The pioneering cancer treatment project went on to win the award for Most Innovative Use of Simulation Technology at the World Congress of the National Agency for Finite Element Methods and Standards—an independent not-for-profit organisation—in Vancouver in 2007.
Prof. Donne says: “The principal technical challenge we faced was to produce a parallel version of our transient boundary element code extremely quickly. It needed to be easily integrated with the previous research coding and achieve rapid debugging.
“We also required a range of diagnostic tools to help with debugging the software and a system with the ability to interoperate with diagnostic tools. If we were going to invest in HPC, we were keen to buy an environment suitable for use in other projects—for example, research into how glass fractures for a major automotive company.”
Solution
SMU initially used a high-performance cluster running Windows Server 2003 with the Windows® Compute Cluster Pack from a third-party supplier in Germany. It has since replaced the operating system with Windows Server 2008 and migrated to Windows HPC Server 2008. The deployment now benefits from new diagnostic features, which are native to Windows Server 2008 and vastly improve the time taken to debug code. Scientists compiled the boundary element modelling feature with the Microsoft® Visual C++® 2005 development system, which has tools for developing and debugging C++ code.
Critically, SMU chose the Microsoft Message Passing Interface (MS MPI) software development kit, which was essential for rapid results. “MS MPI produces a speed up factor of 30 compared to any other solution,” says Prof. Donne. “It also integrated easily with previous research coding, some of which was nearly 40 years old.”
A decision to work with Microsoft rather than Linux owed a great deal to the support from the Microsoft U.K. HPC user group, particularly as SMU researchers had no previous experience with MS MPI. “We made contact with Prof. Richard Hall, Director of the Midlands Simulation Group at the University of Wolverhampton, and, as a result of expert advice, we verified our parallel version of the transient boundary element code,” says Prof. Donne.
Later, when SMU decided to consider changing the operating system to Windows Server 2008, the HPC user group recommended Microsoft Partner dezineforce for the implementation. Dan Adams, Consultant, dezineforce, says: “Windows HPC Server 2008 offers simple yet powerful out-of-the-box deployment and management capabilities based on standard Microsoft technology. With our extensive knowledge in the Windows HPC space, we were able to recommend with confidence that SMU continue with an end-to-end Microsoft solution—there was no viable alternative to Microsoft products for this kind of ground-breaking research.”
Explaining his confidence in Microsoft for HPC in research, Prof. Donne says: “Taking the Microsoft route ensured that we achieved results far more quickly than we would have with Linux, thereby enhancing our reputation for cutting-edge research. We had very little knowledge of MS MPI when we started, but dezineforce and the user group supported us extremely well.”
Benefits
Use of the MS MPI software development kit for parallel testing on a single workstation proved crucial for rapid development of the pioneering solution. It reads three-dimensional (3-D) scan data and undertakes predictions of the chances of successful treatment in less than five minutes. Prof. Donne says: “Simulation of the 3-D oesophagus transient problem for 45 minutes real-time patient therapy now takes less than five minutes. This gives clinicians an almost immediate prognosis and leads to better patient care without the need for conventional surgery.”
· Diagnostic tools native to Windows Server 2008. Diagnostic tools for debugging software that previously required third-party solutions are now native to Windows Server 2008. “This is a huge advantage for the high-performance computing community because it means there is no need to separately introduce Linux diagnostic tools,” says Prof. Donne.
· Fast time to market. With MS MPI, working in parallel on a single workstation for boundary element modelling was simplified—full 3-D code was running and tested in just one week, saving significant amounts of operator time.
· Scalable platform for other research projects. The award-winning solution is easily applicable to other projects. SMU is implementing the boundary element code to model stress analysis for a leading company in the automotive industry.
· Support from Microsoft HPC user group. SMU, with its limited research funds, needed reassurance that it was on the right path, given its lack of experience with MPI. The university benefited from advice and a recommendation to a Microsoft Partner to manage the migration to a new operating system, together with associated training.
This case study is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY.Document published November 2009