JP2017151617A - Simulation device, simulation method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable electric power on a data center to be simulated using a power consumption model for individual apparatuses of the data center.SOLUTION: The present invention is a simulation device comprising: a first power consumption calculation unit (111) for holding a server power consumption model for each type of the server and, upon acquiring the type information and operation amount of the server, deriving the power consumption and temperature of the server after operation of the operation amount at a set intake air temperature using a server power consumption model that corresponds to the type information; a fluid analysis unit (12) for performing a thermal fluid analysis when an air conditioner is operated in a space in which the server is installed, and deriving the intake air temperature of the server after operation; and a second power consumption calculation unit (112) for holding an air conditioning power consumption model for each air conditioner, and deriving the power consumption of an air conditioner in order to bring the space to the intake air temperature after updating by the fluid analysis unit.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a simulation apparatus, a simulation method, and a program for predicting power consumption in a data center.

  With increasing demand for services via data centers such as cloud services, the power consumption is increasing. If it is possible to grasp the power consumption from the operation status of the data center, a more efficient operation policy can be determined.

  In order to ascertain (predict) the power of the data center, all “component power” must be reflected in the state of reflection of all dependencies such as the operating status of all the servers to be configured, their positional relationships, and air conditioning conditions. Needs to be calculated. For this purpose, it is necessary to learn in advance the relationship between the operating state and the actual power consumption at that time, collect huge amounts of data, and then predict the power from the actual operating conditions.

  However, elements such as servers that make up the data center are composed of many dependencies such as the operating status of other servers, their positional relationships, or air conditioning conditions, and it is difficult to estimate the power consumption of the entire data center. For this reason, when a new device is arranged, the arrangement is changed, or the operation is performed in a state different from the previous learning, it is necessary to perform learning again.

Yoshihide Suwa, “Study on High Efficiency of Air-Conditioning Airflow System at Data Center”, Architectural Institute of Japan Environmental Proceedings, Vols. 501-508, May 2011

  Therefore, an object of the present invention is to enable power simulation of a data center using a power consumption model of individual devices in the data center.

  In order to achieve the above object, the present invention makes it possible to perform power simulation of a data center by preparing individual power consumption models of devices constituting the data center and linking the models with the numerical fluid simulator. To do.

Specifically, the simulation apparatus according to the present invention is:
When a heat source device power consumption model that indicates the power consumption and temperature change with respect to the operating amount of the heat source device is held for each model of the heat source device and the model information and operating amount of the heat source device are acquired, the heat source device power consumption corresponding to the model information A first power consumption calculation unit for deriving power consumption and temperature of the heat source device after operating the operating amount at a set intake air temperature using a model;
Using the power consumption and temperature of the heat source device derived by the first power consumption calculation unit, a thermal fluid analysis is performed when the air conditioner is operated in the space where the heat source device is arranged, and the operation amount is operated. A fluid analysis unit for deriving an intake air temperature of a later heat source device and updating the set intake air temperature;
An air conditioner power consumption model indicating power consumption and temperature change with respect to the operating amount of the air conditioner is maintained for each air conditioner, and the power consumption of the air conditioner for deriving the space to the intake air temperature updated by the fluid analysis unit is derived. A second power consumption calculation unit,
Is provided.

Specifically, the simulation method according to the present invention includes:
When the heat source device model information and operation amount are obtained by referring to the storage unit that holds the heat source device power consumption model for each heat source device model indicating the power consumption and temperature change with respect to the operating amount of the heat source device, it corresponds to the model information. A first power consumption calculation procedure for deriving power consumption and temperature of the heat source device after operating the operation amount at the set intake air temperature using the heat source device power consumption model
Using the power consumption and temperature of the heat source device derived in the first power consumption calculation procedure, a thermal fluid analysis is performed when the air conditioner is operated in the space where the heat source device is arranged, and the operation amount is operated. A fluid analysis procedure for deriving an intake air temperature of a later heat source device and updating the set intake air temperature;
An air conditioner for referring to a storage unit that holds an air conditioner power consumption model indicating a change in power consumption and temperature with respect to an operating amount of the air conditioner for each air conditioner and setting the space to the intake air temperature updated in the fluid analysis procedure A second power consumption calculation procedure for deriving power consumption of
Are executed by the simulation apparatus.

