The Federal University of Rio de Janeiro (UFRJ) has deployed a 226-teraflop SGI ICE X supercomputer to support engineering and research projects aimed at the Brazil’s energy industry. The system will serve both commercial and academic interests from its new home in the university’s High Performance Computing Center at the Graduate School of Research and Engineering (COPPE).
The new machine, which will go by the name of “Lobo Carneiro” (after Prof. Lobo Carneiro, who founded COPPE/UFRJ), is expected to increase the available FLOPS at UFRJ by a factor of two, along with a three-fold reduction in power consumption.
Its 226 teraflops are spread across 506 Intel Xeon E5-2670 v3 12-core processors, running at 2.3GHz. The Xeons will be housed in a 253 dual-socket servers and the whole system will contain 16 terabytes of memory. Although nearly a quarter of a petaflop, Lobo Carneiro won’t make the TOP500 list, whose last entry reaches 286 teraflops on Linpack.
Storage will be provided by three SGI InfiniteStorage 5600 systems, which will house 720 terabytes and deliver 17 GB/second of I/O. Intel’s Enterprise Edition of Lustre will be used as the file system.
The new system will support research and development for an array of scientific and engineering applications in oil and gas, wind energy and renewable biofuels. A good of its capacity is expected to be devoted to running seismic imaging models and computational fluid dynamics (CFD) simulations.
According to a blog posted by SGI’s Cori Pasinetti, the new supercomputer has its work cut out for it:
In oil and gas research, the simulation of turbidity currents, seismic imaging and oil spills are ongoing projects involving several academic and industrial partners. The SGI supercomputer will simulate exploration and production of energy, giving researchers a deeper understanding of the behavior of energy reserves. This information will help lead to more precise data to reduce costs associated with discovery, exploration and extraction. In addition, the new system will support international cooperation projects, such as the Brazil EU HPC for Energy, jointly developed with the Barcelona Supercomputing Center (BSC) in Spain.
The system will also be used to research renewable energy sources. COPPE/UFRJ is partnering with BSC on a three-year project, HPC4E, to help the industry address future energy demands and to reduce carbon footprints. The SGI ICE X supercomputer allows COPPE/UFRJ to more efficiently assess resources and respond to demand peaks and output predictions in wind energy. The system can also assist in the development of efficient biomass-derived fuels, reducing greenhouse gas emissions, hydrocarbon dependency and fuel costs.
Versatus HPC, an SGI partner based in Brazil, completed installation of the machine on July 13.