A Tractable Query Cache by Approximation

In this paper we present the organization of a predicate-based query cache suitable for integration with agent-based heterogeneous database systems. The cache is managed using a tractable (sound and complete) query containment algorithm, yet there are no language restrictions placed on the applications. This is accomplished by introducing query approximation. Query approximation is a compilation technique where a query expression in a general query language is mapped to a query expression in a restricted language. We deene a target language such that query containment can be tested in polynomial time. We deene a query approximation method that guarantees that the approximate query contains the original query. In the remainder of the paper we specify the algorithms by which the query engine and the cache manager may negotiate a choice of approximation and the development of a query plan. We use two application workloads and the TPC-D benchmark queries to assess the impact of query approximation and the value of the cache within the architecture of the InfoSleuth heterogeneous database system.