Titan Gets Computational CompanyJennifer Brouner
November 01, 2013
Eos, a Cray XC30 system has 744 nodes divided among four cabinets, is one of the OLCF’s newest computing resources.
OLCF gains new systems and expands the resources available at the center
Two new computing resources are being installed at the Oak Ridge Leadership Computing Facility (OLCF) to enhance the user experience and maximize project outputs.
The new machines will help scientists consolidate the massive amount of data gathered from Titan, allowing researchers to find answers to some of the world’s most puzzling problems.
Eos: OLCF’s newest Cray system
Eos came online October 3 and will initially be prioritized as an additional Innovative and Novel Computational Impact on Theory and Experiment (INCITE) resource for the rest of 2013.
The Cray XC30 system has 744 nodes divided among four cabinets. Each node contains two eight-core Intel Sandy Bridge processors and 64 gigabytes of memory. The compute cluster will be the first machine available to users that relies on the new Spider Lustre file system, which has more I/O bandwidth and storage capacity than the current Spider system. Eos has many of the same debuggers, profiling codes, and software packages that are used on Titan.
“Eos is the newest generation of Cray architecture,” said Suzanne Parete-Koon of the OLCF User Assistance and Outreach group. “The machine initially will be used as an extra computing resource to help INCITE projects reach their goals.”
Rhea: Out with the old
Beginning in November, OLCF users can access a new compute cluster called Rhea that is replacing the aging Lens visualization cluster. The system—which is composed of 196 Dell PowerEdge C600 nodes—offers users a newer resource for post-processing analysis and visualization of information gleaned from Titan.
The nodes—four of which are login nodes—are stored in four racks. Each compute node contains two eight-core Intel processors and 64 gigabytes of memory. Rhea offers a full suite of post-processing software, including ParaView and iPython.
“We are excited about bringing Rhea online so we can offer our users a dedicated analysis and post-processing resource that features the latest commodity hardware,” said Robert D. French, also of the User Assistance and Outreach group. “We expect it to offer a faster time to solution for our users.”
Upon successful deployment of Rhea, the Lens visualization cluster will be retired.