Doctor of Philosophy (PhD)
Modern human-machine systems such as microservices rely upon agile engineering practices which require changes to be tested and released more frequently than classically engineered systems. A critical step in the testing of such systems is the generation of realistic workloads or load testing. Generated workload emulates the expected behaviors of users and machines within a system under test in order to find potentially unknown failure states. Typical testing tools rely on static testing artifacts to generate realistic workload conditions. Such artifacts can be cumbersome and costly to maintain; however, even model-based alternatives can prevent adaptation to changes in a system or its usage. Lack of adaptation can prevent the integration of load testing into system quality assurance, leading to an incomplete evaluation of system quality.
The goal of this research is to improve the state of software engineering by addressing open challenges in load testing of human-machine systems with a novel process that a) models and classifies user behavior from streaming and aggregated log data, b) adapts to changes in system and user behavior, and c) generates distributed workload by realistically simulating user behavior. This research contributes a Learning, Online, Distributed Engine for Simulation and Testing based on the Operational Norms of Entities within a system (LODESTONE): a novel process to distributed load testing by modeling and simulating user behavior. We specify LODESTONE within the context of a human-machine system to illustrate distributed adaptation and execution in load testing processes. LODESTONE uses log data to generate and update user behavior models, cluster them into similar behavior profiles, and instantiate distributed workload on software systems. We analyze user behavioral data having differing characteristics to replicate human-machine interactions in a modern microservice environment. We discuss tools, algorithms, software design, and implementation in two different computational environments: client-server and cloud-based microservices. We illustrate the advantages of LODESTONE through a qualitative comparison of key feature parameters and experimentation based on shared data and models. LODESTONE continuously adapts to changes in the system to be tested which allows for the integration of load testing into the quality assurance process for cloud-based microservices.
Parrott, Chester Ira, "Distributed Load Testing by Modeling and Simulating User Behavior" (2020). LSU Doctoral Dissertations. 5436.
Artificial Intelligence and Robotics Commons, Data Science Commons, Software Engineering Commons