dc.description.abstract |
In previous studies about the subject, hybrid methods were developed to benefit from the advantages of both sparse and dense structures for efficient storage of multidimensional OLAP data. In these previous studies, main concern was to develop efficient sparse - dense region splitting algorithms. Although, previously proposed hybrid methods are efficient, further improvement can be achieved by developing an effective physical storage method. In this study, we defined a chunk based physical storage structure to store multi-dimensional OLAP cubes that consolidates offset-value pairs, multi-dimensional array and sparse-dense split storage methods into a physical structure at chunk level and defined data access methods for this structure. At our hybrid storage, sparse and dense regions of a chunk are stored at spatially close locations on the disk to lower the number of page accessed in range queries. Also, we developed an attribute value order independent dense sub-cube determination heuristic to increase compression ratio. To illustrate the efficiency of our method, we conducted experiments and compared our results with a recent study. |
|