A Hopping Umbrella for Fuzzy Joining Data Streams From IoT Devices in the Cloud and on the Edge

Internet of Things (IoT) is a new technology that changes the image of the current world, yielding new possibilities, but also proliferating data. IoT devices may constantly produce enormous amounts of data as data streams that can be analyzed in real time and also collected for further exploration in data lakes in huge data centers. Due to their scaling capabilities, these data centers are frequently located in the Cloud. However, recent rapid growth in the number of IoT devices and their applications in manufacturing, transport, and health care motivates moving the burden of data processing and analysis to the Edge. One of the phases of data processing is combining data streams from two (or more) IoT devices that monitor the same object and work asynchronously. Since they generate sensor readings at various moments of time, their data streams must be properly combined in order to obtain a complete image of the monitored object or process. In this article, we present the idea of a hopping umbrella which fuzzifies timestamps from sensor readings while joining data streams from asynchronous IoT devices in a flexible way. In contrast to processing data at rest, the hopping umbrella implements the fuzzy join operation in time windows for data streams (data in motion). By using fuzzy sets, the hopping umbrella not only allows combining asynchronous events from multiple sensors, but also facilitates evaluation of the degree of matching of the combined sensor readings, and consequently allows for reduction of the output stream size. Our experiments performed in Cloud and on Edge devices proved that with the use of this idea, we are able to properly join the best matching sensor readings and in some scenarios, reduce the number of data transferred to the Cloud data center without significant overhead in resource utilization of stream processing units.