Titan Gets A Little Help
With the announcement of the 42nd TOP500 list at SC13 in Denver today, Titan, the 17.6 petaflop (Rmax) Cray XK7 system installed at the Department of Energy’s (DOE) Oak Ridge National Laboratory, held onto its number-two spot for another year, but in this era of big data, even the largest supercomputer in the United States could use a little help.
That’s why the Oak Ridge Leadership Computing Facility (OLCF) is installing two smaller machines, Eos and Rhea, whose sole function is to provide support for the center’s crown jewel. The new systems will be tasked with analyzing Titan’s massive data stream in order to enhance user experience and maximize project outputs, according to the OLCF.
Eos, which came online October 3, is the lab’s newest Cray system. The Cray XC30 cluster has 744 nodes housed in four cabinets. Each node packs two eight-core Intel Xeon E5-2670 (“Sandy Bridge”) processors and 64 gigabytes of memory. In total, the Eos partition contains 11,904 processor cores (23,808 logical cores with Hyper-Threading enabled), and 47.6 TB of memory. The compute cluster relies on the new Spider Lustre file system, with extra I/O bandwidth and storage capacity compared to the previous Spider system.
To support compatibility, Eos includes many of the same debuggers, profiling codes, and software packages that are used on Titan. The Cray cluster is spending its first few months supporting the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program. INCITE users get priority access on Eos through the rest of 2013.
OLCF is also welcoming a new Dell PowerEdge C600 cluster, called Rhea, which is replacing the aging Lens visualization cluster. Rhea’s 196 nodes, divided into four racks, each contain two eight-core Intel processors with Hyper-Threading and 64GB of main memory. With a full suite of post-processing software, including ParaView and iPython, Rhea is well-suited for analysis and visualization of simulation data generated by Titan.