Supercomputing
7 of the top 10, 28 of the top 50 and 48 of the top 100 supercomputer sites, including Argonne National Laboratory, Lawrence Livermore National Laboratory, NASA, Oak Ridge National Laboratory, Sandia National Laboratories, San Diego Supercomputer Center, Electricite de France and the Center for Computational Sciences at the University of Tsukuba, rely on S2A Extreme Storage for the highest throughput and bottleneck-free performance to keep data pumping to their petaflop and teraflop computing engines.
Basic and applied research requires the highest levels of computational power to find solutions for the challenges facing the world today. Advanced computing environments at national, academic, and corporate laboratories are enabling scientists and engineers to collaborate in the study of natural phenomenon at all levels from sub-atomic particles, dark matter, and molecular dynamics to climate change, weather forecasting, energy, biomedical, as well as to the development of technologies for long-life battery cells or space exploration.
Leveraging the power of thousands of clustered processing nodes, these advanced computing environments use, process, and generate massive amounts of data in high-speed simulations, analysis, and visualization, placing extreme demands on storage systems. Conventional, general purpose storage systems, designed for transaction processing applications, run applications sequentially, cannot keep data pumping fast enough and become bottlenecks costing more time, more money and more management personnel.
In contrast, DataDirect Networks' eighth generation S2A platform, the industry-leading S2A9900 enables multiple applications to run simultaneously in parallel and provides extreme performance and capacity, managing 2.4 petabytes in only two floor tiles and delivering sustained throughput of up to 6 gigabytes per second for both writes and reads, per appliance. Additionally, the S2A9900 is the only storage platform that writes data as fast as it reads it, with guaranteed QoS, which translates into consistently predictable performance for large content files--critical for supercomputing environments.
Highlights
- Delivers extreme performance beyond 240 GB/s and scalable capacity
- Supports open systems CPU, OS, and file systems
- Shortens checkpoint cycle times
- Provides cost-effective, high-speed "scratch" storage
- Saves space and reduces power for online archives
Industry-Leading Throughput
With throughput in excess of 240GB/s and no disk write performance penalty, the S2A platform is able to deliver consistent high performance even at peak demands. As the core operating system of the S2A, DirectOS features the real-time DirectRAID Engine. This purpose-built system leverages high-speed FPGA parity calculation accelerators to deliver optimum large file data throughput and write performance that is equal to the S2A's read performance.
Unbeatable File System Performance
File systems leverage the S2A architecture to deliver higher, sustained performance levels across clustered storage architectures. DDN delivered one of the world's largest and fastest file systems at Oakridge National Laboratory. Codenamed "Spider," the system is designed to deliver up to 240 gigabytes per second of single Lustre file system performance and support multiple supercomputers as the underlying site-wide storage system for computational science. Spider is roughly twice as fast as the previous performance leader and stores over 10 petabytes of research data, equivalent to 1,000 copies of the printed contents in the U.S. Library of Congress.
Save on Space, Power, Cooling
The rapid growth of simulation data in supercomputing environments requires scalable, cost-effective storage solutions to backup and archive scientific data. The superior high-density storage array configurations from DataDirect Networks hold as many as 1,200 drives in two racks (up to 75TB/ft2) while increasing backup and archive application performance.