  Specifically, the simulation program according to the present invention is a program for causing a computer to execute the first power consumption calculation procedure, the fluid analysis procedure, and the second power consumption calculation procedure described in the simulation method according to the present invention. is there.

  According to the present invention, a power simulation of a data center can be performed by linking a power consumption model of a device constituting the data center and a numerical fluid simulator.

The structural example of the simulation apparatus which concerns on embodiment is shown. An example of a server power consumption model is shown. An example of an air-conditioning power consumption model is shown. The structural example of the data center which performs CFD simulation is shown. An example of the power consumption of the data center estimated using embodiment is shown. An example of the comparison result of the predicted value predicted using the embodiment and the actually measured value is shown. An example of the error between the predicted value and the actually measured value is shown.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In addition, this invention is not limited to embodiment shown below. These embodiments are merely examples, and the present invention can be implemented in various modifications and improvements based on the knowledge of those skilled in the art. In the present specification and drawings, the same reference numerals denote the same components.

  In the embodiment, an example will be described in which a data center includes a heat source device serving as a heat source such as a server and an air conditioner for cooling the heat source device. In the present embodiment, for easy understanding, an arbitrary heat source device disposed in a data center such as a server or a switch is referred to as a “server”. Accordingly, the heat source device power consumption model is referred to as a server power consumption model.

  Since the server occupies most of the heat source of the data center, the prediction of the exhaust temperature of the server is important in predicting the temperature in the data center. The exhaust temperature of the server depends on the amount of heat generated by the server and the amount of air passing through the server. In the data center, the intake air temperature and the air volume differ depending on the server position due to the influence of the device layout. Even if the same task is assigned to the same model server, the server exhaust temperature varies depending on the server position.

  Computational Fluid Dynamics (CFD) is a representative technique for simulating the temperature in a data center in a situation where such interdependencies of various parameters exist. CFD is an analysis method that numerically solves the governing equations of fluid dynamics. Since the CFD simulation calculates the operation of the fluid for the entire analysis region, it is also suitable for analysis of a local air current such as a data center. However, this CFD can simulate temperature, but cannot simulate power.

  The simulation device 91 according to the embodiment prepares an individual power model of each device provided in a data center such as a server or an air conditioner, and combines all the states with a technique for predicting (simulating) the model. There is no need to learn in advance, and the total power consumption is estimated in advance for any new arrangement or operating condition.

  FIG. 1 shows a configuration example of a simulation apparatus 91 according to the embodiment. The simulation apparatus 91 according to the embodiment includes a power consumption model calculation unit 11 and a CFD simulation calculation unit 12. The power consumption model calculation unit 11 includes a server power consumption calculation unit 111 and an air conditioning power consumption calculation unit 112. The server 91 and the air conditioner operation data are input to the simulation device 91 in advance (S1), and server information and air conditioning setting information are input when the simulation is executed (S2).

  The server operation data is data obtained by collecting, for each server, information related to the power consumption of the servers arranged in the data center. The server power consumption calculation unit 111 creates a power consumption model (server power consumption model) for each model by machine learning using server operation data (S1). The operation data of the air conditioner is data obtained by collecting information on the power consumption of the air conditioner arranged in the data center for each air conditioner. The air conditioning power consumption calculation unit 112 creates an air conditioner power consumption model (hereinafter referred to as an air conditioning power consumption model), which is a power consumption model for each air conditioner model, by machine learning using air conditioner operation data. (S1). These models are obtained by operating in advance under various conditions and measuring the power.

  FIG. 2 shows an example of the server power consumption model. The server power consumption model indicates power consumption (W) and temperature (° C.) with respect to the CPU operation rate (%) of the server. The server power consumption model is constructed by operating a server alone. It is only necessary to create a server power consumption model for each server model, and it is not necessary to create a model using a server that is actually arranged in the data center.

  FIG. 3 shows an example of the air conditioning power consumption model. The air conditioning power consumption model indicates temperature (° C.) and power efficiency (COP: Coefficient Of Performance) with respect to the operating amount of the air conditioner. Here, in FIG. 3, as an example of the operation amount of the air conditioner, the ratio (%) with respect to the maximum rotation speed of the fan is shown. The temperature is a supply temperature of air supplied from the air conditioner to the data center. It is only necessary to create an air conditioning power consumption model for each air conditioner model, and it is not necessary to create a model using the air conditioning actually arranged in the data center.

