ABR-Tree: An Efficient Distributed Multidimensional Indexing Approach for Massive Data

In the big data era, there many application scenarios urgently need efficient distributed multidimensional indexing approach to accelerate the data analytics. To address this issue, in this paper, we propose ABR-Tree, a multidimensional distributed indexing approach. ABR-Tree consist of two components, the global append-efficient Bi¾ź+i¾ź-Tree, and the local R*-Tree. Both of them are layered over the cloud database as the index and data store, which not only make ABR-Tree is easy to implement and inherently become a distributed cloud index, but also enable ABR-Tree can sustain high throughput workload and large data volumes, meanwhile, ensuring fault-tolerance, and high availability. We conducted extensive experiments over 1i¾źTB real data set to evaluate its efficiency of processing multidimensional range queries, the results show that it is significantly fast than the existing representative distributed multidimensional cloud index method.