Machine learning applied to chaos engineering

Abstract

Internet-scale systems need to assure their quality, researchers and practitioners have been working on approaches and tools for monitoring, profiling, and testing of internet-scale systems. One of those approaches for testing is Chaos Engineering, which tackles different challenges for the software reliability community.

This thesis aims to improve chaos engineering capabilities with a machine learn- ing model for early detection of system metrics during a chaos engineering experiment. A systematic literature review and an empirical study based on the study insights' is proposed to demonstrate the feasibility of augmenting chaos engineering with machine learning.