Decision provenance in a real-time system with microservice architecture

Abstract

This thesis investigates the feasibility of decision provenance, by implementing it in Carmenta TrafficWatch, a real-time system with a microservice architecture. We add provenance by creating a separate service that listens to the message-bus and stores all events and potential changes to those events. This provenance information can later be queried by other services. To have provenance of how events interact with each other, we also implement a system for event aggregation. Our findings indicate that adding provenance to real-time systems with microservice architecture will not impact the overall performance of the system significantly, and is a viable solution to add accountability and increase the understanding of complex systems.