DARPA taps IBM, Cray for supercomputing work

IBM Corp. and Cray Inc. landed Defense Advanced Research Projects Agency contracts to develop supercomputers that are faster than today's machines and easier to program.

Big Blue's award is worth $244 million, while Cray's contract is worth $250 million. Both contracts are for four years.

IBM of Armonk, N.Y., will develop a machine that provides 100 times the sustained performance of today's general-purpose supercomputers and is simpler to program, administer and use.

The project will increase research and development activities into mainline IBM technologies, such as an upcoming generation of the Power processor, the AIX operating system, IBM's General Parallel File System and IBM's Interconnect and Storage Subsystems. These technologies are planned for delivery in 2010 and beyond.

"We believe this new system will accelerate scientific breakthroughs, improve our nation's competitiveness and create new market opportunities," said Bill Zeitler, senior vice president, IBM Systems and Technology Group.

Cray of Seattle will develop a new supercomputer based on the company's Adaptive Supercomputing vision, a phased approach to hybrid computing that integrates a range of processing technologies into a single scalable platform.

Cray's development program, code-named Cascade, will produce a system with programmability, portability and robustness that is capable of scaling to unprecedented levels of sustained performance on real applications.

Under the Cascade program, Cray will develop a new hybrid system architecture that combines multiple processor technologies, a new high-performance network and an adaptive software layer into a single integrated system. Designed to efficiently scale to large numbers of processors, the system will maximize productivity and performance by automatically matching the most effective processor technology to each application.

One of the most significant barriers to sustained performance of more than a thousand trillion calculations per second, known as petascale computing, is achieving the scalability of the hardware and software across a broad set of existing applications. The goal of the technologies delivered under IBM's contract is to allow a wide spectrum of current applications and programming styles to cross the multi-petascale barrier in sustained performance.

Over the course of the contract, Cray will incorporate elements of the Cascade program into commercially available products, including the peak-petaflops supercomputer, code-named Baker, which will be delivered to the Energy Department's Oak Ridge National Laboratory. In addition, the laboratory will be one of Cray's Phase III partners, focused around scaling from both the systems perspective and the performance of key applications.

Developing computing systems that are more usable by the national security community, science and industry is a key part of DARPA's High Productivity Computing Systems mission. IBM plans to tackle this productivity challenge through an end-to-end holistic approach to advanced system architecture and design as well as software development in the following areas: operating systems, programming models, compilers, libraries, file systems, application development tools, performance tools, systems and data management and serviceability.

IBM, which has more than 329,000 employees and had annual sales of $96.3 billion in fiscal 2005, ranks No. 17 on Washington Technology's 2006 Top 100 list the largest federal IT contractors.

Originally posted at 11:25 a.m. and updated at 6:18 p.m.