Threading Opportunities in High-Performance Flash-Memory Storage

Storage Intensive Supercomputing The storage-intensive supercomputing (SISC) project [1] at LLNL is a research effort that is currently investigating hardware architectures for improving the performance of large, data-intensive applications. In order to achieve nextgeneration performance goals for these applications, it is necessary to consider system architectures that employ both computational accelerators (e.g., FPGAs or GPUs) and high-speed, mass-storage devices. In this work we focus on the latter by examining the performance characteristics of emerging flash-memory storage devices in the context of multicore environments.