The Brazos Cluster at Texas A&M University
Brazos, a major computing cluster at Texas A&M University, is designed to meet the high-throughput computing needs of A&M's computational scientists and engineers. Though capable of executing modest MPI applications, Brazos is optimized for handling large numbers of single-node computations. This form of computing is referred to as high-throughput or capacity computing.
The computing power of Brazos comes from 309 computing nodes, with processors ranging from quad core Intel Xeon (Harpertown) and AMD Opteron (Shanghai), to 8-core AMD Opteron (Bulldozer) with 16GB to 128GB per node. Total peak performance is about 31.3 TFlops with a total of 10.1TB of RAM.
Access to Brazos is via a login nodes load balanced using round-robin DNS. User home directories are supported by a 5TB NFS file system. Data storage is supported using the Fraunhofer Filesystem on a 241TB storage array running on 7 storage nodes. Operating software for Brazos includes the Linux operating system, GNU and Intel compilers, SLURM batch scheduler, several MPI and linear algebra packages, and numerous applications.
The compute nodes and servers of Brazos are connected internally via a modular switch, with Gigabit Ethernet connections to each compute node and 10GbE connections to the login node and the data fileservers. The login nodes are connected to the Science DMZ network with 10GbE. The networking fabric for a large portion of the Brazos cluster is DDR Infiniband.
Funding for Brazos comes primarily from participating stakeholders, faculty from several (initially four) colleges. In addition to managing Brazos, the Academy itself also functions as a stakeholder. The initial stakeholders were:
- College of Agriculture and Life Sciences, Jim Woolley of Entomology, Mariana Mateos of Wildlife Science, and Robert Washington-Allen of Ecosystem Science and Management
- Dwight Look College of Engineering, Akhil Datta-Gupta and Behnam Jafarpour of Petroleum Engineering, and Tiffani Williams of Computer Science
- College of Geosciences, John Nielsen-Gammon of Atmospheric Sciences
- College of Science, Jean-Luc Guermond, Wolfgang Bangerth, and Guido Kanschat of Mathematics
Additional funding from the Vice President for Research supported base infrastructure (e.g., racks and file servers). Support from CIS provides 10GbE networking to the campus network and other operational support such as machine room space, electrical power, and air conditioning. As new stakeholders join and as existing stakeholders increase their contributions, we expect the capacity of Brazos to grow.
Funding sources include: the National Science Foundation, including awards CBET-0736202, DEB-0730616, and DMS-0922866; the Texas Commission on Environmental Quality; Texas A&M University, including the Office of the Vice President for Research, Computer Access Fees, and faculty from the Colleges of Science, Agriculture and Life Sciences, Engineering, and Geosciences; Texas AgriLife Research; and Texas Engineering Experiment Station.