Research And Optimization For Data Management Technology Of MARS

Meteorological scientific data is an important part of Earth science data. Through the quantitative description and analysis of the physical and chemical properties of atmosphere, the intrinsic evolvement of atmosphere can be obtained and the atmospheric phenomenon can also be predicted. With the development of society, economy and scientifical technologies, exploring the huge values of the meteorological scientific data has been in pressing needs. As such, managing and sharing the massive meteorological scientific data efficiently by taking advantages of the computer and information technologies become more and more urgent. In this context, Meteorological Archival and Retrieval System(MARS) is first developed The aim of MARS is to manage the meteorological scientific data coded by self-described information. After years of using, MARS has played a key role in massive meteorological scientific data management. Basing on this data management platform, we study the related techniques of meteorological scientific data management. Particularly, to address the exposed performance bottleneck issues when running MARS, we propose and develop some improved algorithms for the metadata management of MARS. The main work and contribution in this paper are as follows:1?Researching the key technologies and frames of the meteorological scientific data management system developed and used by meteorological scientific research departments all over the world. Based on the analysis of the logical structure and physical storage of the meteorological scientific data coded by self-described information, multidimensional data model and metadata management in Mars are surveyed with the research of the object-oriented database and the interface of the file storage. Furthermore, the efficiency issues are disclosed when processing the massive data archive and retrieval.2?Proposing the Data Cube parallel computing method based on the multiway array in Mars. As a core technique, the index computation of metedata in Mars relies on Data Cube technology. Further, the research of multidimensional data model and Data Cube technologies are investigated. In the meantime, the efficiency issue of the serial computation of the metadata index is deeply analyzed. Then, the Data Cube parallel computing algorithm based on the multiway array is proposed. The final experiment shows that the improved algorithm can greatly reduce the Data Cube computation time.3?Presenting the parallel computing methods respectively on the metadata's retrieval and updation. Based on the logical and physical implementations of Mars' metadata, the parallel algorithms built by multiway array has been optimized. In the addition, We further present twice combining algorithm to merge the parallel computing results by Data Cube methods. The designed expeirments show that our proposed algorithm can greatly improve the efficiency of operating metadata when a large scale of retrieval and archival occur.Basing on the abovementioned work, we summarize our research and key management technologies of the metadata of Mars. Further, we propose the blue print for future work on Mars.