TSQ: An Optimized Framework for Efficiently Answering Time Series Queries

Abstract Time series data are pervasive, with a multitude of applications across various fields including science, industry, Entertainment, medicine, and biology. These data sets often encompass large volumes of information. In the context of databases containing time series data, the increasing frequency of query tasks, such as (i) point queries, (ii) range queries, and (iii) top- k similarity queries, necessitates the enhancement of query processing efficiency. The question thus arises: how can the efficiency of these data processing tasks be enhanced? The solution is twofold. Firstly, dimensionality reduction algorithms can be employed to reduce the complexity of the data. Secondly, we can focus on the optimization of query algorithms. Consequently, based on these ideas, in this paper, we present an optimized framework for efficiently executing time series queries in the openGauss database, called TSQ. The framework is composed of two primary components: a dimensionality reduction (DR) module and a similarity query (SQ) module. The DR module incorporates several algorithms based on space-filling curves, which, through our enhancements, are capable of handling both high and low precision time series data more effectively. In the context of the SQ module, we employ two main optimization strategies, Early Abandoning and Sliding Window, to greatly improve the efficiency of similarity queries. Experimental results on four real-world time series datasets show that our framework not only optimizes general query tasks but also significantly enhances the efficiency of similarity queries.