Operational Data Storage Unification

We see a world where personal system software is self-installing, self-configuring, self-personalizing, self-administering, and self-healing. In this world, personal system data is visible in a single name space regardless of " owning " application and data location, supports uniform non-procedural query, transactions, and security, is automatically backed up, is recoverable, and supports portable and disconnected operation. This paper focuses on how database technology can be applied to help achieve many of these goals. We propose an integrated storage manager, the Virtual Object File System (VOFS), which supports structured files, unstructured files, and relational access. VOFS also supports disconnected operations by caching and replicating all data stored via any of the supported persistence APIs. We show how the combination of support for storage integration and disconnected and portable operation found in VOFS can substantially reduce administrative costs and support more compelling client applications. The VOFS system can be extended to support alternative caching systems, new access methods based upon past navigational paths and temporal access, and different software and data billing schemes. Acknowledgements I would like to thank my supervisor Dr. Jacob Slonim and reader Dr. Grant Weddell for advice, comments, and suggestions throughout this project that have both improved the paper and my understanding of the subject material. Dr. Pat Selinger of the IBM Almaden Research Center provided numerous invaluable technical and editorial comments and was always available for technical discussion. I would also like to thank Dr. Paul Larson of Microsoft Research for time, advice, and education in the early stages of the degree program. The IBM Canada Laboratory provided financial assistance for this work through the CONGESE program.