Performance evaluation of real-time stream processing systems for Internet of Things applications

Abstract In the current scenario, IoT is an ideal and novel technology, which fulfills the needs of most of the commercial, non-commercial, government, and private organizations by its real-time supportive nature and characteristics. However, real-time processing itself a very critical research topic. But, most of the IoT applications are empowered by real-time data processing. Thus, it become a vital part of IoT. In this work, we proposed a four-layer infrastructure for IoT along with stream processing. Further, we use stream processing techniques along with IoT infrastructure for applications and analyze the performance of stream processing techniques for IoT applications. Also, we compare and find the five most suitable distributed stream processing systems for IoT, based on its performance and characteristics. We use two benchmark applications to evaluate the performance of distributed stream processing systems against response time, throughput, jitter, and scalability. Based on that, we suggest the adapted solution for IoT applications. We evaluate the performance with peak stream rates from 100k to 1M along with the various frequencies of benchmark applications. Further, on the basis of results, we conclude that Apache NiFi is the most suitable solution for IoT applications.