A HYBRID INDEX FOR MULTI-DIMENSIONAL QUERY IN HBASE

With the explosive growth of data, distributed databases are widely used in various applications, including e-commerce, social networking, recommendation system, location-based service and etc. Among them, the use of HBase is the most common. However, it does not natively support multi-dimensional query and the existing multi-dimensional index established on HBase has some disadvantages such as unsupporting floating-point numbers and low efficiency of range query. In this paper, we propose a hybrid index for multi-dimensional query in HBase to address these issues. To build the index, we use the z-ordering curve to divide the multi-dimensional space into grids, then we adopt the bit interleaving technique to generate GridID, after that we refer to the coding generation method of Pyramid index in each grid. Combining the z-ordering curve with pyramid technology, our index supports the operation of floating-point numbers, efficient multi-dimensional data processing and range query. Besides, we implement the index structure on HBase and run some experiments on real data. The results of experiments show the index achieves range query and outperforms other index structures.