Multidisk vertical allocation (MDVA) problems intend to find an allocation of relations to disks such that the expected query cost is minimized. Recently, Chang [European Journal of Operational Research 143 (2002) 210] modified Rotem et al.'s [IEEE Transactions on Knowledge and Data Engineering 5 (1993) 882] method for solving an MDVA problem using a smaller number of binary variables. Chang's method however is unable to treat MDVA problems with possible replication of relations. This paper proposes another method to solve MDVA problems, which is more effective than Rotem et al.'s and is able to treat replication problems insolvable by Chang's method.
- Data allocation
- Global optimization