In distributed computing environments, replication of data provides improved availability, isolation between workloads with different characteristics, and improved performance through local access to data. The "real data" is server resident and by "local data" we refer to cached client data. We examine which data should be cached on behalf of a cached query. The minimum requirement for cached data for a query Q is that it enables answering Q locally. We consider the following: 1) Definitions of what data is cached for a cached query. 2) Deciding whether cached data can be used to solve a "new" query. 3) Deciding whether a "new" query to be cached is already effectively cached due to caching of other queries. 4) A simple class of caching rules.