Expiring Data from the Warehouse

Data warehouses are used to collect and analyze data from remote sources. The data collected often originate from transactional information and can become very large. This paper presents a framework for incrementally removing warehouse data (without a need to fully recompute views), o ering two choices. One is to expunge data, in which case the result is as if the data had never existed. The second is to expire data, in which case views de ned over the data are not necessarily a ected. Within the framework, a user or administrator can specify what data to expire or expunge, what auxiliary data is to be kept for facilitating incremental view maintenance, what type of updates are expected from external sources, and how the system should compensate when data is expired or other parameters changed. We present algorithms for the various expiration and compensation actions, and we show how our framework can be implemented on top of a conventional RDBMS.