Research On The Index Mechanism In The Uniifed Access Of The Seismic Data

The seismic data is typical large-scale static data sets, which consist massive, specificrecords or data block. Because of the large scale, these data usually store in the distributed filesystem, and they are rarely modified once written. And it has a variety of storage format. If theapplication gives every interface for every format, not only its own code will greatly increase, butalso it will limit the flexibility, portability and scalability. In order to solve the problems above,people research on a variety of solutions, including the unified data access of multi-format bigdata. It proposes the unified logical description for data establishing. Through the standardizedinterface of the logic layer, it realizes the unified access of the physical data. The design has awell flexibility. At the same time, because of the logic layer between the physic layer and theapplication layer, it also added extra operation in the physical layer and the potential drawbacksin efficiency.In this paper, based on the characteristics of the seismic data, we optimize the indexstructure of the logic layer in the unified access of the seismic data (UASD) model, and constructthe distributed index structure based on the UASD. At the same time, the HDFS-based levelindex structure, which is applied to the B-tree, R-tree and its variants, is proposed. And give theparallel building ways of the hierarchical index. Research and analysis the model of the unifiedaccess of the seismic data. Optimize the index structure of the logic layer in the model. Andextract an index formula which can quickly locate the index. Propose the distributed indexstructure based on the UASD and HDFS, which improves the index performance of the UASD.And present the strategy of building index in different levels based on the keyword of the seismictrace header, avoiding the extra space for the inactive index of keyword index and themaintenance overhead. Research and analysis the common index mechanism in big data. Andconstruct the HDFS-based level index structure, which is applied to the B-tree, R-tree and itsvariants and improves the disadvantages of key-value storage structure in the non-primary keyindex. Give the parallel building ways of the hierarchical index and accelerate the index buildingspeed. Finally, synthesizing the content of this study, give the application of the UASD in theparallel processing platform of the seismic data, which verifies the feasibility and effectiveness ofthe techniques and algorithms proposed above.The experiments show that the series method reduces the query response time, the indexspace and data transfer overhead, and effectively enhances the index performance of the UASD.