2 page Case Study - Posted 11/16/2009

Views: 29

Rate This Evidence:

University of Bologna

University Physicists Reach New Research Horizons with High-Performance Computing

The Department of Physics at the University of Bologna needed to increase its computation speed for complex tomographic analysis of large works of art. After successful testing, it now conducts its calculations on a cluster that runs the Windows® HPC Server 2008 operating system. Doing so made it possible for university researchers to avoid rewriting code, reduce calculation times from 20 days to 6 hours, and expand into new areas of research.

Business Needs

Mystery surrounds various works of art, from Leonardo da Vinci’s Mona Lisa to the statues on Easter Island. At the Department of Physics at the University of Bologna, physicists and computer science experts in the department’s X-Ray Imaging Group work together to develop computerized tomography techniques that can be applied to not only the art world but to medicine and business as well.

Tomography—imaging by sections through the use of energy waves—is nothing new, but a group led by Professor Franco Casali has developed innovative tools and methods for high-resolution tomographic analysis of large works of art. These new three-dimensional analysis techniques require considerable data processing. In fact, the university recently conducted tomographic analysis in collaboration with the Conservation and Restoration Center in Turin, Italy. The joint team studied a two-meter, thirteenth-century Japanese statue known as Kongo Rikishi and generated 24,000 radiographs and 120 gigabytes of data in the process, which took two months.

The X-Ray Imaging Group ran the study using software, written in C++, that was installed on a computer with an Intel dual-core processor and running the Windows® XP operating system. It took 20 days for the computer to complete the calculations necessary for the visual restoration of the tomographic analysis. “Our new techniques deliver better resolution than the medical standard and use a complex data-processing algorithm, so the computations are complicated,” says Professor Giuseppe Levi of the University of Bologna Department of Physics. “But taking that long to compute data isn’t feasible for our analysis because researchers often have to travel to the artwork to conduct studies.”

The university needed a hardware and software platform that was capable of carrying out complex calculations at high speeds and doing so from any location. However, in moving to a more powerful platform, it didn’t want to have to rewrite its code.

Solution

The university assessed the information systems that could offer it the calculation power necessary to reduce the tomographic image reconstruction process to just a few hours. Its research centers already had Linux-based clusters for high-performance computing (HPC), but using those clusters would have required at least a partial rewrite of its code. The university also believed that using Linux clusters would result in compatibility problems.

We get results on the cluster that are faster than our previous setup by a factor of up to 75 …  [W]e’ve reduced image processing time from 20 days to just 6 hours. Giuseppe Levi Professor, University of Bologna Department of Physics

After defining automation scripts for the graphic-restoration process, the university conducted tests with increasingly complex objects. The results exceeded all expectations, and the university decided to deploy Windows HPC Server 2008 with a similar architecture in its own laboratories. “Following the positive results obtained in Redmond, we configured our own local cluster very easily. Within a few hours, we defined the node image and sent it to all machines, thus preparing the new parallel calculation architecture,” says Levi. “A similar operation on Linux systems would have presented a greater level of complexity. With the Windows platform, we could use Visual Studio 2008 to debug our code, eliminate defects, and make the code more robust, all of which contributed to an increase in performance.”

The X-Ray Imaging Group laboratory now has a cluster with four nodes and eight cores running Windows HPC Server 2008. The group uses the system’s Job Scheduler and Node Management utilities as its primary management tools. “The simplicity of using integrated tools for the configuration of parallel calculations is impressive, and the Job Scheduler has shown itself to be extremely intuitive," says Dr. Rosa Brancaccio of the Department of Physics at the University of Bologna. “Using those tools, we now have the ability to check the outcome of every single task in real time.”

Benefits

The Department of Physics at the University of Bologna now has a system that provides significantly faster tomographic analysis than was previously available, thus making it possible to conduct new, innovative studies. Benefits include:

• Accelerated calculation times. The university has experienced a significant increase in speed since its move to Windows HPC Server 2008. “We get results on the cluster that are faster than our previous setup by a factor of up to 75,” says Levi. “For the analysis of the Kongo Rikishi statue, for instance, we’ve reduced image processing time from 20 days to just 6 hours.”
  
  
• Ability for real-time analysis. Because the university can carry out calculations quickly, it will be able to conduct real-time analysis, which is particularly important in medical research. “The possibility of real-time analysis will resolve the problems connected with small movements of patients during CAT [computed axial tomography] scans,” says Brancaccio.
  
  
• Support for scientific advancements. For the university, using Windows HPC Server 2008 has opened up new possibilities in the fields of cultural and medical research. Its X-Ray Imaging Group is currently the only research group in the world capable of carrying out tomographic analyses on large objects. “We’re using Windows HPC Server 2008 to develop algorithms for applications that we could never have considered before with traditional data-processing tools,” says Levi. “We’ve also experienced a further advantage with the efficient recycling of code.”