NOTE: Carver will be decommissioned on August 31, 2015. Users are encouraged to begin migrating their Carver workloads to Edison. Please see Retirement Plans for more information.
Current Status: Up
Carver, named in honor of American scientist George Washington Carver, is an IBM iDataPlex cluster. Its unique features are nodes with relatively high memory per core, a batch system that supports long running jobs, and a rich set of third party software applications. Carver also serves as the interactive host for various experimental "testbeds", including the Dirac GPU cluster and the Hadoop testbed.
Carver, a liquid-cooled IBM iDataPlex system, has 1202 compute nodes (9,984 processor cores). This represents a theoretical peak performance of 106.5 Teraflops/sec. All nodes are interconnected by 4X QDR InfiniBand technology, providing 32 Gb/s of point-to-point bandwidth for high-performance message passing and I/O. Read More »
There are three compiler suites available on Carver: Portland Group (PGI), Intel, and GCC. The PGI compilers are the default. The MPI implementation on Carver is Open MPI, which is descended from LAM. In particular, note that Open MPI is not part of the MPICH family of MPI implementations. For each supported compiler suite, NERSC provides a version of Open MPI that is compatible with that compiler. Read More »
The Carver system will be retired at the end of August, 2015. Carver users should prepare for this by migrating their workload to one of the other NERSC systems, primarily Edison. This page is provided to assist users in that migration. Schedule Carver's final day of operation will be Monday, August 31, 2015. The precise timeline has not yet been determined. It is likely that the system will not be "drained"; that is, batch jobs will be allowed to run until the system is shut down. Read More »