This very interesting whitepaper explains how selecting a proper parallel file system for your application can increase the performance of complex simulations and reduce time to completion.
“The data that I presented from the Sanger Institute is typical of the profiles that we come across: a mix of good streaming IO (ie the larger reads), but unexpectedly high numbers of small reads and writes. These small reads and writes are potentially harmful to the file system. We’ve profiled HPC applications in various different life sciences organizations, not just the Sanger Institute, and we’ve found these IO patterns throughout. We’ve also seen similar IO patterns in EDA and oil and gas applications.”
Kyoto University Thinks Widening SIMD Will be Key to Performance Gains in New Intel Xeon Phi processor-based Cray System
“With an imminent switchover to a new Cray system with next-generation Intel Xeon Phi Processors (codenamed Knights Landing) planned for October, the ACCMS team at Kyoto University is eagerly looking forward to a potential two-fold application performance improvements from its new system. But the lab is also well aware that there is significant recoding work ahead before the promise of the new manycore technology can be realized.”
“Data Science and Information Systems researchers at UQ are tackling the challenges of big data, real-time analytics, data modeling and smart information use. The cutting-edge solutions developed at UQ will lead to user empowerment at an individual, corporate and societal level. Our researchers are making a sustained and influential contribution to the management, modeling, governance, integration, analysis and use of very large quantities of diverse and complex data in an interconnected world.”
DDN has opened a research and development center in Paris, in response to strong demand the French market and across Europe for high-performance big data storage solutions. DDN’s Newly Created Research and Development Centre, located in the Meudon area of Paris, will develop advanced end-to-end data lifecycle management technology, and a local technical presence for its European customer base – increasing scope for innovation and collaboration.
Today DDN announced it has increased its leadership as the dominant storage provider of the world’s fastest supercomputers for the eighth consecutive year with storage solutions that power more than 70 percent of identifiable sites on the most recent TOP500 list. To learn more, we caught up with the company’s Chief Marketing Officer, Molly Rector. “We’ve led the TOP500 list for eight consecutive years and continue to increase our leadership margin – most recently boosting our presence to an unprecedented 70% of the identifiable largest supercomputing sites worldwide.”
“With more than a decade of experience in designing, installing and supporting Lustre-based storage, DDN is the most experienced Lustre provider and has worked closely with us over many years to design optimized Lustre-based storage systems. DDN’s latest ES14K offering delivers a high-performing, high density appliance for the HPC market built on Intel Enterprise Edition for Lustre,” said Brent Gorda, GM of Intel’s High Performance Data Division.
This week, the Women in HPC organization announced a series of special events coming up at ISC 2016. To learn more, we caught up with WHPC Director Dr. Toni Collis from EPCC at the University of Edinburgh. “Most people don’t notice how un-diverse HPC really is. But when you start counting the number of women in the room, at the table, or in the C-suite, it is quite surprising.”
With ISC 2016 coming up in June, a number of ancillary events have been scheduled in Frankfurt to take advantage of this annual gathering of over 2500 supercomputing professionals. We’ve compiled a full listing for what looks to be an exciting week in the history of high performance computing.
Today DataDirect Networks announced DDN Flashscale, a new family of all-flash scale-out and scale-up storage solutions, which delivers the highest performance and capacity in a cost-optimized and feature rich platform designed for Enterprise Big Data and Analytics, Web Scale Cloud, and HPC environments. “DDN Flashscale’s fast embedded PCIe fabric delivers full native performance and extreme low latency from 48 NVMe or 72 SAS or any mix of SSDs while offering cost optimized sub $1/GB all-flash storage up to 576TB, 6 Million IOPS and 60GB/s per 4U node.”