Energy Modeling Method for Containerized Applications on Physical Hosts

The energy consumption problem in data centers is becoming increasingly severe, and energy consumption models can reveal the relationship between data center energy consumption and its corresponding factors, which is crucial for reducing energy consumption. Containerization is the current mainstream trend in data centers. Therefore, this paper proposes energy consumption models for physical hosts, containers, and containerized applications. First, we constructe a physical host energy consumption model based on polynomial regression with lasso. Then, by analyzing the resource utilization, we propose a simple and accurate method for modeling the container energy consumption. Finally, based on the container energy consumption model, we propose a method for constructing a parameter-based energy consumption models for containerized applications. The experimental results show that our modeling methods are valid. The advantages of our modeling method are that it can analyze data center energy consumption from different levels, such as physical hosts, containers, and applications, and it has the advantages of simplicity, high accuracy, ease of implementation.