Powerful Blue Gene/P supercomputer at Argonne to address most-challenging science problems

November 9, 2007

ARGONNE, Ill. — One of the world's fastest supercomputers will soon reside at the Argonne Leadership Computing Facility at the U.S. Department of Energy's Argonne National Laboratory, thanks to a recently completed contract for the acquisition of a 445-teraflops IBM Blue Gene/P cluster.

The ALCF's second major acquisition, the Blue Gene/P supercomputer will boost the facility's total computing power to 556 teraflops, representing a fivefold increase in system capability. This advance will help to initiate the coming era of petascale computing and enable experts to answer questions that have confounded America's scientists for years.

"The ALCF has been a valuable contributor in the development of Blue Gene/P," said Leo Suarez, head of deep computing at IBM. "The close working relationship that we enjoy will deliver a machine that will propel scientific discovery in the most profound way since Galileo's telescope."

The ALCF Blue Gene/P system relies on technology provided by IBM, Myricom and DataDirect Networks in order to pair world-class processing speeds with advanced data management capabilities to meet the intense computational strength and data-storage demands of petascale computing. "Researchers can employ this new computing resource to attack cutting-edge problems in science and engineering at unprecedented scale and speed,” said Ray Bair, Argonne's ALCF director.

The Blue Gene/P can carry out 445 trillion calculations – or flops (short for floating-point operations) – every second. If all six billion people on Earth tried to work as fast as the supercomputer, each person would need to do more than 70,000 additions or multiplications per second to keep up with it. 

The new Blue Gene/P will open to researchers entire realms of scientific inquiry, from in-depth modeling of climate change to more accurate and complete representations of exploding supernovas. The complex computations needed to produce the fine resolution of these models create tremendous volumes of data, from tens to hundreds of terabytes per run. 

The ALCF provides resources for the U.S. Department of Energy's Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program, which seeks computationally intensive research projects from industry, scientific researchers and research organizations. The 445-teraflops Blue Gene/P will dramatically increase ALCF resources available to INCITE.

INCITE computations will commonly produce tens to hundreds of terabytes of data in a single day. "This architecture enables our Blue Gene/P systems to support the gigantic data flows of petascale applications," Bair said, "like the time series data that come from biomolecular dynamics, climate models and astrophysics simulations, to name just a few."

Because the Blue Gene/P can execute so many computations so quickly, the ALCF also requires a state-of-the-art data-storage system. This system will take the form of a bank of more than 8,000 disk drives that will send and receive data from the Blue Gene/P's more than 100,000 processors. Altogether, this system can deliver nearly 80 billion bytes per second to and from disk – the equivalent of transferring the content of 100 full CDs every second.

Myricom's economical, low-latency modular switches represent the heart of the ALCF's data-management system. The nine-switch complex supports up to 2,048 connections, each of which simultaneously exchanges data at around 1 billion bytes per second.

IBM will also supply 68 file servers to run Argonne's high-performance Parallel Virtual File System (PVFS2). This bank of servers will manage more than 8 petabytes of storage on 17 next-generation DataDirect Networks' massively scaleable S2A storage appliances, which can deliver more than 78 gigabytes per second of sustained throughput from the array. “Our design can enable the IBM Blue Gene/P to maximize computational time instead of waiting for I/O operations,” said Alex Bouzari, CEO of DataDirect Networks.