This article is from the archives of the UB Reporter.

New Dell cluster nearly doubles CCR’s capacity

Published: July 21, 2005

Contributing Editor

In response to the soaring demand for computational power by the hundreds of researchers who depend on it, UB has expanded the computing capacity of the Center for Computational Research in its New York State Center of Excellence in Bioinformatics and Life Sciences with the installation of a new Dell high-performance computing cluster.

The cluster, with 1,668 processors, nearly doubles the Center of Excellence's computing capacity. If this cluster was listed on the current top 500 list of the world's fastest supercomputers, (, it would be among the 40 fastest individual machines.

The cluster was purchased with $2.3 million in federal funds appropriated for the Center of Excellence as a result of the efforts of Rep. Thomas M. Reynolds, Sen. Hillary Rodham Clinton and Sen. Charles E. Schumer.

"The Center of Excellence has been working diligently over the past year to establish the necessary programs and infrastructure to pursue quantum leap approaches in medical discovery," said Bruce A. Holm, executive director of the Center of Excellence and UB senior vice provost.

"This CCR upgrade will facilitate our statewide biomedical grid computing efforts and was an instrumental part of creating new scientific programs affiliated with the Center of Excellence, like the new $16.9 million structural genomics program that will span upstate New York and be housed in the Buffalo Life Sciences Complex in the Hauptman-Woodward Medical Research Institute. Needless to say, this new cluster also will be available to the larger UB community as well."

The new cluster will support the work of researchers affiliated with the Center of Excellence, as well as UB faculty members working in numerous areas ranging from chemistry to engineering, visualization to data mining.

"This installation will increase dramatically the pace of research and discovery at UB," said Russ Miller, director of CCR and UB Distinguished Professor in the Department of Computer Science and Engineering, School of Engineering and Applied Sciences.

"The expansion effectively will double our processing speed," he continued. "It will expand the capability of computing we can offer to our users, where large numbers of processors are used simultaneously to solve an individual problem, as well as our capacity computing, where the results of many independent, single-processor results are combined to solve a problem.

"This increase in throughput will enable researchers to undertake calculations at an entirely new scale," he said. "Researchers will be able to tackle larger problems, consider them in more detail and address problems that would have been intractable without such a system."

Even during the installation, Miller explained, the Dell cluster has proven to be very productive, as it was used extensively in a capacity fashion on a joint project with IBC Digital, a Buffalo animation company, in order to render a series of animated videos currently being shown on MTV2.

The Dell cluster's benefits to the center—and to society—will range as broadly as the fields of study at the university, from modeling combustion to tracing tiny amounts of contaminants in groundwater to genomic approaches to developing vaccines and worldwide monitoring of human rights abuses for peoples with disabilities.

One of the first projects that will be supported by the new Dell cluster will be a cutting-edge combination of demanding theoretical calculations and experimental studies designed to help understand how certain forms of cancer originate in the body. If successful, such an understanding could lead to improved methods to treat certain forms of cancer.

According to Miller, in today's society, research and discovery often are driven by simulation and modeling, which require high-end computing, including storage, networking, computing and visualization.

"To achieve this, it is critical to build a balanced high-end infrastructure that incorporates the efficient collection, organization, processing, visualization and distribution of data," he said.

"When analyzing our budget, the type of computing that would satisfy our users' needs in a cost-effective fashion and the quality of support and maintenance that we could expect, we chose a standards-based system from Dell that incorporates IBRIX and an EMC storage system, as well as interconnects from Force10 and Myricom," Miller said.

John Mullen, vice president for Dell's higher education business, said clusters have become the platform for a growing number of the fastest supercomputers in the world.

"More than 300 of the world's fastest supercomputers are using Intel processors, and most of those are labeled 'clusters,'" Mullen said. "Dell's supercomputing strategy has always centered on a standards-based, scalable enterprise architecture, which offers customers flexibility and the most computing power for their money."

The new Dell system has a theoretical peak performance of more than 10 trillion floating point operations per second (TFlops), with an anticipated sustained performance of more than 7 Tflops. The internal EMC storage system contains 30 TBytes of disk.

The installation brings the number of processors in CCR to nearly 7,000, extending the center's theoretical peak performance from approximately 12 Tflops to approximately 22 TFlops. (One floating point is the number of calculations a computer can perform each second.)

The new cluster is made up of Dell PowerEdge SC1425, 1850 and the award-winning Dell PowerEdge 2850 servers.

This high-end commodity-based Dell system consists of more than 1,600 of Intel's latest processors, packaged two processors per server, connected to each other by two networks: a high-speed (GigE) Ethernet network from Force10 and a state-of-the-art high-performance network from Myricom.

In addition, the system features a large internal storage network consisting of hardware from storage leader EMC and parallel file system software from IBRIX that is essential to accommodate the demanding file-transfer rates needed to support the cluster's computing power.