  The server information includes the task amount allocated to each server, the task allocation status to the server, the operation pattern of each server, and the intake air temperature of each server. The air conditioning setting information includes the settings of the air conditioner used for the simulation. Settings include temperature, wind direction, and wind power.

  When acquiring the server information and the air conditioning setting information, the simulation device 91 inputs the server information and the air conditioning setting information to the power consumption model calculation unit 11 and inputs the air conditioning setting information to the CFD simulation calculation unit 12 (S2).

  The server power consumption calculation unit 111 functions as a first power consumption calculation unit, and uses the server power consumption model to operate with the operation pattern specified in the server information at the intake air temperature specified in the server information. The power consumption and heat generation amount of each server are calculated for each server (S3).

  The CFD simulation calculation unit 12 functions as a fluid analysis unit, and executes a CFD simulation (thermal fluid analysis) using the estimated heat generation amount and air conditioning setting information of each server (S3). Here, the CFD is a numerical analysis / simulation method for observing a flow by solving various equations relating to fluid motion with a computer, and is particularly used as a method for simulating a fluid temperature distribution (for example, non-patent literature). 1). As a result, when the server in the data center reaches the power consumption and heat generation amount calculated by the server power consumption calculation unit 111 in an environment where the air conditioner operates under the conditions specified in the air conditioning setting information, the intake air of each server Derived how many times the temperature is.

  FIG. 4 shows a configuration example of a data center that performs CFD simulation. An example is shown in which 12 servers 83 and one air conditioner 82 are arranged in the data center 81. The air conditioner 82 takes in air from the air inlet 821, cools the taken-in air, and supplies the air to the data center 81 from the back side. According to the CFD simulation, in the data center 81, the temperature is about 50 ° C. in the region A1, about 45 ° C. in the region A2, about 40 ° C. in the region A3, about 35 ° C. in the region A4, and about 30 ° C. in the region A5. It was before and after. In this case, the intake air temperature of the server 83 # 3 is around 40 ° C. in the region A3, and the intake air temperature of the server 83 # 5 is around 40 ° C. in the region A4. In this way, the CFD simulation calculation unit 12 derives the intake air temperatures of the servers 83 # 1 to # 12.

  Steps S2 and S3 are repeated until the state inside the data center converges (S4). At this time, the server power consumption calculation unit 111 estimates the power consumption of each server using the server power consumption model from the intake air temperature of each server obtained from the result of the CFD simulation and the assigned task amount. The CFD simulation calculation unit 12 obtains the return air temperature to the air conditioning by the CFD simulation in which the power consumption of each server and the setting information of the air conditioning are set as parameters. As a result, the power consumption of each server and the return air temperature of the air conditioning in the data center can be obtained.

  When the state inside the data center converges, the air conditioning power consumption calculation unit 112 executes step S5. For example, the temperature difference between the intake air temperatures of the servers 83 # 1 to # 12 used by the server power consumption calculator 111 and the intake air temperatures of the servers 83 # 1 to # 12 output from the CFD simulation calculator 12 is as follows. When it is within the set constant temperature, step S5 is executed. Whether or not the temperature difference is within a certain temperature is determined for each server. For all servers, when the temperature difference falls within a certain temperature, it can be determined that the state inside the data center has converged.

  After that, the air conditioning power consumption calculation unit 112 functions as a second power consumption calculation unit, and uses the air conditioning power consumption model in the condition specified by the air conditioning setting information to return the air conditioning obtained from the result of the CFD simulation. The power consumption of the air conditioner at the air temperature is estimated (S5).

  The power consumption model calculation unit 11 sums the estimated power consumption values of each server and each air conditioner (S6). Thereby, the simulation apparatus 91 can estimate the power consumption of the data center.

  FIG. 5 shows an example of power consumption of the data center predicted using the embodiment. L1 is the total power consumption of the server, L2 is the total power consumption of the air conditioner, and L3 is the total power consumption value of the data center. It can be seen that the power consumption of the data center varies depending on the set temperature of the air conditioner.

