"Secure, Energy-Efficient, Evolvable, Long-Term Archival Storage"

Users are storing ever-increasing amounts of information digitally, driven by many factors including government regulations and the public's desire to digitally record their personal histories. Unfortunately, we have yet to demonstrate that we can reliably preserve digital data for more than a few years, putting a generation's cultural legacy at risk. Much of the problem is rooted in our approach to building long-term storage systems; currently archival systems are developed using the same approaches, access patterns and techniques used to design higher-performance, shorter-term storage systems. As a result, current archival storage systems still rely on strategies that fail in long-term scenarios, waste money and energy, and perpetuate the endless cycles of "fork-lift" upgrades and wholesale migrations needed to remain efficient and up to date.

In my thesis, I demonstrate that archival storage is a first class category of storage that requires specialized solutions. To this end, I present several techniques tailored specifically for the unique demands of long-lived data. To explore the security needs of archival data, I have developed POTSHARDS, which offers secrecy through unconditionally secure secrecy techniques, and survivability through increased attack detection and built-in data recovery. To study cost savings, I have created Pergamum, a distributed system of intelligent storage appliances that stores data reliably with multi-level encoding and a hierarchical auditing scheme, and energy-efficiently by leveraging existing MAID techniques, while extending them by exploiting the different access patterns of data and metadata. Running atop of Pergamum is Logan, a management layer being developed that actively identifies and decommissions wasteful devices in order to continuously maximize system efficiency. These systems combine to demonstrate significant progress towards effective, secure, energy-efficient, and evolvable archival storage.