Research And Implementation Of Olap Query Optimization

As time goes on, the data in the data warehouse, which contains a wealth of information, will expand rapidly. The data warehouse technology just solves the problem of the storage of the data and improves its efficiency. Only by combining with the data analysis techniques, can we abstain the decisional information from the large and complex data environment. But the data analysis dealing with mass data needs a large complex operation with those data, time-consuming are very serious. Specifically Mondrian OLAP system is based on ROLAP system, which results in many deficiencies in handing multi-dimensional data, so we have to decrease the system's response time by optimizing it.This thesis focuses on studying in depth to how to improve the query performance of Mondrian OLAP system, and proposes to bring forecast management and data-caching technology into the Mondrian OLAP system, in additional, adds a cold start process to obtain aggregation data from the relational database to the cache. Those would reduce the response time.Firstly, this thesis analyses the existing problems of the Mondrian OLAP system, and puts forward the overall framework of cache management proposal, which is based on forecast management and cold start process Then,discusses its three key technologies in-depth: forecast management technology, management techniques based on the result set cache and cold start process. Forecast management system can abstain the user's pattern of query multi-dimensional data behaviors by tracking their query logs, and base on these behaviors to predict the user's query behavior in the future, making the cache system always save the best result set for the future query. In additional, considering the characteristics of Mondrian OLAP system, this thesis chooses the best cached object and the algorithm of cache management, and designs the L2 cache to store the user's query result set. The cold-start process loads the data result set to the cache system when the OLAP system initials start-up or when restart, making the system user's query could always have a faster response time. Experimental results show that, after improved the Mondrian OLAP system, the performance bottleneck due to the rapid growth of data warehouse to some extent is solved indeed, and the overall Mondrian OLAP system's query performance is improved.