OLAP database computation with a splitcube on a multi-core system

In On-Line Analytical Processing (OLAP), the computation of a full computed data cube has exponential complexity due to the cube dimensionality. An efficient alternative is splitcube which does not emphasize on a full computed data cube. In this report, we explore the use of newly published Microsoft Parallel Extension to build and query splitcube in parallel on a multi-core system. We provide an implementation of a parallel splitcube building program as well as an implementation of a parallel splitcube query processing program. Experiment data shows we can achieve speedup in parallel splitcube building process which is bounded by the performance of disk I/O operation. Experiment data also shows slowdown in parallel splitcube query process. The slowdown is caused by the limited disk I/O bandwidth and the increased latency of synchronous disk read operations result from multiple non-contiguous accesses to the physical disk.