  FIG. 6 shows the relationship between the actual power value of the server and the estimated value of the server power value obtained based on the simulation method of the embodiment. FIG. 7 is a histogram (occurrence frequency distribution) representing the relationship between the actually measured values and the estimated values in FIG. From these two graphs, it can be seen that the power value of the server can be estimated with extremely high accuracy by the simulation method of the present embodiment, and consequently the power of the entire data center can also be estimated with high accuracy.

  The data center is not limited to video-on-demand or cloud storage, and includes an arbitrary space in which a plurality of heat source devices are arranged and temperature adjustment by air conditioning is desirable.

  The simulation apparatus 91 according to the embodiment may be realized by causing a computer to function as the power consumption model calculation unit 11 and the CFD simulation calculation unit 12. In this case, each configuration is implemented by a CPU (Central Processing Unit) in the simulation apparatus 91 executing a computer program stored in a storage unit (not shown). The computer program may be recorded on a computer-readable recording medium.

  By deallocating tasks to some servers and stopping unused servers, server power consumption can be reduced. Moreover, the power efficiency of the air conditioner can be improved by appropriately setting the air conditioner based on the temperature distribution in the data center.

  Since the simulation apparatus 91 according to the embodiment can control the devices in the data center in a coordinated manner, the power reduction of the entire data center can be achieved in an environment in which various devices are operating simultaneously, such as a data center. Can do.

  Here, the simulation apparatus 91 according to the embodiment is configured in advance by combining a technology for predicting (simulating) an individual power consumption model for each component (server or air conditioner) in advance. Therefore, it is possible to grasp the entire power in advance for any new arrangement or operating condition.

  In addition, since the simulation apparatus 91 according to the embodiment can grasp the total power reflecting each interaction and relative position only by preparing individual power consumption models of each component, Regardless of any element (such as a server) that has no track record of installation, even if the configuration is changed, it is possible to grasp the entire power when it is mounted without operating the entire system in advance.

  The present invention can be applied to the information communication industry.

11: Power consumption model calculation unit 111: Server power consumption calculation unit 112: Air conditioning power consumption calculation unit 12: CFD simulation calculation unit 81: Data center 82: Air conditioner 821: Inlet 83: Server 91: Simulation device

Claims (3)

When a heat source device power consumption model that indicates the power consumption and temperature change with respect to the operating amount of the heat source device is held for each model of the heat source device and the model information and operating amount of the heat source device are acquired, the heat source device power consumption corresponding to the model information A first power consumption calculation unit for deriving power consumption and temperature of the heat source device after operating the operating amount at a set intake air temperature using a model;
Using the power consumption and temperature of the heat source device derived by the first power consumption calculation unit, a thermal fluid analysis is performed when the air conditioner is operated in the space where the heat source device is arranged, and the operation amount is operated. A fluid analysis unit for deriving an intake air temperature of a later heat source device and updating the set intake air temperature;
An air conditioner power consumption model indicating power consumption and temperature change with respect to the operating amount of the air conditioner is maintained for each air conditioner, and the power consumption of the air conditioner for deriving the space to the intake air temperature updated by the fluid analysis unit is derived. A second power consumption calculation unit,
A simulation apparatus comprising: When the heat source device model information and operation amount are obtained by referring to the storage unit that holds the heat source device power consumption model for each heat source device model indicating the power consumption and temperature change with respect to the operating amount of the heat source device, it corresponds to the model information. A first power consumption calculation procedure for deriving power consumption and temperature of the heat source device after operating the operation amount at the set intake air temperature using the heat source device power consumption model
Using the power consumption and temperature of the heat source device derived in the first power consumption calculation procedure, a thermal fluid analysis is performed when the air conditioner is operated in the space where the heat source device is arranged, and the operation amount is operated. A fluid analysis procedure for deriving an intake air temperature of a later heat source device and updating the set intake air temperature;
An air conditioner for referring to a storage unit that holds an air conditioner power consumption model indicating a change in power consumption and temperature with respect to an operating amount of the air conditioner for each air conditioner and setting the space to the intake air temperature updated in the fluid analysis procedure A second power consumption calculation procedure for deriving power consumption of
Is a simulation method in which the simulation apparatus executes.   A program for causing a computer to execute the first power consumption calculation procedure, the fluid analysis procedure, and the second power consumption calculation procedure according to claim 2.

Images