LoneStar: Design and evaluation of an energy-efficient, disk-based archival storage system.

LoneStar: Design and Evaluation of an Energy-E cient, Disk-Based Archival Storage System by Matthias Grawinkel Tape has been the predominant storage target for digital archives, but with workloads becoming more active and storage media like hard disk drives catching up on costs-perbyte, the archival storage stack needs to be reinvestigated. Reliability, integrity, and durability are the main goals of digital preservation, but the role of performance increases with active archives that provide near-online access to all stored data. A tape-only solution cannot deliver the required parallelism, latencies, and throughput, which leads to disk-based caches as compensation. In this thesis, we explore the challenges and opportunities of building a hard disk drive based storage system that aims to be highly reliable and energy e cient and is suited for active archives as well as cold storage environments. First, we analyze the storage landscape and access behavior of a large digital archive to present a use case for the proposed architecture. Then, we develop mechanisms to improve the reliability of a single disk drive and propose and evaluate a new energy e cient two-dimensional RAID scheme that is optimized for “write once, read sometimes” workloads. Finally, we present journaling and caching schemes that support the underlying goals and evaluate the RAID scheme in a le system environment.