UMBC High Performance Computing Facility
High Performance Computing @ UMBC
The following high performance computers have been operated by the
UMBC High Performance Computing Facility:
tara.rs.umbc.edu is the current cluster
in the UMBC High Performance Computing Facility (HPCF).
Installed in Fall 2009, tara is an 86-node distributed-memory cluster
with two quad-core Intel Nehalem processors and 24 GB per node,
an InfiniBand interconnect, and 160 TB central storage.
For a detailed description and more information
see resources for tara users.
The purchase of tara was supported by faculty contributions,
by grants from the National Science Foundation, and by the UMBC administration.
Researchers can purchase nodes for long-term priority access.
If you are interested in joining the effort, see
the contact page.
hpc.rs.umbc.edu was a distributed-memory cluster
with 33 compute nodes plus 1 development and 1 combined user/management node,
each equipped with two dual-core AMD
Opteron processors and at least 13 GB of memory,
connected by an InfiniBand network
and with an Infiniband-accessible 14 TB parallel file system. The
purchase in 2008 pooled funds from several researchers with seed funding
from UMBC. This machine was replaced by tara in 2009.
For the historical record of hpc and the projects that used it, see the
archived hpc webpage.
kali.math.umbc.edu was a distributed-memory cluster
with 32 compute nodes including a storage node connecting the 0.5 TB
central storage, each with two Intel Xeon 2.0 GHz
processors and 1 GB of memory, connected by a Myrinet interconnect,
plus 1 combined user/management node.
This machine was purchased in 2003 by researchers
from the Department of Mathematics and Statistics
using a grant from the National Science Foundation.
It became part of HPCF in 2008 and was shut down in early 2009
after the cluster hpc had become operational.
For the historical record of kali and the projects that used it, see the
Other high performance computers available on the UMBC campus are:
hercules.rs.umbc.edu is a shared-memory computer with 8 Intel Xeon 2.0 GHz and 24 GB total memory maintained by UMBC
Office of Information Technology (OIT). To obtain an account on the machine, file a ticket with the OIT Helpdesk or mail
email@example.com. hercules does not have any scheduling software, so users must monitor system activity themselves (e.g.,
using the command "top") before starting jobs. The machine can run MATLAB and other OIT-licenced software with large memory usage.
bluegrit.cs.umbc.edu is a distributed-memory cluster with 47 blades using IBM PowerPC chips (33 blades with 0.5
GB memory and 14 dual-processor blades with 2 GB memory) and 12 blades with two IBM Cell Processors and 0.5 GB memory. See the
bluegrit webpage for more information. This machine is part of the
Multicore Computational Center (MC2) at UMBC.