STORAGE MAPPINGS FOR MULTIDIMENSIONAL LINEAR DYNAMIC HASHING.

Simulation experiments which we carried out confirm the high performance of the Multidimensional Linear Dynamic Hashing (MLDH) method. For example, for uncontrolled splits, with b equals 5 and b prime equals 1, an average of 1. 09 accesses is necessary, irrespective of d (the dimension of the space), while maintaining an average load factor of 0. 67. On the other hand, for controlled splits, even with a threshold load factor of 0. 9 for b equals 10, b prime equals 2, we obtained an average number of accesses of 1. 74. Storage mappings for these splitting strategies were also developed, and as a byproduct we obtained a quite general class of extendible array storage mappings. We further intend to experiment with different distributions of attribute values in real files to see which splitting strategies are suitable.