Vacuuming XML

A transaction-time XML database archives XML data versions to create a complete history of the data and processes transaction-time queries to fetch requested versions of data. The database will grow over time as new versions are created, and its unrestrained growth will eventually exceed storage capacity. To become an autonomous, self-managing database this growth must be controlled by vacuuming versions of data, which reclaims space and prevents the data from exceeding storage capacity. Vacuuming policies specify which versions to vacuum. This paper develops micro and macro vacuuming policies for an XML data collection. A micro vacuuming policy targets specific elements in the data, while a macro policy applies to the entire collection. We show how the mix of policies can effectively manage the growth of a database. Vacuuming hierarchical data differs from vacuuming relational data since vacuuming data high in the hierarchy also potentially removes data versions, not explicitly vacuumed, lower in the hierarchy. Vacuuming controls the growth of the archive by sacrificing the completeness of the data's history. Some requests can no longer be satisfied since a requested version might have been vacuumed. An autonomous transaction-time database must also be self-repairing, so this paper also presents a query repair strategy to redirect queries to data that has not been vacuumed. All of the new functionalities were designed to be backwards compatible with existing protocols (e.g., SOAP) and standards (e.g., XML), so a data collection can become a vacuum-enhanced, transaction-time collection at any time.