The Week in HPC News
Happy Fourth of July holiday to all of our American readers….it’s a short week on this side of the pond and for the HPC community, a slightly quieter one following the influx of news from the International Supercomputing Conference, which just rounded out in Germany. On that note, we look forward to returning next year at the new site in Frankfurt—a location that promises to be a bit simpler to get to, settle in, and enjoy.
We were able to spend quite a bit of time talking with many from the community at the show and will be bringing more stories from our conversations and a few key presentations in the weeks ahead. Although ISC week with the Top 500 announcements and related items from the vendor and end user community is past, there were still some lingering bits of note. In particular, the full results of the Green 500 rankings were released. If you missed it, there’s some solid analysis here.
This week we dug more in depth on the emerging companion benchmark called HPCG, which produced some early results to compare with Linpack. We also talked with the newly announced executive director at the Texas Advanced Supercomputing Center, Dan Stanzione about the diverse stable of systems they’re designing and upgrading, took a close look at one of the more vocal newsmakers at ISC this year, Bull, and their new approach to the market, in addition to many other stories.
For now, let’s review just a couple of stories from the week’s news that stood out:
This Week’s Top News Items
A consortium of companies including Arista Networks, Broadcom Corporation, Google, Mellanox, and Microsoft have announced the availability of a specification optimized to allow data center networks to run over a 25 or 50 Gigabit per second (Gbps) Ethernet link protocol. This new specification will enable the cost-efficient scaling of network bandwidth delivered to server and storage endpoints in next-generation cloud infrastructure, where workloads are expected to surpass the capacity of 10 or 40 Gbps Ethernet links deployed today.
The goal is to create an industry-standard, interoperable Ethernet specification that boosts the performance and slashes the interconnect cost per Gbps between the server Network Interface Controller (NIC) and Top-of-Rack (ToR) switch. The specification adopted by the Consortium prescribes a single-lane 25 Gbps Ethernet and dual-lane 50 Gbps Ethernet link protocol, enabling up to 2.5X higher performance per physical lane or twinax copper wire between the rack endpoint and switch compared to current 10 Gbps and 40 Gbps Ethernet links.
Oil and gas company Eni has just fired up a new large-scale HPC cluster, this time selecting an 1500 IBM iDataPlex dx360 M4 nodes, built on more than 30,000 processing cores, each equipped with two NVIDIA GPUs. This combination offers up 3 petaflops of performance, coupled with 7.5 petabytes on the storage end. The new HPC cluster is working alongside with the first GPU computing system installed last year (1500 IBM iDataPlex nodes and 1300 NVIDIA Tesla GPU accelerators).
Dell has announced that the International Center for Biosaline Agriculture (ICBA) in Dubai has deployed a new HPC system that will support analyses of the impact of climate variability and change on water and agriculture in the Middle East & North Africa (MENA) region. While system details were not made public, Dell says as part of ICBA’s Modeling and Monitoring Agriculture and Water Resources Development) funded by the U.S. Agency for International Development, ICBA is collaborating with NASA’s Goddard Space Flight Centre and selected U.S. university experts to use the Weather Research and Forecasting system to downscale climate data at regional and local scale.
Again, a quieter than usual week, but as always, we’ll be back again after the holiday with more as it develops.