IBM and GENCI, France's high performance computing agency, today announced a collaboration aimed at speeding up the path to exascale computing—the ability of a computing system to perform at least 1 exaflop, or 1 billion billion calculations, in 1 second.
The collaboration, planned to run for at least 18 months, will focus on preparing complex scientific applications for systems that are under development and expected to achieve more than 100 petaflops—a solid step forward on the path to exascale, IBM said.
Working closely with supercomputing experts from IBM, GENCI will use the high performance computing technologies from the rapidly expanding OpenPOWER ecosystem. Supported by more than 140 OpenPOWER Foundation members and thousands of developers worldwide, the OpenPOWER ecosystem includes a variety of computing solutions that use IBM's open, licensable POWER processor technology.
Currently, the fastest systems in the world perform between 10 and 33 petaflops, or 10 million billion to 33 million billion calculations per second—roughly 1 to 3 percent the speed of exascale, IBM said. Put into context, if exascale computing is the equivalent of an automobile reaching 1,000 miles per hour, today's fastest systems are running within a range between 10 and 33 miles per hour, IBM said.
As part of the collaboration, GENCI will look into the impact and requirements of the OpenPOWER architecture on scientific applications to gain a deeper understanding of application requirements as the computing industry advances toward exascale computing. Special emphasis will be placed on accelerator technologies.
"The work we are doing with GENCI—bringing together some of the best minds in science and information technology—is a collaborative effort on a grand scale involving not just GENCI and IBM, but thousands of developers contributing to the rapidly expanding OpenPOWER ecosystem worldwide," said Michel Teyssedre, CTO of IBM France, in a statement. "We fully expect our collaborative efforts will produce innovations capable of moving the supercomputing industry that much closer to exascale."
Moreover, the collaboration will take advantage of OpenPOWER-based innovations such as the connection of Nvidia GPU accelerators to POWER processors through the high-speed Nvidia NVLink interconnect, as well as how Mellanox EDR 100Gb/s InfiniBand switches can exploit IBM's Coherent Application Processor Interface (CAPI) to improve solution performance.
In addition, IBM scientists along with experts from GENCI and French research organizations plan to work on forwarding the evolution of programming models, considering MPI and OpenMP as a first step for shared memory multiprocessing programming. Alternative application program interfaces will also be considered, given potential changes may be required as systems move closer toward exascale.
"If we want to continue to address the challenges of the French scientists and engineers, we need to anticipate the rise of new high performance computing architectures that bring us closer to exascale and prepare our communities," said Catherine Riviere, CEO of GENCI, in a statement.
IBM will provide dedicated technical experts to support application porting and optimization efforts as well as organizing along with GENCI education and porting sessions. IBM also will support this collaboration with its newly created POWER Acceleration and Design Center in Montpellier as part of the partnership established with both Nvidia and Mellanox. The center will provide technical expertise around scientific applications, programming models and systems, as well as early access to forthcoming 2016 platforms, the IBM high performance computing (HPC) software stack and the Nvidia Tesla Accelerated Computing Platform.