News

Jackson State/U.S. Army Research Project Powers up Infiniband-Based Cluster

• By Dian Schaffhauser
• 04/13/09

Jackson State University in Mississippi has installed an SRC-7 cluster system in support of a joint research project between the university and the United States Army Engineer Research and Development Center (ERDC). The system being used in the research project is an Infiniband-based SRC-7 MAPstation cluster from SRC Computers. Each node provides the performance of hundreds of microprocessors; but, said the vendor in a statement, the overall system consumes 90 percent less power than a microprocessor-based cluster. It's capable of scaling to tens of thousands of nodes.

Khalid Abed, a faculty member in the Department of Computer Engineering at the university, is the principal investigator on this research effort, which deals with accelerating scientific applications with high performance reconfigurable computing (HPRC). "Microprocessor-based HPC clusters are facing formidable challenges such as run time performance, memory bottlenecks, floor space, and power dissipation. HPRC is emerging as a new research area that may provide solutions to some of these challenges. The SRC-7 system will allow us to investigate how to map complex parallel codes onto HPRC clusters."

This type of "reconfigurable supercomputer" allows researchers to reconfigure the hardware for different computing applications to achieve optimum performance, Abed said.

"A significant benefit of this system is the SRC Carte Programming Environment," said Gerald Morris, a computer scientist at the ERDC Department of Defense (DoD) Supercomputing Resource Center (DSRC), who is the Government technical lead on the project. "Carte allows the development of hardware-based computational kernels using either high-level languages like C or Fortran or hardware description languages like VHDL or Verilog."

Carte abstracts away some of the underlying design details, Carte explained. "This is particularly important when trying to map floating-point applications onto reconfigurable hardware, which is still a challenging research area."