ACADEMIA

Cluster File Systems (CFS), announced that the open source Lustre network file system will start using Sun's open source OpenSolaris ZFS disk file system on Lustre servers running Linux operating systems. Lustre is a leading HPC file system, running on more than 50 percent of the 30 biggest computers in the world. Sun Microsystems’ OpenSolaris ZFS file system contains revolutionary scalability, storage management and robustness enhancements that are ideally suited for use on Lustre servers, which handle 100's of GigaBytes per second in I/O coming from clusters with many 10,000's of client systems, hundreds of servers and petabytes of storage. The transition to the OpenSolaris ZFS will help Lustre servers achieve greater scalability and robustness. Lustre verifies and recovers data integrity over the network while OpenSolaris ZFS automatically verifies and maintains data integrity. This is critical as Lustre installations usually contain many thousands of disks. "Our customers are asking for features in Lustre servers which are available through the use of OpenSolaris ZFS. This will shorten our time to market with future versions of our product that scale to trillions of files and exabytes of storage with greatly enhanced storage management," said Peter Braam, CEO of Cluster File Systems. "We are encouraged to see the Lustre file system transition to ZFS as a backend because it is a natural fit to take Lustre to the next level of scalability and robustness that we will need in operation for petascale computing and beyond,” said Karl Schulz, Assistant Director for High Performance Computing (HPC) at the University of Texas at Austin (TACC). "Lustre is a proven, scalable, storage solution which is used in some of the world's largest supercomputers, from Tokyo Institute of Technology’s (Tokyo Tech) TSUBAME - the fastest supercomputer in Asia which has been using Sun storage running Lustre for over a year - to the TACC Ranger system currently being installed," said Marc Hamilton, vice president of Solaris Marketing at Sun. "With more than 1.5 petabytes of Sun storage each, the TSUBAME and Ranger supercomputers - as well as smaller HPC systems, often critical elements of a workflow in commercial companies – will benefit from the transition to Lustre running with OpenSolaris ZFS and Solaris, bringing new levels of data integrity and scalability that were previously not feasible." "This 'combination' of Lustre and OpenSolaris ZFS is an dream integration come true," said Satoshi Matsuoka, the Professor in charge of the TSUBAME supercomputer project at Tokyo Tech. "Lustre and OpenSolaris ZFS combined constitute a dominant portion of TSUBAME's storage which has grown to 1.6 petabytes, and have provided a solid operational track record. Their tight coupling would only serve to further enhance their excellence as we continue to grow our storage."