Query caching and view selection for XML databases

TLDR

The paper proposes a method to maintain a semantic cache of materialized XPath views. The method stores and maintains cached XPath views—both previously queried and selected ones—in relational tables, using a notion of query/view answerability to reduce tree operations to string operations, and selects views based on a warm‑up workload, allowing the cache to reside inside or outside the database. Experiments show that the caching technique is efficient and yields performance gains.

Abstract

In this paper, we propose a method for maintaining a semantic cache of materialized XPath views. The cached views include queries that have been previously asked, and additional selected views. The cache can be stored inside or outside the database. We describe a notion of XPath query/view answerability, which allows us to reduce tree operations to string operations for matching a query/view pair. We show how to store and maintain the cached views in relational tables, so that cache lookup is very efficient. We also describe a technique for view selection, given a warm-up workload. We experimentally demonstrate the efficiency of our caching techniques, and performance gains obtained by employing such a cache.