A collaborative control optimization method for near-field wind wall micromodules

Abstract

The present invention provides a collaborative control optimization method applicable to near-field wind wall micromodules, comprising the following steps: Step 10, obtaining IT system scheduling data and air conditioning system data of the near-field wind wall micromodule within a preset time period, and respectively establishing a micromodule server real-time power consumption model, an air conditioning system power consumption model, and a server temperature rapid prediction model; Step 20, based on the micromodule server real-time power consumption model, the air conditioning system power consumption model, and the server temperature rapid prediction model, adopting an optimization method combining real-time combination and timing decoupling to collaboratively control virtual machine deployment and the air conditioning system. The collaborative control optimization method applicable to near-field wind wall micromodules provided by the present invention can solve the technical problems of local hot spots and overcooling that are prone to occur in energy-saving scheduling of near-field wind wall micromodules, so as to improve the overall energy efficiency and operation stability of the near-field wind wall micromodules.

Description

Technical Field

[0001] The present invention belongs to the technical field of air cooling control of data centers, and specifically, relates to a collaborative control optimization method applicable to near-field wind wall micromodules. Background Art

[0002] Near-field wind wall air conditioner is one of the data center air cooling technologies with the highest heat exchange efficiency at present, and it is widely used in various high-power density data centers. However, in actual engineering, the energy-saving scheduling of the IT system and the cooling system of the near-field wind wall micromodule is currently executed separately, that is, the cloud platform of the IT system usually adopts the scheduling strategy of the least active servers, while the near-field wind wall air conditioner adopts a feedback control strategy based on the supply / return air temperature (usually controlling the fan frequency according to the supply air temperature, and controlling the coil evaporation temperature according to the return air temperature). There is a lack of effective coordination between the two, which leads to problems such as local hot spots and overcooling, resulting in the increase and decrease of energy consumption of the IT system and the cooling system, and has a great impact on the safety and stability of the data center.

[0003] In order to improve the precise cooling capacity of near-field wind wall air conditioners, an effective way is to predict the chip temperature of the server based on the real-time power changes of the server in the micromodule, so as to regulate the air volume at each terminal of the near-field wind wall air conditioner and the evaporation temperature of the coil. However, due to the limitations of fluid distribution and heat capacity characteristics, the cooling system represented by the near-field wind wall air conditioner has significant time lag and control accuracy problems during the cooling process, and it is difficult to dynamically match the heat dissipation changes of each server. Therefore, this technology has failed to fundamentally solve the problem of hot and cold mismatch in data centers. Summary of the invention

[0004] The present invention provides a collaborative control optimization method applicable to a near-field wind wall micromodule, which solves the technical problems of local hot spots and overcooling that are prone to occur in the near-field wind wall micromodule during energy-saving scheduling, so as to improve the overall energy efficiency and operation stability of the near-field wind wall micromodule.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] The present invention provides a collaborative control optimization method applicable to a near-field wind wall micromodule, comprising the following steps:

[0007] Step 10, obtaining IT system scheduling data and air conditioning system data of the near-field wind wall micromodule within a preset time period, and establishing a real-time power consumption model of the micromodule server, a power consumption model of the air conditioning system, and a rapid prediction model of the server temperature;

[0008] Step 20, based on the real-time power consumption model of the micromodule server, the power consumption model of the air conditioning system and the rapid prediction model of the server temperature, an optimization method combining real-time combination and timing decoupling is adopted to coordinately control the deployment of virtual machines and the air conditioning system.

[0009] As a further improvement of the present invention, the IT system scheduling data includes the CPU resource requirements of the virtual machine, the mapping relationship between the virtual machine and the server, the CPU utilization rate of each server, the real-time power consumption of each server and the chip temperature of each server; the air-conditioning system data includes the fan frequency of each terminal module, the coil evaporation temperature, the outdoor unit condensing temperature, the load rate and the total power consumption of the air-conditioning system.

