Energy-Aware Allocation of Approximate Query Processing over Data Streams with Error Guarantee

Xiaohui Wei,Yuanyuan Liu,Shang Gao,Xingwang Wang

DOI: https://doi.org/10.1109/iwqos.2018.8624123

2018-01-01

Abstract:With increasing real-time and resource-intensive requirements, approximate computing is widely adopted to improve the performance of query processing over data streams. However, existing works concentrate on simple queries with single-step operations, such as point or join queries. There are a large number of nested queries with selection or filtering operations before aggregation. In this poster, we focus on approximate nested stream queries. We first propose a novel approximate model, SCM-sketches, that makes two-stage approximation for nested query answering with guaranteed errors. In the first stage for nested filtering operations, we use the sampling method to compress the arriving data. Then in the second stage, a sketch is used for further aggregation or join operations. We also theoretically analyze the effect of error propagation on approximate errors. Compared with existing sketch-based methods, experiment results with real-life datasets verify the effectiveness of SCM-sketches.