Next supercomputer is taking shape here, but will reside in Illinois.
Ravi Arimilli is the Austin-based head of global team
URBANA, Ill. — The world's next huge scientific supercomputer will have strong Austin roots, but it will make its home on the sprawling campus of the University of Illinois in the heart of the Prairie State's farm country.
The big machine, called Blue Waters, will be built by IBM Corp. with an extraordinary level of innovation. Much of the work is being done in Austin, where the computer's crucial brain chip, Power7, is being designed.
Ravi Arimilli is the Austin-based chief architect for Blue Waters, heading a global technical team of a few thousand engineers and software experts driving the $208 million project.
When it goes online in 2011, Blue Waters is expected to be 30 times more powerful than Ranger, the current scientific computing speed champion at the University of Texas. Ranger has roughly the power of 30,000 desktop computers; Blue Waters will equal 900,000.
Scientists already are lining up to get time on the new computer, which is expected to deliver big payoffs in such areas as advanced modeling of human cells and climate forces, and better ways of coupling hurricane models with ways of predicting onshore storm surge, according to officials at the University of Illinois.
IBM also plans to use the design as a foundation for other high-powered computers it will market to government agencies, aerospace and energy companies, and corporations that need intense computing power to handle mountains of data.
Arimilli's team and Illinois' National Center for Supercomputing Applications are working in a tight partnership to make sure that Blue Waters delivers on the promises made to the National Science Foundation, which is footing the bill.
Several universities competed for the project. Arimilli says he pushed for Illinois because it was where he saw the best chance of building the deep technical partnership that the project required.
IBM has had to share some of its most prized technical secrets with Illinois so the school can begin crafting new software to run on the computer.
"One thing that led us to Illinois was their culture," Arimilli said. "The model of trust and collaboration is so sincere on both sides."
The teams face a formidable workload over the next three years before they can claim victory with Blue Waters. "There will only be a medal to pin on when we are successful," said Thom Dunning, director of the National Center for Supercomputing Applications.
Scientists who are using UT's Ranger say the Austin computer is delivering research payoffs, some of which will be disclosed in formal research papers being prepared in the next few months. Blue Waters will provide an even more powerful tool for more complex research problems.
Klaus Schulten, a leading University of Illinois researcher in molecular dynamics, is making extensive use of Ranger to model how the microscopic components of cells interact to handle vital functions. He has already applied for access to Blue Waters, which, he said, "gets us closer to the mystery of life than we have every got before."
Arimilli drives his team with the vision of making computer history. Most of the computer industry pushes forward at the rate of the celebrated Moore's Law, in which performance doubles every 18 months.
But with Blue Waters, the goal is to prove his newly coined "Arimilli's Law," which holds that a dedicated and talented technical team that has five years and hefty financial support can improve performance by 100 times.
When Arimilli described his "law" to IBM's senior management, chief executive Sam Palmisano issued a challenge. If the law were true, he said, then the Blue Waters team ought to be able to accomplish another 100-times leap in performance in the five years after the Illinois computer is completed.
Arimilli agreed that IBM can accomplish the next great leap forward if the substantial money can be found to support the necessary intensive study and technical innovation.
To support its current round of technical innovations, IBM has depended on a $242 million research grant that it won in 2006 from the federal Defense Advanced Research Projects Agency to develop the hardware and software technology required for very high performance, easy-to-manage technical computers.
IBM has yet to release many technical details of Blue Waters. The computer contains thousands of high-powered Power7 processors, each of which has eight separate computing cores
The chip itself has extensive internal networking circuitry that speeds the interplay between the thousands of processors in the chip.
Along the way to building the computer, IBM had to invent a new networking and interconnect system for tying together different parts of the machine with very high-speed communications links.
"This is the dream of my professional lifetime," Arimilli said. With the huge challenge of completing Blue Waters "the next three years are spoken for."
Ravi Arimilli is the Austin-based head of global team
URBANA, Ill. — The world's next huge scientific supercomputer will have strong Austin roots, but it will make its home on the sprawling campus of the University of Illinois in the heart of the Prairie State's farm country.
The big machine, called Blue Waters, will be built by IBM Corp. with an extraordinary level of innovation. Much of the work is being done in Austin, where the computer's crucial brain chip, Power7, is being designed.
Ravi Arimilli is the Austin-based chief architect for Blue Waters, heading a global technical team of a few thousand engineers and software experts driving the $208 million project.
When it goes online in 2011, Blue Waters is expected to be 30 times more powerful than Ranger, the current scientific computing speed champion at the University of Texas. Ranger has roughly the power of 30,000 desktop computers; Blue Waters will equal 900,000.
Scientists already are lining up to get time on the new computer, which is expected to deliver big payoffs in such areas as advanced modeling of human cells and climate forces, and better ways of coupling hurricane models with ways of predicting onshore storm surge, according to officials at the University of Illinois.
IBM also plans to use the design as a foundation for other high-powered computers it will market to government agencies, aerospace and energy companies, and corporations that need intense computing power to handle mountains of data.
Arimilli's team and Illinois' National Center for Supercomputing Applications are working in a tight partnership to make sure that Blue Waters delivers on the promises made to the National Science Foundation, which is footing the bill.
Several universities competed for the project. Arimilli says he pushed for Illinois because it was where he saw the best chance of building the deep technical partnership that the project required.
IBM has had to share some of its most prized technical secrets with Illinois so the school can begin crafting new software to run on the computer.
"One thing that led us to Illinois was their culture," Arimilli said. "The model of trust and collaboration is so sincere on both sides."
The teams face a formidable workload over the next three years before they can claim victory with Blue Waters. "There will only be a medal to pin on when we are successful," said Thom Dunning, director of the National Center for Supercomputing Applications.
Scientists who are using UT's Ranger say the Austin computer is delivering research payoffs, some of which will be disclosed in formal research papers being prepared in the next few months. Blue Waters will provide an even more powerful tool for more complex research problems.
Klaus Schulten, a leading University of Illinois researcher in molecular dynamics, is making extensive use of Ranger to model how the microscopic components of cells interact to handle vital functions. He has already applied for access to Blue Waters, which, he said, "gets us closer to the mystery of life than we have every got before."
Arimilli drives his team with the vision of making computer history. Most of the computer industry pushes forward at the rate of the celebrated Moore's Law, in which performance doubles every 18 months.
But with Blue Waters, the goal is to prove his newly coined "Arimilli's Law," which holds that a dedicated and talented technical team that has five years and hefty financial support can improve performance by 100 times.
When Arimilli described his "law" to IBM's senior management, chief executive Sam Palmisano issued a challenge. If the law were true, he said, then the Blue Waters team ought to be able to accomplish another 100-times leap in performance in the five years after the Illinois computer is completed.
Arimilli agreed that IBM can accomplish the next great leap forward if the substantial money can be found to support the necessary intensive study and technical innovation.
To support its current round of technical innovations, IBM has depended on a $242 million research grant that it won in 2006 from the federal Defense Advanced Research Projects Agency to develop the hardware and software technology required for very high performance, easy-to-manage technical computers.
IBM has yet to release many technical details of Blue Waters. The computer contains thousands of high-powered Power7 processors, each of which has eight separate computing cores
The chip itself has extensive internal networking circuitry that speeds the interplay between the thousands of processors in the chip.
Along the way to building the computer, IBM had to invent a new networking and interconnect system for tying together different parts of the machine with very high-speed communications links.
"This is the dream of my professional lifetime," Arimilli said. With the huge challenge of completing Blue Waters "the next three years are spoken for."
Comments