[0010] As a further improvement of the present invention, in step 10, a real-time power consumption model of the micromodule server is established according to IT system scheduling data;

[0011] Specifically include:

[0012] Step 101, filtering out power consumption data of each server under different chip temperatures and CPU usage rates from IT system scheduling data;

[0013] Step 102, establishing a server static power consumption matrix according to the power consumption data of each server when the CPU usage rate is 0;

[0014] Step 103, according to the data of the change of the real-time power consumption of each server with the CPU usage rate, a conversion coefficient between the CPU usage rate and the dynamic power consumption of each server is fitted by a multivariate linear regression method to form a conversion coefficient matrix between the CPU usage rate and the dynamic power consumption;

[0015] Step 104, according to different temperature data of the chip of each server at the same CPU usage rate, a multivariate linear regression method is used to fit the expression of the leakage power consumption of each server to form a server leakage power consumption matrix;

[0016] Step 105, combining the server static power consumption matrix, the conversion coefficient matrix of CPU usage and running power consumption, and the server leakage power consumption matrix, to obtain the real-time power consumption model of the micromodule server shown in formula (1):

[0017] P PM (t) = P SC (t)+u VM (t)*MM VM→PM (A(t),F(t))*f h +D(T PM (t)) Formula (1)

[0018] Where P PM (t) represents the server real-time power consumption matrix, P SC(t) represents the server static power consumption matrix, u VM (t) represents the CPU resource requirement matrix of the virtual machine, MM VM→PM (A(t), F(t)) represents the mapping matrix between virtual machines and servers, A(t) represents the virtual machine real-time deployment matrix, F(t) represents the virtual machine migration matrix, u VM (t)*MM VM→PM (A(t), F(t)) represents the CPU usage of the server, f h The conversion coefficient matrix between CPU utilization and dynamic power consumption, D(T PM (t)) represents the server leakage power consumption matrix.

[0019] As a further improvement of the present invention, in step 10, based on the air conditioning system data, an air conditioning system power consumption model shown in formula (2) is established:

[0020] P FC (t) = P CP (T co (t),T ev (t),PLR(t))+ΣP FX (FP(t)) Formula (2)

[0021] Where P FC (t) represents the total power consumption of the air conditioning system, P CP (T co (t),T ev (t),PLR(t)) represents the power consumption of the external unit, T co (t) represents the coil evaporation temperature, T ev (t) represents the condensing temperature of the external unit, PLR(t) represents the load factor, FP(t) represents the fan frequency matrix of each terminal module, P FX (FP(t)) represents the power consumption matrix of each terminal module.

[0022] As a further improvement of the present invention, in step 10, a server temperature rapid prediction model is established based on IT system scheduling data and air conditioning system data;

[0023] Specifically include:

[0024] Step 131, filtering out the mapping relationship between virtual machines and servers, the CPU resource requirements of virtual machines, and the chip temperature of each server from the IT system scheduling data, and filtering out the fan frequency and coil evaporation temperature of each terminal module from the air conditioning system data to form a modeling data set;

[0025] Step 132, based on the modeling data set, the fan frequency of each terminal module, the coil evaporation temperature and the CPU utilization rate of each server are used as input variables of the BP neural network model, and the chip temperature of each server is used as the output variable of the BP neural network model. The BP neural network model is trained to obtain a server temperature rapid prediction model.

[0026] As a further improvement of the present invention, in step 20, an optimization method combining real-time association and timing decoupling is used to coordinately control virtual machine deployment and the air conditioning system, including:

[0027] Step 21, when the cloud platform receives a new virtual machine deployment request, the deployment location of the new virtual machine and the fan frequency of each terminal module are jointly optimized to achieve the lowest overall power consumption of the server and each terminal module under a single virtual machine deployment;

[0028] Step 22, at each given time, first migrate the virtual machines running on each server as a whole to minimize the total power consumption of the server; then optimize the coil evaporation temperature of the air-conditioning system and the fan frequency of each terminal module to achieve the lowest overall power consumption of the air-conditioning system.

[0029] As a further improvement of the present invention, the step 21 specifically includes:

[0030] Step 211, when the cloud platform receives the deployment request of the new virtual machine, it traverses the remaining CPU resources of all active servers according to the CPU resource requirements of the new virtual machine to find a server available for deployment;

[0031] Step 212: If there are servers available for deployment, the locations of the servers available for deployment are used as optimization variables for the deployment of the new virtual machine, and step 213 is executed; if the remaining CPU resources of all active servers cannot deploy the new virtual machine, a dormant server is randomly switched to an active state, the new virtual machine is deployed, and step 214 is executed;

