Cutting-edge research and development (R&D) is a competitive differentiator for many organisations, allowing them to attract leading talent and solve some of the world's largest challenges. However, emerging scientific applications are consuming more and more technology resources, making computational performance, power and space limiting factors in the research equation.
HP is helping solve this problem by creating hyperscale computing systems that combine extreme levels of performance, power and cooling innovations to optimize energy consumption.
"Organisations across the board are struggling to keep pace with rapidly increasing demands for compute resources", stated Paul Santeler, vice president, Hyperscale Server Business Segment, HP. "The technologies driving the fastest supercomputers of today will come to bear on traditional data centres of tomorrow, making this a critical sector for both HP and our customers."
Combining breakthrough innovations from HP Labs - the company's central research arm - with more than 16 years of server market leadership, HP is helping customers like Tokyo Institute of Technology, BP, Purdue University, the U.S. Department of Energy's National Renewable Energy Laboratory and the Texas Advanced Computing Center define the next wave of hyperscale computing.
Tokyo Institute of Technology's TSUBAME 2.5 supercomputer ranked no. 11 on this year's TOP500 list.
TSUBAME 2.5 is a recent upgrade of the TSUBAME 2.0 system that entered the TOP500 at no. 4 in November 2010. TSUBAME 2.5 was created by upgrading the system's 1,408 HP ProLiant SL390s G7 Servers from NVIDIA Tesla M2050 GPUs to the latest NVIDIA Tesla K20x GPUs. The upgrade increased peak performance from 2.4 petaflops to 5.7 petaflops, and increased the system's LINPACK performance benchmark from 1.19 petaflops to 2.84petaflops. These performance improvements were achieved without increasing the power consumption or data center footprint of the system.
TSUBAME 2.5 allows the Tokyo Institute of Technology to speed scientific and research achievements by supporting applications in medical simulations, climate forecasting, earthquake and tsunami simulations, and computational fluid dynamics. Research on the TSUBAME 2.0 system won numerous awards, including the 2011 Gordon Bell Prize for Scalability and Time-to-Solution.
BP recently unveiled the largest supercomputing complex for commercial research in the world.
The supercomputing facility, based in Houston, will serve as a worldwide hub for BP to process and manage huge amounts of geologic and seismic data from across its portfolio, allowing BP scientists to produce clear images of rock structures deep underground and advance its global hunt for oil and natural gas. The system reached a peak processing capacity of 2.2 petaflops of peak performance of the overall computing capacity in the data center, leveraging more than 5,400 HP ProLiant SL230s Gen8 servers for extreme density, scale and energy efficiency.
HP collaborated with Purdue University to develop the Conte supercomputing cluster, the highest-ranking campus supercomputer on the TOP500 list, which placed at no. 33 overall.
Conte boasts approximately five times the sustained performance of the previous fastest university-owned supercomputer, Purdue's Carter system, while using only 87 percent of the number of servers.
The system performs at a sustained maximum speed of 943.38 teraflops and a peak performance of 1.342 petaflops. This speed would allow Conte to process problems approximately 15,000 times faster than a high-end consumer laptop.
The Conte supercomputer leverages 570 HP ProLiant SL250s Gen8 servers to support a wide variety of research programmes for the university, including chemical engineering, aeronautics, astronautics and planetary sciences.
HP collaborated with the U.S. Department of Energy's National Renewable Energy Laboratory (NREL) and Intel to develop a highly energy-efficient, warm-water liquid-cooled supercomputer dubbed Peregrine.
In addition to achieving a peak performance of 1.2 petaflops, Peregrine is housed in one of the most energy-efficient data centres in the world, with a power usage effectiveness (PUE) ranking of 1.06, compared to a national average of 1.91.
Peregrine leverages an innovative approach to cooling that was designed in collaboration with HP. The system uses warm liquid to capture heat generated by the servers; the excess heat is then used to heat NREL's Energy Systems Integration Facility, the nation's first major research facility focused on clean energy grid integration and wide-scale deployment.
NREL estimates that the system will save the organisation up to $1 million annually in server cooling and building heating costs.
The Texas Advanced Computing Center (TACC) at the University of Texas at Austin has worked with HP to design Maverick, an interactive, remote visualization and data analysis system that will enable users to make sense of large-scale scientific data.
Maverick will be ideal for Big Data analysis. Every node will have a large memory, state-of-the-art GPU accelerator, and be connected to massive data storage.
The system comprises 132 HP ProLiant SL250s compute nodes and 14 HP ProLiant DL360 and DL380 management, login and Lustre router servers, all connected by a Mellanox FDR InfiniBand interconnect to provide a high-performance platform.
Maverick will be fully deployed in January 2014.