Observabilité et gestion des ressources dans les environnements cloud-natifs
Nicolas Marie-Magdelaine · HAL (Le Centre pour la Communication Scientifique Directe) · 2021
Cloud Computing and Cloud-Native technologies have become the backbones of the modern Internet. Users and organizations are now relying on cloud applications for their everyday needs. However, outages and Quality of Service degradations can have disastrous impacts on our society. Moreover, web applications have become complex distributed systems, difficult to understand and operate, thus more prone to failures if not managed accordingly. As a result, it is paramount to understand, observe, prevent, detect and correct any issues that may lead to failures.In this thesis, we propose a framework for achieving Observability in cloud-native environments. Observability envisions a deeper understanding of the complex distributed system that web applications have become, representing an improvement from traditional monitoring strategies. The proposed Observability framework is demonstrated via proof-of-concept and deployment in production environments. Furthermore, following the principle of autonomic computing, we also propose an architecture for Observability-driven auto-scaling in Cloud-Native environments. This architecture enables us to correlate auto-scaling to the application workload. We also push a step further by leveraging Machine Learning and enable proactive auto-scaling. We focus on using automation and Observability to increase the Quality of Service metrics during scaling events. Additionally, we explore and demonstrate the benefits and possibility of porting the cloud-native architecture and principles to the Internet of Things. We propose to leverage technologies such as Software-Defined Networking and Software-Defined Radio to provide flexible and re-configurable generic Internet of Things devices.