An Observability Framework for Predicting the Behavior of IoT Systems

Thesis (Master's)--University of Washington, 2020Internet of Things (IoT) systems typically consume a number of microservices for completing a business process or task. To this extent, the Quality of Service (QoS) of individual microservices is crucial for measuring the overall performance of the system while identifying potential downtime or sources of failure. By observing the behavior of microservices, it is then possible to identify the root causes of bottlenecks or performance-related issues. In this thesis, we present the Distributed Observability Framework (DOF) for observing the behavior of microservices in a distributed IoT system. Through DOF, it is then possible to identify sources of failure and predict the overall performance of an IoT system more effectively. We have built our DOF observability framework such that it is capable of predicting the overall behavior of IoT systems composed of microservices. In this thesis, we describe the concepts, related work, methodology, implementation, experimentation, and results. In addition, we present a working prototype that was developed for the purpose of clearly demonstrating the usefulness of this proposed framework. Throughout the thesis, we discuss possible applications and impact of this research work in the development and deployment of IoT systems and how our research work can be incorporated within the current paradigm of distributed tracing