CityBench: A configurable benchmark to evaluate RSP engines using smart city datasets

Abstract

With the growing popularity of Internet of Things (IoT) and IoT-enabled smart city applications, RDF stream processing (RSP) is gaining increasing attention in the Semantic Web community. As a result, several RSP engines have emerged, which are capable of processing semantically annotated data streams on the fly. Performance, correctness and technical soundness of few existing RSP engines have been evaluated in controlled settings using existing benchmarks like LSBench and SRBench. However, these benchmarks focus merely on features of the RSP query languages and engines, and do not consider dynamic application requirements and data-dependent properties such as changes in streaming rate during query execution or changes in application requirements over a period of time. This hinders wide adoption of RSP engines for real-time applications where data properties and application requirements play a key role and need to be characterised in their dynamic setting, such as in the smart city domain. In this paper, we present CityBench, a comprehensive benchmarking suite to evaluate RSP engines within smart city applications and with smart city data. CityBench includes real-time IoT data streams generated from various sensors deployed within the city of Aarhus, Denmark. We provide a configurable testing infrastructure and a set of continuous queries covering a variety of data- and application- dependent characteristics and performance metrics, to be executed over RSP engines using CityBench datasets. We evaluate two state of the art RSP engines using our testbed and discuss our experimental results. This work can be used as a baseline to identify capabilities and limitations of existing RSP engines for smart city applications.