Application Performance Monitoring in Wireless Sensor Networks

Wireless Sensor Systems (WSNs) are widely deployed in industrial environments to ensure the flawless operation of all entities in a factory. It is therefore crucial to install a WSN that the user can count on. In fact, the overwhelming majority of people talks about reliability when asked for the most important features in a WSN [1]. Because the sensors inside a WSN are usually extremely limited in terms of size and cost, they only offer a limited amount of battery lifetime. This makes it understandable that vendors tend to completely neglect Application Performance Monitoring (APM) capabilities in such constrained devices. Isn’t this inconsistent when recalling that reliability is the most important feature from a customer’s point of view? – Yes, it is! 

The IEEE802.15.4e standard ​[2]​ has proven to be a very reliable and robust basis for industrial IoT networks ​[3]​. We want to find methods to introduce APM in such constrained systems while keeping the impact on the performance in terms of latency, storage, and power consumption as low as possible. A common method is to periodically collect performance metrics ​[4]​ in fixed time intervals and deliver them to the user.  This concept is implemented in different state-of-the-art applications under various names, such as “health reports” ​[5]​, “snapshots” ​[6]​ or “heartbeats” ​[7]​. In our research, we evaluate these approaches and develop new strategies that optimize the balance between the verbosity of APM and the benefits from the visibility that it brings. 

Once the user is equipped with a powerful monitoring tool of their WSN, they have the opportunity to detect crucial trends, find root causes of failures, or even foresee critical events. Having full visibility of the network and its “health condition” is an essential feature when it comes to WSNs with moving elements in a dynamic sensor swarm. Since APM is a topic which is unsatisfactorily addressed in the research community until now, we work on closing this gap with our contributions to the OpenSwarm project.