NIH Biowulf Expands to 9000 Compute Cores

In June, 336 new compute nodes came online as part of the NIH Biowulf Cluster. The increase in compute power will be applied to a wide range of biomedical research applications including those in genomics, imaging, molecular dynamics, and statistical analysis.

The compute nodes

Each compute node consists of two Intel quad-core processors and 24 gigabytes (GB) of memory. At eight cores per node, the additional nodes add 2688 processing cores to Biowulf for a total of just under 9000 cores. In addition, sixteen of these new nodes have a larger memory complement of 72 GB for compute tasks requiring large memories, such as those common in sequence assembly and comparative genomics applications.

Issue of cost and power consumption

The competitive procurement for the nodes was begun in January and was awarded in April. The evaluation of proposals was based on application performance benchmarks and a total cost of ownership analysis based on initial cost and power consumption. In recent years, power consumption has become one of the largest cost factors for data centers housing scientific supercomputers. The system chosen was an IBM iDataPlex, which could provide more compute cycles for the same power utilization than the competition.

The Biowulf Cluster

At nearly 9000 processing cores, the NIH Biowulf Cluster is the largest single compute resource at the NIH and will help meet the ever increasing demand for computer power by biomedical research.

For more information about the Biowulf Cluster, please visit the Biowulf at NIH web page at http://biowulf.nih.gov/or view CIT's scientific computing resources web page at http://cit.nih.gov/Science/.

Back to top of page