New Windows HPC cluster now available to Purdue researchers
February 8, 2012
Forestry and natural resources Professor Bryan Pijanowski’s research looks at how people and land use play into the state of ecosystems, including issues like projecting loss of prime farmland to urbanization over the next 50 years and the potential effects of climate change on fish habitats.
His lab examines such questions with computer modeling, among other techniques, sometimes on a national scale and at almost-unheard-of resolutions as fine as every 30 meters. The work can involve hundreds of gigabytes to terabytes of data to begin with, trillions of data points and software that runs — and in some cases will only run — in Microsoft Windows.
In the past, that could mean running a model for months on a desktop computer. But a new cluster supercomputer using the high-performance computing version of Windows, now available to Purdue researchers through the Rosen Center for Advanced Computing, has reduced the time it takes to get results to a few hours and is enabling more detailed and accurate modeling.
“We could actually run the entire world at 30-meter resolution if we wanted,” Pijanowski says. “What this has done, to be able to say we can do cutting-edge research at this scale, is put us at the head of the line.”
Faculty interested in using the new Windows HPC cluster can email email@example.com for more information.
ITaP built the Windows HPC cluster to make high-performance computing accessible to more campus researchers, along with additional research computing software.
“We want to serve everybody,” says Preston Smith, manager of research support for the Rosen Center, ITaP’s research computing unit.
The new cluster runs Windows HPC Server 2008 R2, the latest version of Microsoft’s operating system for high-performance computing. Windows HPC is an alternative to Linux-based clustering software packages, with similar features such as cluster management tools, a job scheduler and a message passing interface (MPI) library.
But while similar in functionality to Purdue’s other community cluster supercomputers, the Windows HPC cluster offers potential advantages for some users and developers.
Users get an entry point to high-performance computing with the familiar Windows user interface.
“Once you get it up and going, it makes a lot of sense,” says Jarrod Doucette, a GIS and database specialist for the Forestry and Natural Resources Department who works with Pijanowski.
The Windows HPC cluster makes high-performance computing available to researchers who rely on Windows-only software like ArcGIS and Microsoft Excel. In addition, the new cluster can run Windows HPC versions of research software common on Purdue’s Linux community clusters, such as MATLAB and SAS.
ArcGIS is a standard tool for Pijanowski and Doucette and a lot of the data they employ, from government agencies for instance, comes in ArcGIS format.
Another key tool is the Land Transformation Model Pijanowski began developing in the early 1990s. David Braun, an ITaP research computing specialist, worked closely with Pijanowski’s lab to migrate the model from desktop Windows to Windows HPC.
For developers, the Windows HPC cluster can work with familiar tools such as Microsoft’s .NET Framework, C# programming language and Visual Studio.
Forestry and natural resources Professor Patrick Zollner’s lab, with assistance from Braun, is porting models used to study the effects of landscape changes on animals to the Windows HPC cluster. Zollner, a wildlife scientist, looks at questions like the impact of traffic on maternity colonies of the endangered Indiana brown bat.
Zollner’s models run on Windows desktop computers now and, as for Pijanowski, the Windows HPC cluster promises to radically speed up the time to results and allow his lab to model in far greater detail. He says moving the code to one of the Linux-based community clusters would be no small matter.
“We would have to reinvent the wheel,” he says. “It would not be straightforward at all.”
Through community clustering, faculty partners and ITaP make more computing power available for Purdue research projects than faculty and campus units could afford individually. The Rosen Center installs, administers and maintains the community clusters, including security, software installation and user support, so researchers can concentrate on doing research rather than on running a high-performance computing system.
Community clustering also maximizes the use of resources by sharing computing power among faculty partners whenever it is idle. Researchers always have ready access to their share and potentially more if they need it.
Writer: Greg Kline, science and technology writer, Information Technology at Purdue, 765-494-8167 (office), 765-426-8545 (cell), firstname.lastname@example.org