Observability: The Missing Piece of Management in NFV-based Network Environments

Networks are dynamic entities, as the demands on their resources, functions, and services vary over time, leading to continuous changes in their state. In recent years, the Network Function Virtualization (NFV) paradigm has emerged as a practical approach to address this dynamic nature, offering high flexibility, elasticity, and mobility for managing network functions and services. To fully leverage this flexibility, it is essential to observe the network environment by monitoring metrics that reflect its state, enabling the inference of complex indicators and trends. Despite its importance, observability in NFV-based networks remains underexplored in the literature. Existing works often lack discussions that explicitly integrate observability concepts into the NFV reference architecture’s working domains and operational elements, as well as analyses of the practical implications of enforcing them. This paper presents an investigation of the application of observability in the context of NFV, linking the fundamentals of observability with the latest NFV architectures and technologies. It also demonstrates, through a case study on detecting network state changes caused by malicious activity, the impact of adopting different observability measurement models. Our results show that distinct measurement models present varying overheads on network traffic and network function instances, underscoring the importance of a well-planned integration of observability in NFV-based networks.