[0032] Step 213, with the optimization goal of minimizing the sum of the total real-time power consumption of the server after deployment and the total power consumption of the terminal module, with the optimization constraints that the sum of the resources used by all virtual machines running on each server does not exceed the total rated resources of the server, and the chip temperature of each server does not exceed the preset temperature threshold, a genetic algorithm is used to jointly optimize the deployment position of the new virtual machine and the fan frequency of each terminal module to obtain the optimal real-time deployment position of the virtual machine and the optimal fan frequency of each terminal module; the cloud platform implements the deployment of the new virtual machine according to the optimal real-time deployment position of the virtual machine, and the air conditioning system controls the fan frequency of each terminal module according to the optimal fan frequency of each terminal module;

[0033] Step 214, with the optimization goal of minimizing the sum of the power consumption of each terminal module after the new virtual machine is deployed, and with the constraint that the temperature of each server chip cannot exceed the preset temperature threshold, a genetic algorithm is used to optimize the fan frequency of each terminal module to obtain the optimal fan frequency of each terminal module; the air-conditioning system controls the fan frequency of each terminal module according to the optimal fan frequency of each terminal module.

[0034] As a further improvement of the present invention, the step 22 specifically includes:

[0035] Step 221, at a given moment, calculating the minimum number of active servers required based on the sum of the number of all virtual machines running on the server and their CPU resource requirements;

[0036] Step 222, based on the required minimum number of active servers, select servers as active servers for the next period in the order of rack numbers and server numbers on each rack, until the number of selected servers reaches the minimum number of active servers;

[0037] Step 223, according to the load balancing strategy, evenly migrate the existing virtual machines to the selected active servers; hibernate the idle servers;

[0038] Step 224, with the goal of minimizing the total power consumption of the air conditioning system, and with the temperature of each server chip not exceeding a preset temperature threshold as a constraint, a genetic algorithm is used to optimize the coil evaporation temperature and the fan frequency of each terminal module to obtain the optimal coil evaporation temperature and the optimal fan frequency of each terminal module;

[0039] Step 225 , the air conditioning system regulates the coil evaporation temperature according to the optimal coil evaporation temperature, and controls the fan frequency of each terminal module according to the optimal fan frequency of each terminal module.

[0040] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0041] The present invention provides a collaborative control optimization method applicable to a near-field wind wall micromodule. First, IT system scheduling data and air conditioning system data of the near-field wind wall micromodule within a preset time period are obtained, and a real-time power consumption model of a micromodule server, an air conditioning system power consumption model, and a server temperature rapid prediction model of the near-field wind wall micromodule are established respectively; then, in the control stage, when real-time deployment optimization is adopted, the deployment position of the newly arrived virtual machine and the fan frequency of each terminal module are jointly optimized to improve the real-time matching degree between the server heat dissipation and the air conditioning terminal cooling, so as to reduce the probability of local hot spots and overcooling in the micromodule, thereby achieving the lowest overall power consumption of the server and the air conditioning terminal under a single virtual machine deployment; when timing decoupling optimization is adopted, the virtual machines running on each server are first migrated as a whole at a given time node to minimize the total power consumption of the server. On this basis, the coil evaporation temperature of the air conditioning system and the fan frequency of each terminal module are optimized to achieve the lowest overall power consumption of the air conditioning system, and promote the timing height matching between the server heat dissipation and the air conditioning cooling, thereby further improving the utilization efficiency of the cooling capacity; through the cooperation of real-time joint optimization and timing decoupling optimization, the overall energy efficiency and operation stability of the near-field wind wall micromodule can be improved. The method of the present invention starts with the coordinated control of the server and the near-field wind wall air conditioner, aiming to fundamentally solve the cold and hot mismatch of the data center to achieve the optimal overall energy efficiency of the server and the near-field wind wall air conditioner. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0043] Figure 1 is a flow chart of a collaborative control optimization method according to an embodiment of the present invention;

[0044] Figure 2 It is a flow chart of real-time joint optimization of virtual machine deployment and air conditioning system in the method of the embodiment of the present invention;

[0045] Figure 3 It is a flow chart of regularly performing decoupling optimization on the overall migration of virtual machines and the air conditioning system in the method of the embodiment of the present invention;

[0046] Figure 4 It is a schematic diagram of the structure of a near-field wind wall micromodule to which the embodiment method of the present invention is applicable.

