In the context of multidimensional databases implemented on relational DBMSs through star schemes, the most effective technique to enhance performances consists of materializing redundant aggregates called views. In this paper we investigate the problem of vertical fragmentation of views aimed at minimizing the workload response time. Each view includes several measures which not necessarily are always requested together; thus, the system performance may be increased by partitioning the views into smaller tables. On the other hand, drill-across queries involve measures taken from two or more views; in this case the access costs may be decreased by unifying these views into larger tables. After formalizing the fragmentation problem as a 0–1 integer linear programming problem, we define a cost function and outline a branch-and-bound algorithm to minimize it. Finally, we demonstrate the usefulness of our approach by presenting a set of experimental results based on the TPC-D benchmark.