Transforming computing

01.10.12 -

We've had several challenging months at NCSA, but what a comeback. Blue Waters is firmly back on track.

Our new partners at Cray will begin delivering a system in the next several weeks. It will be based on their XE6 hardware and the next generation of their XK6 hardware.

Science and engineering researchers will be able to begin work on a 1.4 petaflops early science system by early 2012, nearly doubling the computing capacity in the National Science Foundation's supercomputing centers. By mid 2012, the entire Cray system will be in place.

The Blue Waters system that Cray will put in the National Petascale Computing Facility will be a CPU-GPU hybrid. With its 8.3 petaflops of CPU capability, it will bring sustained-petaflop performance to a broad range of science and engineering applications in fields like climate change, the spread of epidemics, earthquakes, fundamental chemistry and physics, and materials science. And, with its 4.3 petaflops of GPU capability, it will serve as a bridge to the technologies on which future supercomputers will be based.

The capability and performance of the Cray system is just as relevant today as it was the first time we sat down to plan the project. Our colleagues conducting open scientific research need a system dedicated to the largest, most complex problems that can be simulated today. They'll get that with Blue Waters. Blue Waters has been configured as a balanced computing system, able to handle the most compute-, memory- and data-intensive problems in science and engineering.

It's been several months of work to bring this new plan into focus. But a few things stood out while the plan was still emerging that were absolutely essential to our success then and will continue to be essential as we move forward.

First among those is the NCSA staff—knowledgeable, experienced, and dedicated. All organizations do well when times are good. What distinguishes an outstanding organization is how well it does when times are uncertain. I am extremely proud of the way that our staff handled themselves during these trying times—they kept their focus on the goal and brought years of experience and expertise to bear to ensure the right outcome for the science and engineering community.

They had the know-how to right the project, and they have the know-how to see that the project lives up to its fullest potential. We have everything we need to ensure that the researchers who use Blue Waters will have an outstanding computing system and technical support. With this, they will be able to take full advantage of the extraordinarily capabilities that Blue Waters provides.

Cray and its impressive staff were also crucial. They entered a tricky situation with great resolve, and have been an able partner from the outset. To say they hit the ground running is an understatement. Their commitment is obvious and greatly valued.

The entire team—at NCSA, Illinois, Cray, and our academic partners throughout the country who are part of the Great Lakes Consortium for Petascale Computation—understands the core principal behind Blue Waters. That is: It's more than just a hardware acquisition, more than just a chase for a high peak performance number or a ranking on the Top 500 list.

Instead, Blue Waters is a project that was focused from the beginning on transforming the role of computing in science and engineering. The external networking for the system is up and running, and the archive system will follow soon. The science teams who will use Blue Waters were already working with NCSA staff to prepare their codes to run at the scale required. That work is continuing unabated and has been expanded to embrace GPUs. Similarly, the Blue Waters education program has been training hundreds of students in the skills they need to exploit extreme-scale computing systems—including GPUs.

This mindset allowed us all to evaluate our options and move quickly to select a new hardware solution with Cray. The Cray hybrid system will allow NCSA to serve the national science and engineering community as it has for the last 25 years—driving new discoveries and improving our world.

These are lofty goals, and we are keenly aware that they come with a certain amount of risk. But they are also timeless and vital and exciting.

These are the kinds of risks NCSA has taken throughout our 25-year history. These are the kinds of risks we will continue to take, and the kind of goals we will continue to pursue.

Thom Dunning