The U.S. Energy Department's Argonne National Laboratory has commissioned IBM to build what will become one of the world's most powerful supercomputers, IBM announced Tuesday.
The computer, a model nicknamed Mira, will be capable of executing 10 quadrillion calculations per second, or 10 petaflops, and be operational in 2012, IBM said. It will be built on the next version of IBM's Blue Gene supercomputer architecture, called Blue Gene/Q.
Through funding from the National Science Foundation, Argonne will allow industry, academia and government researchers to access the computer for large-scale research and development projects, said Dave Turek, head of IBM's deep computing group.
Such computational muscle "should give rise to a lot of interesting innovations" across the fields of computational chemistry, aerodynamics, materials science, alternative energies and many other disciplines, he said.
In his annual State of the Union address last month, President Barack Obama highlighted the use of supercomputing as a way for the U.S. to maintain its competitive economic edge. He alluded to the progress that other countries, such as China, have made in building their own supercomputers.
The 10-petaflop performance far outstrips what is commonly thought of as today's most powerful supercomputer, the recently built Tianjin National Supercomputer Center's Tianhe-1A system, which benchmarked a performance of 2.67 petaflops for the last Top500 twice-annual ranking of the world's most powerful supercomputers.
The additional horsepower could be used to cut the time it takes to run computerized models, and even execute jobs that were previously too large for even the world's largest machines, IBM predicts. A model of how a human heart reacts to a medicine would take only two minutes to run on a 10-petaflop machine, compared to two years with today's top supercomputers. The speedier run times allow industry and academia to conduct more research to bring new products and innovations to market.
By 2012, Mira will be one of three U.S. IBM systems capable of 10 petaflops or greater. IBM is also ramping up production of another supercomputer for the Energy Department's Lawrence Livermore National Lab (LLNL), a 20-petaflop model to be called Sequoia. And it is also building the 10-petaflop Blue Waters system for the NSF-funded University of Illinois at Urbana-Champaign's National Center for Supercomputing Applications.
Mira will have more than 750,000 IBM PowerPC A2 1.6 Ghz, 16-way SMP (symmetric multiprocessing) A2 processors. Each compute node will have a single processor, and they will be housed in racks of 1,024. Each compute rack will also have from eight to 128 I/O nodes -- also running the A2 processor -- that will be dedicated to moving data on and off the compute nodes.
Each node will have either 8 or 16 gigabytes of memory, aggregating to 750 terabytes of memory across the entire system. Communications among the nodes will go over IBM 5D Torus interconnects, capable of 40 gigabits-per-second throughput.
For an operating system, the compute nodes will run a Compute Node Lightweight open-source scalable kernel, and the I/O nodes and the front-end and service nodes will run a modified version of Red Hat Enterprise Linux. The system will be mostly water-cooled and consume an average of 60 kilowatts per rack.
IBM did not reveal the price for Mira, though it did say Argonne had purchased it with funds from a $180 million grant.
While this system will be capable of 10 petaflops, the Blue Gene/Q architecture should be able to scale to 50 petaflops and perhaps more, Turek said.
IBM developed the Blue Gene architecture in 1993 as part of a US$100 million development effort in conjunction with LLNL. The research effort was aimed at building out a highly scalable yet energy-efficient architecture for large supercomputers.