In this paper we propose a new cache management scheme for online analytical processing (OLAP) systems based on the usability
of query results in rewriting and processing other queries. For effective admission and replacement of OLAP query results, we
consider the benefit of query results not only for recently issued queries but for the expected future queries of a current query. We
exploit semantic relationships between successive queries in an OLAP session, which are derived from the interactive and navigational
nature of OLAP query workloads, in order to classify and predict subsequent future queries. We present a method for
estimating the usability of query results for the representative future queries using a probability model for them. Experimental
evaluation shows that our caching scheme using the past and future usability of query results can reduce the cost of processing
OLAP query workloads effectively only with a small cache size and outperforms the previous caching strategies for OLAP systems.