Self-Healing Cloud Applications

Abstract

Cloud computing offers on-demand services to deploy and run applications with flexible and scalable resource pooling. The techniques adopted by cloud systems introduce reliability issues that challenge the design of cloud applications. In my PhD I work on the key problem of improving the reliability of cloud applications. In particular, I am investigating on the definition of effective and efficient self-healing approaches that integrate failure prediction, fault localization and fault fixing mechanisms in the context of cloud-based systems. In the first part of my research I investigated the problem of automatic failure prediction, which constitutes the first step of a complete self-healing approach. I identified an original approach based on a combination of data analytics and machine learning techniques, and developed an early prototype to collect experimental data about the proposed approach. The data collected so far indicate that the approach has both high precision and recall rate.