TRACC’s Massively Parallel Computer Delivered; Acceptance Testing Underway

February 5, 2008

The massively parallel computer that is central to Argonne National Laboratory’s Transportation Research and Analysis Computing Center (TRACC), was delivered January 14, 2008, and was operational the following day. System and acceptance testing are now complete.

The key computing hardware component of TRACC is the computational cluster, a 512-core, customized LS-1 system from Linux Networx, consisting of 128 compute nodes, each with two dual-core AMD 2216 Opteron CPUs and 4 gigabytes of RAM.

The comprehensive test plan consisted of several phases:

• In the functional test phase, Linux Networx and Argonne staff tested the system’s basics to demonstrate that its hardware and software are operating and configured correctly.
• In the stability test phase, the entire cluster ran the HPL Linpack benchmark for 48 hours to ensure that the system remains operational under a continuous high load.
• In the performance test phase, standard test cases of the LS-DYNA®, STAR-CD, and TRANSIMS application codes, as well as low-level benchmarks of the cluster’s memory, storage, and networking subsystems, were run.