[0047] In the figure, there are near-field wind wall micromodule 1, rack 11, server 12, terminal module 21, fan 22, evaporator coil 23, refrigerant pipeline 24, and external unit 25. DETAILED DESCRIPTION

[0048] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0049] The embodiment of the present invention provides a collaborative control optimization method applicable to near-field wind wall micromodules, which is applied to Figure 4 In the near-field wind wall micromodule 1 shown. There are N racks 11 in the near-field wind wall micromodule 1, and each rack is provided with M servers 12. The near-field wind wall has N terminal modules 21 corresponding to the N racks 11 one by one, and the fan 22 in a single terminal module can only be uniformly frequency-converted, and the evaporator coil 23 is connected to the external machine 25 (including the condenser, the compressor and the throttling device) through the refrigerant pipe 24.

[0050] The embodiment of the present invention provides a collaborative control optimization method applicable to near-field wind wall micromodules, such as Figure 1 As shown, the following steps are included:

[0051] Step 10, obtain IT system scheduling data and air-conditioning system data of the near-field wind wall micromodule within a preset time period; establish a real-time power consumption model of the micromodule server of the near-field wind wall micromodule based on the IT system scheduling data; establish an air-conditioning system power consumption model of the near-field wind wall micromodule based on the air-conditioning system data; establish a server temperature rapid prediction model of the near-field wind wall micromodule based on the IT system scheduling data and the air-conditioning system data.

[0052] Step 20, based on the real-time power consumption model of the micromodule server, the power consumption model of the air conditioning system and the rapid prediction model of the server temperature, an optimization method combining real-time combination and timing decoupling is adopted to coordinately control the deployment of virtual machines and the air conditioning system.

[0053] In the above step 10, IT system scheduling data is obtained from the scheduling data stored in the cloud platform. The IT system scheduling data includes the CPU resource requirements of the virtual machine, the mapping relationship between the virtual machine and the server, the CPU usage of each server, the real-time power consumption of each server and the chip temperature of each server. The air conditioning system data is obtained from the management system of the near-field wind wall air conditioning system. The air conditioning system data includes the fan frequency of each terminal module, the coil evaporation temperature, the outdoor unit condensing temperature, the load rate and the total power consumption of the air conditioning system. Preferably, the preset time period is not less than 24 hours.

[0054] In the above step 10, it is assumed that the servers are homogeneous, and a real-time power consumption model of the micromodule server is established by using a matrix modeling method according to IT system scheduling data.

[0055] Specifically include:

[0056] Step 101 , filtering out power consumption data of each server under different chip temperatures and CPU usage rates from IT system scheduling data.

[0057] Step 102: Establish a server static power consumption matrix based on the power consumption data of each server when the CPU usage rate is 0.

[0058] Step 103, based on the data of the change of the real-time power consumption of each server with the CPU usage rate, a multivariate linear regression method is used to fit the conversion coefficient of the CPU usage rate and the dynamic power consumption (under the standard working conditions of the server, the real-time power consumption minus the static power consumption is the dynamic power consumption) of each server to form a conversion coefficient matrix of the CPU usage rate and the dynamic power consumption.

[0059] Step 104 , according to different temperature data of chips of each server at the same CPU usage rate, a multivariate linear regression method is used to fit the expression of server leakage power consumption to form a server leakage power consumption matrix.

[0060] Step 105, combining the server static power consumption matrix, the conversion coefficient matrix of CPU usage and running power consumption, and the server leakage power consumption matrix, to obtain the real-time power consumption model of the micromodule server shown in formula (1):

[0061] P PM (t) = P SC (t)+u VM (t)*MM VM→PM (A(t),F(t))*f h +D(T PM (t)) Formula (1)

[0062] Where P PM (t) represents the server real-time power consumption matrix, matrix P PM Each element in (t) represents the real-time power consumption model of a server, that is, each element corresponds to each server in the near-field wind wall micromodule, and one element represents the power consumption model of a server; P SC (t) represents the server static power consumption matrix, u VM (t) represents the CPU resource requirement matrix of the virtual machine. VM→PM (A(t), F(t)) represents the mapping matrix between virtual machines and servers. It is a function of the virtual machine real-time deployment matrix A(t) and the virtual machine migration matrix F(t) and can be expressed as a (0,1) matrix. For example, Indicates that virtual machine 1 is deployed on server 1, and virtual machine 2 is deployed on server 2. VM (t)*MM VM→PM (A(t), F(t)) represents the CPU usage of the server, f h The conversion coefficient matrix between CPU utilization and dynamic power consumption, D(T PM(t)) represents the server leakage power consumption matrix.

