High-Performance Computing and Communications

Among supercomputers in an academic setting, HPCC's cluster is the 22nd fastest in the world and the 10th fastest in the United States. Among all supercomputers in the world, it is ranked 63rd. HPCC's cluster claimed these distinctions by achieving a benchmark in August 2007 of 17.11 teraflops, or 17.11 trillion floating-point calculations per second.

HPCC comprises a diverse mix of computing and data resources. The principal computing resources are a Linux cluster supercomputer and Sun x4600 large memory systems. In addition, HPCC has a central facility that provides more than 325 terabytes of combined disk storage and potential access to nearly a petabyte of tape storage, as well as a Condor cluster that uses spare cycles on Unix workstations in USC's public user rooms.

Upcoming HPCC Downtimes

HPCC Research		HPCC System Updates
Thomas Jordan Using seismic tomography to help reduce earthquake risks More... Thomas Katsouleas High-energy physics demands high-power computing More... Carl Kesselman Middleware guru and grid pioneer More... Kevin Knight and Daniel Marcu Statistical approach to natural language translation More... Aiichiro Nakano and Ashish Sharma Large-scale scientific visualization More... Priya Vashishta Multi-teraflop supercomputing More... Arieh Warshel Computer simulation of biological processes More... Richard Weinberg Computer animation and visual effects More...		Mpic-mx build The cluster is currently running Mpich-mx. MPI programs compiled with mpic-bm must be recompiled with updated mpich-mx build. To see if you binaries are affected, run 'ldd $executable|grep libgm'. If "libgm" is printed, then the binary will not work without recompiling. /usr/usc/mpich/default/mx-gnu34 /usr/usc/mpich/default/mx-intel /usr/usc/mpich/default/mx-gnu4 /usr/usc/mpich/default/mx-pgi pbstop Pbstop is installed on hpc headnodes, hpc-login1 and hpc-login2 (/usr/bin/pbstop). To learn about its features, visit the manpage or help screen (type h).