Master of Science (MS)
Bottom-Up Computation (BUC) is one of the most studied algorithms for data cube generation in on-line analytical processing. Its computation in the bottom-up style allows the algorithm to efficiently generate a data cube for memory-sized input data. When the entire input data cannot fit into memory, many literatures suggest partitioning the data by a dimension and run the algorithm on each of the single-dimensional partitioned data. For very large sized input data, the partitioned data might still not be able to fit into the memory and partitioning by additional dimensions is required; however, this multi- dimensional partitioning is more complicated than single-dimensional partitioning and it has not been fully discussed before. Our goal is to provide a heuristic implementation on multi-dimensional partitioning in BUC. To confirm our design, we compare it with our implemented PipeSort, which is a top-down data cubing algorithm; meanwhile, we confirm the advantages and disadvantages between the top-down data cubing algorithm and the bottom-up data cubing algorithm.
Yeung, Kenneth, "Multi-Dimensional Partitioning in BUC for Data Cubes" (2008). Master's Projects. Paper 118.