$\textrm{B}^s$-tree: A Self-tuning Index of Moving Objects

Abstract

Self-tuning database is a general paradigm for the future development of database systems. However, in moving object database, a vibrant and dynamic research area of the database community, the need for self-tuning has so far been overlooked. None of the existing spatio-temporal indexes can maintain high performance if the proportion of query and update operations varies significantly in the applications. We study the self-tuning indexing techniques which balance the query and update performances for optimal overall performance in moving object databases. In this paper, we propose a self-tuning framework which relies on a novel moving object index named $\textrm{B}^s$-tree. This framework is able to optimize its own overall performance by adapting to the workload online without interrupting the indexing service. We present various algorithms for the $\textrm{B}^s$-tree and the tuning techniques. Our extensive experiments show that the framework is effective, and the $\textrm{B}^s$-tree outperforms the existing indexes under different circumstances.