SC16 – SQream Technologies, a GPU accelerated database startup from Israel, teaming up with Orange Silicon Valley, EchoStreams, and Caltech, and using the world’s highest density “Supercomputer-in-a-Box” (Klimax-210) from Cocolink (S. Korea), has developed a system deployed on the exhibit floor of Supercomputing 2016 (SC16) in Salt Lake City that runs SQream’s powerful GPU accelerated database on a compact 100 TeraFLOP 4 RU Supercomputer to load and process 20 TB/hr of science data.
By using the server platform’s capability of hosting up to 20 NVIDIA GPUs, and codes developed by Caltech researchers at the LHC, the team plans to demo and benchmark systems handling up to 50 TB/hr. This is a feat for a single database node that will prove to be extremely useful to Big Science programs such as the high energy physics experiments at the LHC, as well as other Big Data experiments in basic energy sciences, as well as genomics, climate, and other domain sciences.
High energy physics experiments, with their high volume, velocity, and variability of data already approaching an exabyte under management at sites around the world, are a unique platform to architect this new class of exascale architecture, and meet the future needs of data intensive sciences. The experiments at the LHC are instrumented with close to a hundred million electronic channels and are exposed to data rates of up to a petabyte per second. The data need to be carefully filtered in real time and analyzed to provide crucial knowledge on fundamental questions, including the nature of the dark matter that makes up 85% of the matter density in the universe, and the long-term stability of our universe.
“As we enter the era of the High-Luminosity LHC in the next decade, we will be in entirely new territory in terms of Big Data in science,” said Caltech’s Professor Harvey Newman, who leads a research program that uses high energy physics as a testbed for exascale ecosystem architectures of the future.
“SQream is excited to take on the world’s biggest data science challenges – exploring the basic ingredients of the universe, together with Orange, Caltech, and EchoStreams,” said Ami Gal, CEO and Co-Founder at SQream Technologies.
“We are pleased to be part of this collaboration with Caltech, Orange, and SQream. EchoStreams working closely with Cocolink provides a very High Performance Computing (HPC) Klimax-210, supporting up to 20 GPU in a 4U server all in the same PCIe root complex. This enables researchers to fully utilize all GPU computing capabilities without any compromise,” said Andy Lee, Director of Product Marketing at EchoStreams.
“We are working on providing solutions to Big Data processing challenges in science such as the ones of the largest science data repository in the world: the distributed data of the Large Hadron Collider program. Such science applications can be extremely valuable and the principles can be ported downstream to various industry verticals in the Telecom and Enterprise domain,” said Gabriel Sidhom, Vice President of Technology and Development at Orange Silicon Valley.
With Business Wire
© HPC Today 2019 - All rights reserved.
Thank you for reading HPC Today.