Parallelizing The Data Cube.

Frank Dehne, Todd Eavis, Susanne Hambrusch, Andrew Rau-Chaplin

Abstract: This paper presents a general methodology for the efficient parallelization of existing data cube construction algorithms. We describe two different partitioning strategies, one for top-down and one for bottom-up cube algorithms. Both partitioning strategies assign subcubes to individual processors in such a way that the loads assigned to the processors are balanced. Our methods reduce inter-processor communication overhead by partitioning the load in advance instead of computing each individual group-by in parallel as is done in previous parallel approaches. In fact, after the initial load distribution phase, each processor can compute its assigned subcube without any communication with the other processors. Our methods enable code reuse by permitting the use of existing sequential (external memory) data cube algorithms for the subcube computations on each processor. This supports the transfer of optimized sequential data cube code to a parallel setting. The bottom-up partitioning strategy balances the number of single attribute external memory sorts made by each processor. The top-down strategy partitions a weighted tree in which weights reflect algorithm specific cost measures like estimated group-by sizes. Both partitioning approaches can be implemented on any shared disk type parallel machine composed of p processors connected via an interconnection fabric and with access to a shared parallel disk array. Experimental results presented show that our partitioning strategies generate a close to optimal load balance between processors.

Download paper in .pdf format
Download paper in .ps format

Home * Publications