Cloud Adaptation with Control Theory in Industrial Clouds

The volatility of web software systems, for example due to traffic fluctuations, can be addressed through cloud resource elasticity. Cloud providers offer specific services to automate the process of elasticity, so that application developers can efficiently and effectively manage their cloud resources. Current autoscaling methods mostly employ rule-based or threshold-based techniques. In this work, we discuss a more sophisticated and robust method based on control theory. We present the design for a simple controller and how it can be applied on real cloud environments. We demonstrate the applicability of our controller by deploying it on two cloud environments, one public and one private. Our experiments show that the same controller functions accordingly and maintain the set performance goal in both environments, indicating the potential portability of the controller across clouds.