Research On Data Center Energy Consumption Based On Server Power Consumption Model

In the context of global information explosion,big data has become an important basic strategic resource for the country,leading a new round of scientific and technological innovation,and promoting economic transformation and development.With the rapid development of cloud computing,big data,artificial intelligence,blockchain,5G and other technologies,the number and scale of data centers,as the carrier of data,are increasing,and the problem of energy waste is becoming more and more prominent.The call for energy conservation is rising.How to realize the thermal management and energy management of data centers on the premise of ensuring the safe operation of IT equipment to achieve the pursure of energy saving of data centers has become the focus of the industry.The energy consumption model of data centers is the key to realize thermal management and energy management.IT equipment and cooling sysytems comsume the most of electricity in data centers.Cooling systems serve for IT equipment,hence,its energy consumption depends directly on the amount of heat dissipation of IT equipment.Thus,it can be seen that accurate prediction of IT equipment energy consumption is the basis of the establishment of energy consumption model of data centers,and the guarantee for data centers to achieve fine thermal management and energy management.The study is mainly focus on the energy consumption model and its application of data centers based on the power consumption characteristics of the server.Research on the power trend and the characteristics of existing power consumption models of servers.In this paper,a total of 641samples of servers are collected to analyze the relationship and development trend among idle power,peak power and rated power of servers by considerding different heights and CPU numbers.The existing power consumption models are categorized as additive models,BA models,and other models based on calculation formula.The main variables of 47 statistic models are summarized,and it can be seen that CPU utilization is the most widely used variable in the power consumption models,followed by CPU frequency,performance counters,and temperature.Additionally,the accuracy of the models is compared,and the applicability,advantages and disadvantages of each model are analyzed based on the fitting factors,accuracy and simplicity of the model,which lays a foundation for the establishment of the server power consumption model.Research on model of server power consumption coupled with inlet air temperature in different operating modes.As the basic unit of heat flow and energy flow in data centers,the power characteristics of the server not only affect the cooling capacity of cooling systems,but also influence the overall energy consumption of data centers.The server experiment platform of power consumption and heat dissipation is built,and 176experiments are carried out to study the characteristics of power and airflow rate under different operation modes,by using SPECpower＿ssj2008 to change the load of the server,temperature controller to change the server inlet air temperature.On this basis,in order to enable the model of server power consumption to be applied in different application scenarios,CPU utilization and average temperature are coupled with the inlet air temperature to establish the model of server power consumption.Study on the balance between airflow supply and demand of CRACs and servers.Different from conventional cooling system,the cooling capacity and airflow rate requirement need to be met between cooling system and servers in data centers.In order to accurately calculate the airflow rate,a two-dimensional diagram of airflow and heat transfer in the data center is put forward based on the airflow demand of the server.Theoretical analysis and calculation of the influence of leakage airflow and bypass airflow on the airflow supply of CRACs,and the functional relationship between the air temperature difference of CRACs and that of servers are also established,which can be expressed as(T_so-T_si)/（T_r-T_s）=β（1+η）（1+ψ）and is the criterion for the balance of air supply and demand between CRACs and IT equipment.Besides,the effects of leakage and bypass airflow on the thermal environment of the data center are studied by using 6Sigma Room software.The results show that when the airflow of CRACs considering the leakage airflow and the bypass airflow,the average maximum temperature of the thermal environment of the data center is 2.5°C lower than that without consideration.Study on energy consumption model of data centers and error amplification of power consumption model of servers.In this artical,the energy consumption model of the data center is raised by referring to bottom-up calculation,combining power consumption model of servers,airflow supply-demand balance between CRACs and servers,and the static heat transfer model.According to the parameters involved in the energy consumption model of the data center,a monitoring framework for the prediction of energy consumption is constructed.Additionally,the power errors between the predictive values by proposed model and linear regression model and measured values are calculated and analyzed.On this basis,the overall energy consumption error is also studied,in the case of different supply airflow rate of CRACs and inlet air temperature of servers.The results show that the overall data center energy consumption errors caused by the proposed model and linear regression model are-1.32%～-1.24%and-5.55%～-3.10%,respectively.A quantitative study on the impact of various energy-saving measures on the overall energy consumption in the data center.The common energy saving measures of data centers are summarized.According to the energy consumption model proposed in this paper,operation modes,electric load,the average CPU utilization,operation condition,inlet air temperature,floor gap ratio,open ratio of raised floor,blank panel installation,PLR of chiller,inlet water temperature of condenser and free cooling,a total of 11 factors,2888 cases are considered in the energy calculation and analysis.By changing different levels of factors,the energy saving is calculated,and then the priority of the application of energy saving measures in IT equipment,cooling system and airflow organization is obtained,which provides a theoretical basis for energy management in a data center.