A Collaborative and Distributed Learning-Based Solution to Autonomously Plan Computer Networks

The high programmability provided by Software Defined Networking (SDN) paradigm facilitated the integration of Machine Learning (ML) methods to design a new family of network management schemes. Among them, we can cite self-driving networks, where ML is used to analyze data and define strategies that are then translated into network configurations by the SDN controllers, making the networks autonomous and capable of auto-scaling decisions based on the network’s needs. Despite their attractiveness, the centralized design of the majority of proposed solutions cannot keep up with the increasing size of the network. To this end, this paper investigates the use of a multiagent reinforcement learning (MARL) model for auto-scaling decisions in an SDN environment. In particular, we study two possible alternatives for distributing operations: a collaborative one, where controllers share the same observations, and an individual one, where controllers make decisions according to their own logic and share only some basic information, such as the network topology. After an experimental campaign performed both on Mininet and GENI, results showed that both approaches can guarantee high throughput while minimizing the set of active resources.