Reinforcement Learning Based Service Provisioning for a Greener Cloud

Cloud computing is an emerging distributed computing model consisting of massive data enters for making different services available to the users. In the current scenario where energy consumption and wastage in the IT field is looked upon with growing apprehension, Green Computing encourages the design of energy-efficient computing approaches that can be applied to cloud computing to address and reduce the factors which influence power consumption alias energy cost. Other evolving technologies like Virtualization and VM (Virtual Machine) migration technologies are employed widely for energy efficient consolidation of resources. The existing work on green cloud service provisioning aids energy aware cloud service provisioning by incorporating the Trigger Engine Agent which uses the static pre-processed information of service usage to initiate live VM migration. This paper proposes to take the dynamic environment into consideration to substantiate the decisions made by the existing model and incorporate learning agents into the model. Our model comprises of two parts: a dynamic Post-Processing Agent and a Learning Agent. The Post-Processing Agent verifies the migration decisions made by the Trigger Engine and corrects them if they do not conform to the energy-aware provisioning approach. The significant portion of this work is the Learning Agent, which will learn the optimal policy to follow in the current environment by incorporating the actions of the Post-Processing agent into the pre-processed data of the existing system using Q-learning methodology.