Final-phase success celebrated
Laboratory officials celebrated Los Alamos’s decision to pursue, pending approval by the National Nuclear Security Administration, the final phase of Roadrunner, a high-performance computer (HPC) slated to become the computational cornerstone of Laboratory mission-related work.
“Roadrunner ushers in a new era in high performance computing,” said Terry Wallace, principal associate director for science, technology, and engineering (PADSTE). Wallace praised the hard work and dedication of the entire Roadrunner Project Team at the celebration last Thursday.
Roadrunner is designed to achieve a sustained operating speed of 1,000 trillion calculations each second, or a “petaflop/s” in computer jargon—peta signifying the number 10 followed by 15 zeros, and flop/s meaning “floating point operation per second.” The fastest current supercomputer is rated at 478 teraflop/s (teraflop/s meaning one trillion floating point operations per second); Roadrunner would be roughly 3 times faster.
The computer will be developed in partnership with IBM and will utilize commercially available hardware, including aspects of commercial gaming and graphics technologies. Because of its off-the-shelf design, the computer costs significantly less than a one-of-a-kind machine. It uses a Linux operating system.
Roadrunner will be the first computer in the world to operate at sustained petaflop/s speeds. Although the Phase 3 Roadrunner Final system will be a very large cluster of nodes linked together by a high-speed interconnect, it is a unique hybrid petascale system. Each compute node in this cluster consists of two AMD Opteron™ dual-core processors plus four Cell™ processors used as computational accelerators. The Cell processors used in Roadrunner are a special IBM-developed variant of the Cell processor used in the Sony PlayStation 3®. The node-attached Cell accelerators are what make Roadrunner different than typical clusters.
The Lab and IBM have been in the development phase of Roadrunner since early 2006. The first phase of the project included delivery of an initial cluster that operates at a speed of 71 teraflop/s. The full-scale Roadrunner machine will operate more than 10 times faster on real applications than the current installed system. Laboratory researchers are using the 71-teraflop/s-version machine for classified weapons applications.
Phase 2 of the Roadrunner project was completed in October. Two external assessments, one by NNSA HQ and one by an independent team of HPC experts, evaluated the machine’s potential use for Laboratory applications, the Laboratory’s ability to successfully manage the computing system, IBM’s ability to deliver the product and whether computer programs could be adapted to the new system. Technical information from the assessments is available here.
Based on the positive outcome of the assessments, the Laboratory has decided to pursue the final phase of the Roadrunner project: development of a full-scale Roadrunner machine, pending NNSA approval.
The powerful clusters of nodes will process information to enable the Laboratory to use Roadrunner for advanced physics simulations and predictive simulations of complex scientific processes. Weapons science applications that can be processed by Roadrunner are applicable to all three of the U.S. Department of Energy weapons laboratories. The machine will also be well equipped to tackle the intricacies of modeling processes, ranging from the biomolecular to the cosmological.
After NNSA approves the Laboratory’s decision to acquire a full-scale Roadrunner system, named in honor of New Mexico’s official state bird, the machine should arrive at the Laboratory by next fall. The first computing applications are expected to begin running on the machine in January 2009.
The cost of the Roadrunner project is expected to be about $120 million. More than 100 Laboratory employees have been involved in this effort.
“This has been a remarkable partnership effort, exemplifying Los Alamos’s and IBM’s leadership in high performance computing,” Wallace said.