[0063] In the established micromodule server real-time power consumption model, the server real-time power consumption consists of static power consumption, dynamic power consumption and leakage power consumption. Among them, static power consumption is a fixed value related to server performance, dynamic power consumption is a function of server CPU utilization, and leakage power consumption is a function of server CPU chip temperature.

[0064] In the above step 10, based on the air conditioning system data, the air conditioning system power consumption model shown in formula (2) is established:

[0065] P FC (t) = P CP (T co (t),T ev (t),PLR(t))+ΣP FX (FP(t)) Formula (2)

[0066] Where P FC (t) represents the total power consumption of the air conditioning system; P CP (T co (t),T ev (t),PLR(t)) represents the power consumption of the outdoor unit, and is the coil evaporation temperature T co (t), outdoor unit condensing temperature T ev (t) and load rate PLR(t); P FX (FP(t)) represents the power consumption matrix of each terminal module, which is a function of the fan frequency matrix FP(t) of each terminal module.

[0067] In the above step 10, a BP neural network model is established, with the fan frequency of each terminal module, the coil evaporation temperature and the CPU utilization rate of each server as input variables, and the chip temperature of each server as the output variable, to construct a dynamic mapping relationship between the fan frequency of each terminal module, the coil evaporation temperature, the utilization rate of each server CPU and the chip temperature of each server.

[0068] Specifically include:

[0069] Step 131, the mapping relationship between virtual machines and servers, the CPU resource requirements of virtual machines and the chip temperature of each server are screened out from the IT system scheduling data, and the fan frequency and coil evaporation temperature of each terminal module are screened out from the air conditioning system data to form a modeling data set. The modeling data set is divided into a training set and a test set in a ratio of 4:1.

[0070] Step 132, the fan frequency of each terminal module, the coil evaporation temperature and the CPU utilization rate of each server (calculated according to the mapping relationship between the virtual machine and the server and the CPU resource demand of the virtual machine) are used as input variables of the BP neural network model, and the chip temperature of each server is used as the output variable of the neural network model. The BP neural network model is trained using the training set, and the hyperparameters such as the number of layers of the model and the number of neurons in each layer are tuned during the training process.

[0071] The accuracy of the trained BP neural network model is tested using a test set. If the model error meets the requirements (for example, less than 1%), the model is considered qualified and a server temperature rapid prediction model is obtained.

[0072] In the above step 20, an optimization method combining real-time integration and timing decoupling is used to coordinately control the virtual machine deployment and the air conditioning system, including:

[0073] Step 21, when the cloud platform receives a new virtual machine deployment request, it jointly optimizes the virtual machine deployment location and the fan frequency of each terminal module to achieve the lowest overall power consumption of the server and each terminal module under a single virtual machine deployment.

[0074] Step 22: At each given moment, first migrate the virtual machines running on each server as a whole to minimize the total power consumption of the server. On this basis, optimize the coil evaporation temperature of the air conditioning system and the fan frequency of each terminal module to minimize the overall power consumption of the air conditioning system. Preferably, the time period between two adjacent given moments is 3 to 15 minutes.

[0075] Preferably, Figure 2 As shown, step 21 specifically includes:

[0076] Step 211, when the cloud platform receives a deployment request for a new virtual machine, it traverses the remaining CPU resources of all active servers according to the CPU resource requirements of the new virtual machine to find a server available for deployment.

[0077] In step 212, if there are servers available for deployment, the locations of the servers available for deployment are used as optimization variables for the deployment of the new virtual machine, and step 213 is executed. If the remaining CPU resources of all active servers cannot deploy the new virtual machine, a dormant server is randomly switched to an active state, the new virtual machine is deployed, and step 214 is executed.

[0078] Step 213, with the optimization goal of minimizing the sum of the real-time total power consumption of all servers after deployment and the total power consumption of all terminal modules, and with the constraints that the sum of the used resources of all virtual machines running on each server does not exceed the total rated resources of the server, and the chip temperature of each server does not exceed the preset temperature threshold, a genetic algorithm is used to jointly optimize the deployment position of the new virtual machine and the fan frequency of each terminal module to obtain the optimal real-time deployment position of the virtual machine and the optimal fan frequency of each terminal module.

