A Linux cluster deployed at Lawrence Livermore National Laboratory and codenamed "Thunder" yesterday delivered 19.94 teraflops of sustained performance, making it the most powerful computer in North America - and the second fastest on Earth.

The 4,096 Itanium 2 processor based cluster is the largest Itanium 2 processor deployment, as well as the largest implementation of Quadrics' low-latency QsNet(II) interconnect technology - technologies which, in combination, allow Thunder to achieve record cluster efficiency of 86.9%, an important metric in measuring cluster scalability.

"Thunder" uses 1,024 California Digital 6440 servers, and California Digital CEO B.J. Arun was understandably proud, saying: "Thunder sets important benchmarks for massively-parallel Linux computing. We're proud to have successfully delivered such a ground-breaking Linux cluster with world-record performance and efficiency."

"Working with California Digital and Lawrence Livermore has been a great opportunity to demonstrate the absolute performance and scalability that can be achieved with Intel's Itanium2 processor," said Intel Enterprise Platforms General Manager Richard Dracott.

An official announcement declared:

Thunder's efficiency and scalability rest on the strength of its sophisticated interconnect technology, Quadrics' QsNet(II) offering. QsNet(II) (Elan4) provides the underlying high bandwidth and low latency MPI communications required by today's demanding scalable applications. With support for broadcast in hardware and scalable collective operations, QsNet(II) scales clusters efficiently to over 4,000 nodes.

Despite the technical sophistication of Thunder and the incorporation of new technologies, California Digital deployed "Thunder" in five months, speeding delivery of computing solutions to support Lawrence Livermore's national security and science programs in fields such as materials science, structural mechanics, electromagnetics, atmospheric science, seismology, biology, and inertial confinement fusion.

"Thunder represents the next generation of Linux cluster for scientific simulation," remarked Mark Seager, Livermore's Assistant Department Head for Advanced Technology. "Our applications are seeing a 50% to 400% speed up over our Intel Xeon based clusters."

California Digital has released a number of the innovative tools that manage the cluster effectively under GPL/ open source licenses as part of its FreeIPMI project for server management and configuration.

• Los Alamos Lab to Get Another Linux Supercomputer
• 'Super-Linux' Cluster Declared Third-Fastest Computer On Earth
• "Powell Cluster" Becomes Army Research Lab's Most Powerful Linux Cluster
• What About the Linux End Users?