Testing a Q-learning approach for derivation of scaling policies in cloud-based applications

In this demonstration, we show the applicability of a new management paradigm based on Reinforcement Learning approach for the control of systems' behavior in complex dynamically evolving environments, without requiring preliminary specifications of the system models. The learning agent identifies the most adequate control policies in live interaction with a partially observed system and provides it with autonomous management capabilities. We present an experimentation with a simulated and a real Cloud-based application, and compare the results with other approaches.