With the development of information technology, how to make more effective use of the data with explosive growth and implement the data analyzing task better is the main problem that needs to be solved urgently in the field of database in these years. One of the feasible solutions to solve these problems is main memory database which solves performance bottleneck of disk I/O, effectively improves the speed of data access. According to the characteristics of the main memory database, some research has been carried out on the indexing mechanism, but the existing indexing mechanism either have poor query performance or are unable to adapt to complex and variable analyzing tasks.In view of the above problems, indexing mechanism is researched based on the main memory database Monet DB. In the first place, the storage management mechanism of Monet DB that the index mechanism dependent on is analyzed. Then the data storage structure Binary Association Table and object pool which is closely related to the index mechanism is mainly analyzed. According to the underlying implementation of Monet DB storage management, the index mechanism is designed finally.In order to perform the analysis tasks better, this research focuses on the typical adaptive index algorithms, such as cracking index, self-adaptation merging index, and holistic index, and deeply analyzing the advantages and disadvantages of these algorithms. The holistic index mechanism monitor the usage of CPU and use idle CPU resources to create new thread for cracking data to speed up the construction of the index. Considering the poor performance of holistic index method, the improved indexing algorithm which based on holistic index use the newly created thread to sort the smaller data segments, eventually ordering the index completely and achieving better query performance.Experiment is designed to verify the improved index algorithm. The experimental results show that the more query task execution, improved indexing algorithm compared with holistic index has a 20% improvement in performance. Finally, by setting the different threshold size o f data sorting, experiment results show that the performance of the algorithm will be 8% fluctuations because of the cache hit rate, and the cache sensitivity of the improved indexing algorithm is verified. |