Metrics for Self-Adaptive Queuing in Middleware for Internet of Things

Abstract

Internet of Things (IoT) is a cornerstone technology for automation in the physical world. In particular, IoT allows industrial automation, also known as Industry 4.0. With overwhelming amount of sensor types and communication protocols, management of IoT middleware becomes unfeasible. This problem might be addressed by implementing self-adaptive functionality in IoT middleware. The presented paper contributes to the studies of the self-adaptive message queuing in IoT middleware: an estimated waiting time (EWT) metric for automating the scaling of message queuing subsystems is proposed and evaluated on CPU-intensive and blocking I/O-intensive tasks. Mixed metrics (with conventional CPU utilization and processing capacity) were also evaluated. Evaluation of the proposed metrics based on Google Kubernetes Engine revealed cost reduction potential of EWT and the well-balanced quality of queuing IoT middleware deployments provided by processing capacity metric.