Los Alamos receives innovation and promise award for computer storage research

LOS ALAMOS, N.M. -- Los Alamos National Laboratory has been honored for innovations in large-scale computer storage. The Laboratory has received one of five "Best Practices in Storage Awards” from Computerworld and the Storage Network Industry Association. The award was presented at the Storage Networking World meeting in Orlando. The award, for “Innovation and Promise,” recognizes the role played by the Laboratory, industry partner Panasas Inc. of Fremont, Calif., and the National Nuclear Security Administration in developing secure, object-based storage and the scalable file systems that use it. The storage system enables the Laboratory and others to efficiently use some of the largest commodity cluster computer systems in the world, such as Los Alamos’ new Lightning system, with a peak speed of 11.26 trillion operations per second, or 11.26 teraOPS. For about five years, Los Alamos has strongly supported the development of scalable file systems through design, funding, testing and prototyping, and has worked with Panasas and other partners to identify the best uses of the object-based storage model “Secure, scalable storage for commodity supercomputers will benefit Los Alamos’ national security mission as well as large web-server operations and anyone who manages huge quantities of data,” said Gary Grider of Los Alamos’ High Performance Computing Environments Group. “Los Alamos and the National Nuclear Security Administration have been major forces in technical efforts with academia and industry partners such as Panasas to shape where scalable storage and file systems are headed.” The parallel file systems that store data from cluster computers require extremely complex software. Before the advent of object-based storage, the parallel file system had to keep track of all the blocks of data on all storage devices, increasing complexity and causing scalability problems due to accounting in parallel. Object-based storage, developed in large part by Garth Gibson of Carnegie-Mellon University and now also with Panasas, greatly simplifies parallel storage software with a set of secure commands that tell devices to store and manage a variable quantity of data, thereby making for a simpler and more secure approach to the complexities that parallel file systems pose.