Research On Parallelled Data Cube Computing Method Based On Multi-core CPU

With the era of big data, the fields of online analytical processing of large dataare also facing serious challenges. Online analytical processing based onmulti-dimensional analysis constructs data cube and presents it to users afterprocessing. Data cube calculation has been a research focus in the fields of onlineanalytical processing. With the dimensions and data size increasing, the traditionaldata cube calculation experience severe bottlenecks that it is difficult to generate thedata cube with high dimensions and large amount of data.With the recent popularization and application of multi-core CPU and memorycapacity continues to grow, how to create a memory-based data structure andefficiently use the concurrent multi-core CPU processing power to handle massivehigh-dimensional data cube becomes a research hotspot. This paper presents amemory-based cube tree data structure model and proposed a cube calculationmethod which take advantage of the multi-core CPU concurrency to accelerate thecube calculation.For high-dimensional data cube calculation problem, the main researches are asfollows:(1) Summarize the basic knowledge of data cube and the classification of theaggregation functions in data cube aggregation. Described the basic operations ondata cube and characterized the classic data cubes;(2) Discusses the Statistical Tree data structure and the algorithms based onStatistical Tree in depth. I propose Cube Tree storage structure which overcomes theshortcomings of Statistic Tree which takes up a lot storage space and propose aparallel algorithm to create cube trees which take advantage of multi-core CPUs.(3) In this paper I propose the Rearrangement Sub-cube Generation Algorithmwhich achieves fast data cube query speed after take advantage of the parallelism ofmulti-core CPUs. Finally, I proved the effectiveness of the algorithm throughexperiments.