ML Enhanced Control Planes for Adaptive Distributed Systems

The growth of adaptive distributed systems requires smarter and more resilient control planes that adapt to system behaviors and failures.Machine Learning (ML) is becoming an important enabler for advancing these control planes' capabilities with predictive analytics, failure detection, energy efficiency, and efficient data routing.This paper reviews recent research on ML-enabled control planes in adaptive distributed system contexts, including fault tolerance, data replication, energy management, optical networking, and security.The authors systematically reviewed some of the most significant research articles in these areas, and discuss how ML can be embedded within different layers of control plane architecture, and its consequent impact on performance, scalability, reliability, and so on.The review points out important trends, issues, and technology that impact how control planes are developing and incorporates some aspect of reinforcement learning, active learning, and root cause analysis frameworks.The review also reinforces the ability of ML to make distributed systems resilient and context aware, as well as the important need for resilient, secure ML models.In contrast to other reviews that have pointed out the power of machine learning in disconnect domains, this paper takes an integrated view that correlates fault tolerance, the energy efficiency and security and optical networks with ML-enabled control planes-beginning to fill a significant gap in the literature with respect to adaptive distributed systems and supporting a holistic view.The insights also serve as a lead into direction for future research with respect to self-optimizing and secure machine learning-based control systems.