A data structure and search method for a data base management system. The structure and method allow the locating of a stored record in a massive system in a controlled and small number of mass memory accesses. The data structure is arranged into a plurality of search trees, each defining patent nodes and terminal nodes. The nodes of a tree are hierarchically arranged, and the trees are hierarchically arranged as a whole into levels. The initial search tree and an initial subset of trees, in some cases, are designed to be maintained in a main fast access memory. The remaining trees are kept in mass memory. A plurality of first storage files maintained in the mass memory are associated with terminal nodes of each of the trees except the final trees in the hierarchical structure. Terminating storage files, which are the ultimate repository for information, are associated with terminal nodes of the final trees. An input search parameter is partitioned into a plurality of subparameters, one for each level of search trees. The subparameters are used to search a tree in each level of the data structure until the location of a terminating file is determined.