Los Alamos taps Panasas storage for supercomputer
- By John Moore
- Dec 28, 2006
The Los Alamos National Laboratory is tapping storage technology from Panasas for the lab’s Roadrunner supercomputer.
Los Alamos will use Panasas’ ActiveScale Storage Cluster for its Roadrunner platform. Roadrunner, when completed, aims to run at a sustained performance level of 1 petaflop, or 1,000 trillion calculations per second, according to the lab. Los Alamos will expand its Panasas storage installation to support the petascale system.
Meanwhile, the lab is testing two recently released Panasas products, ActiveStor 3000 and ActiveScale 3.0, said Larry Jones, vice president of marketing at Panasas. ActiveStor 3000, announced in November, is a storage cluster that is designed for high-capacity, batch-oriented Linux cluster applications, Panasas said. ActiveScale 3.0, also announced last month, is the latest version of the storage cluster’s operating environment.
While testing the new storage cluster, “the largest benefits we have seen are in performance for some particularly difficult [input/output] applications, faster single-client bandwidth -- which helps us with many applications -- and better protection from failure though the scalable rebuild features, which at our scale is a concern,” said Gary Grider, group leader of Los Alamos’ High Performance Computing Systems Integration Group.
Jones said faster single-client bandwidth supports larger individual compute nodes. Roadrunner features 4x dual-core Advanced Micro Devices Opteron systems, he added.
Roadrunner also uses IBM’s Cell Broadband Engine processors. Los Alamos will use the supercomputer to run nuclear weapons calculations.