View materialization vs. indexing: Balancing space constraints in data warehouse design

View materialization and indexing are the most effective techniques adopted in data warehouses to improve query performance. Since both materialization and indexing algorithms are driven by a constraint on the disk space made available for each, the designer would greatly benefit from being enabled to determine a priori which fractions of the global space available must be devoted to views and indexes, respectively, in order to optimally tune performances. In this paper we first present a comparative evaluation of the benefit (saving per disk page) brought by view materialization and indexing for a single query expressed on a star scheme. Then, we face the problem of determining an effective trade-off between the two space fractions for the core workload of the warehouse. Some experimental results are reported, which prove that the estimated trade-off is satisfactorily near to the optimal one.