An Active Storage System for High Performance Computing

Traditionally, active storage devices execute custom application code on large amounts of data by utilizing the unused processing power of the storage nodes. For computation-intensive applications, the performance might be quite low due to insufficient processing power on the storage devices. In this paper we present a reconfigurable computing solution that can provide flexible, high-performance processing capabilities for the storage nodes. We have made a change over the existing active storage framework for object-based storage device by off loading computation-intensive request-driven services to a reconfigurable hardware accelerator. In the experiment, we use two applications (AES encryption and FFT) as the case studies, and the results show that the reconfigurable computing-based system can dramatically increase the performance comparing with the traditional software-based active storage systems.