Parallel algorithms for executing join on cube-connected multicomputers

The authors present parallel processing of database joins in a hypercube system based on a dynamic and order-preserving data partitioning storage organization referred to as DYOP. A DYOP file system corresponds to a grid-based file organization. A grid-based file is any file organization where the n-field (attribute) records of a file are represented as points in an n-dimensional space. Such a hyperspace is divided into a set of hyperrectangles, each of which corresponds to a disk block. In DYOP, values along each attribute are stored in an order-preserving fashion. As the file grows, its partitions are split along a dimension (attribute) by preserving the order of values. As the file growth continues, further splits are performed in such a way that each split occurs along a different attribute in a cyclic fashion. There have been two previous studies in implementing a database join operation on hypercubes. A performance comparison of these studies is presented