Reactive Prediction Models for Cloud Resource Estimation

The main objective of this thesis is to propose a new reactive resource estimation model to achieve more reasonable resource allocation and higher server utilization for cloud providers. The model is flexible enough to adapt to different situations given the current server utilization, the customer loyalty, the price of the service, etc. More precisely, four mathematical models are first proposed to deal with different situations. Then, a reactive model combining these four models is introduced. Simulations based on CloudSim are designed and implemented. The simulation results for all models meet the expectations derived from the mathematical analysis. Finally, the resource utilization of the combined model is improved by 25% in a real-time simulation compared to previous work.