Systematic literature review of AI/ML in software-defined networks using the snowballing approach

Abstract

Current networks need to host an array of heterogeneous devices with different resource requirements and traffic outputs while maintaining acceptable QoS. To meet these requirements, networks have become increasingly more complex and difficult to manage, con gure and monitor. To make networks more easily manageable and controllable, researchers and operators proposed to use software programs that can monitor the network and configure it on-demand automatically. With Software-De ned Networks, we can build programs to efficiently manage the network through intelligent algorithms. In this study we conducted a systematic analysis focused on the use of AI/ML algorithms to improve SDN functions. We used a snowballing approach to organize and select articles to review. Following the analysis of 1200 articles (and the acceptance of 38), we present an overview of the state-of-the-art.