Today’s high-speed clusters can easily use the latest interconnect technologies for node-to-node communications, but the I/O storage systems have failed to keep pace. This lag in advancement is of major concern because science, national security, and business computing are becoming more data intensive. Data management challenges will, if not already, exceed the compute-power challenge. Therefore, a versatile storage architecture that is scalable, sharable, and secure is critical to meeting the performance demands for all sectors of High Performance Computing (HPC).
The Scalable Computing R&D Department at Sandia National Laboratories is conducting an R&D project to address the I/O challenges in HPC. Through research and industry partnerships, this study explores Parallel and Remote Direct Memory Access (RDMA)technologies for scalability, promotes standardized common filesystem interface to facilitate sharing, and is compatible with Sandia’s corporate security infrastructure.