Research and Application of a Framework for Evaluating Energy Efficiency in Internet Data Center

With various energy-saving technologies are introduced into computer room scenarios. The evaluation of energy-saving effects in data centers has become an important part of judging the effectiveness of relevant methods. The traditional energy-saving evaluation method is T-test which evaluates the changes in energy consumption that before and after implementation as the basis for energy-saving effectiveness. However, there are external environmental disturbances in the computer room environment, such as the implementation of other energy-saving technologies, load changes and etc. which make the evaluation cycle cannot be too long. A shorter evaluation period will lead to lower confidence, and the energy-saving effect of the data center is generally small. Therefore, changes in energy consumption caused by environmental changes will have an impact on the energy efficiency evaluation, thereby affecting the overall energy-saving effect evaluation. With the introduction of Reinforcement learning and other technologies, energy efficiency assessment has entered the fully automatic mode, and it is necessary to automatically identify the scenarios of energy consumption changes caused by the external environment within the assessment cycle. This article proposes a set of energy-saving effectiveness evaluation framework by calculating whether the change point position and implementation day within the evaluation period are the same day and combining confidence. Through experiments, it is shown that under this framework, the accuracy of identifying abnormal data is 77%, and the recall rate is 100%.