Dynamic Frequency Adjustment and Dispatching for Cloud Computing System with Energy Cost Considerations

This research studies the dynamic frequency adjustment and task dispatching problems for cloud computing systems under user demand uncertainties. The objective of the research is to minimize energy consumption and waiting costs. A dynamic programming model is presented to minimize total costs by adjusting CPU frequency dynamically. In this model, we implement real-time job dispatching based on the current state. With the objective of minimizing the total waiting time for tasks. Additionally, the model employs a stochastic process to accurately estimate the transition probability between various events. Backward induction is used to solve the model and make the optimal decision for the different states. Compared with other control methods from the literature, our numerical results show that the proposed method reduces total energy and waiting costs. The improvement is particularly significant when cloud servers and tasks have higher heterogeneity. This approach provides a comprehensive solution to the challenges of cloud computing, balancing performance and cost efficiency.