[0079] As follows:

[0080] Optimization goal: minP total,PM+FX (t)=ΣP PM (t)+ΣP FX (FP(t))

[0081] Optimization constraint: T PM (t)≤T max,PM andΣu VMiin PMk(t) ≤C PM

[0082] Optimization variables: {A(t), FP(t)}

[0083] Where, ΣP PM (t) represents the total real-time power consumption of all servers, ΣP FX (FP(t)) represents the total power consumption of all terminal modules, T PM (t) represents the chip temperature of each server, T max,PM Indicates the preset temperature threshold, u VMiin PMk(t) represents the sum of resources used by all virtual machines running on each server, C PM Indicates the total amount of server rated resources.

[0084] The cloud platform implements the deployment of new virtual machines according to the optimal real-time deployment position of the virtual machines, and the air-conditioning system controls the fan frequency of each terminal module according to the optimal fan frequency of each terminal module.

[0085] Step 214, with the optimization goal of minimizing the sum of the power consumption of each terminal module after the new virtual machine is deployed, and with the constraint that the temperature of each server chip cannot exceed the preset temperature threshold, a genetic algorithm is used to optimize the fan frequency of each terminal module to obtain the optimal fan frequency of each terminal module.

[0086] As follows:

[0087] Optimization goal: minP total,FX (t) = ∑P FX (FP(t))

[0088] Optimization constraint: T PM (t)≤T max,PM

[0089] Optimization variable: {FP(t)}.

[0090] The air conditioning system controls the fan frequency of each terminal module according to the optimal fan frequency of each terminal module.

[0091] In the method of this embodiment, during real-time optimization, first find a server from the active servers that can be used to deploy a new virtual machine. If there is one, then jointly optimize the deployment location of the new virtual machine and the fan frequency of each terminal module, so that the total energy consumption of the server and each terminal module is minimized; if not, then randomly start a dormant server for deployment, and optimize the fan frequency of each terminal module, so that the total energy consumption of each terminal module is minimized. Improve the real-time matching between server heat dissipation and air conditioning terminal cooling to reduce the probability of local hot spots and overcooling in the micromodule.

[0092] Preferably, Figure 3 As shown, step 22 specifically includes:

[0093] Step 221 , at a given moment, the minimum number of active servers required is calculated based on the sum of the number of all virtual machines running on the server and their CPU resource requirements.

[0094] Step 222, based on the minimum number of active servers required, select servers from bottom to top on each rack as active servers for the next period of time in the order of rack numbers and until the number of selected servers reaches the minimum number of active servers. Considering that the heat dissipation effect at the bottom of each rack is the best, when the method of this embodiment migrates the virtual machines as a whole, servers are selected from bottom to top in a single rack for deployment.

[0095] Step 223: According to the load balancing strategy, the existing virtual machines are uniformly migrated to the selected active servers. The idle servers are hibernated.

[0096] Step 224, with the minimum total power consumption of the air-conditioning system as the optimization goal and the chip temperature of each server not exceeding the preset temperature threshold as the constraint condition, the coil evaporation temperature and the fan frequency of each terminal module are optimized using a genetic algorithm to obtain the optimal coil evaporation temperature and the optimal fan frequency of each terminal module.

[0097] As follows:

[0098] Optimization goal: minP FC (t) = P CP (T co (t),T ev (t))+∑P FX (FP(t))

[0099] Optimization constraint: T PM (t)≤T max,PM

[0100] Optimization variables: {T ev (t), FP(t)}.

[0101] Step 225 , the air conditioning system regulates the coil evaporation temperature according to the optimal coil evaporation temperature, and controls the fan frequency of each terminal module according to the optimal fan frequency of each terminal module.

[0102] In the method of this embodiment, during the timing optimization, the virtual machines are first migrated as a whole, and the idle servers after the migration are hibernated, so that the location of the active servers in the micromodule is more concentrated and the total power consumption of the servers is minimized. Then, on this basis, the coil evaporation temperature and the fan frequency of each terminal module are optimized to minimize the total energy consumption of the air conditioning system. This promotes a high degree of timing matching between server heat dissipation and air conditioning cooling, so as to further improve the utilization efficiency of cooling capacity.

[0103] The method of this embodiment performs real-time deployment optimization control on the virtual machine deployment and air conditioning system when there is a new virtual machine deployment request, so as to improve the real-time matching between the heat dissipation of the server in the near-field wind wall micromodule and the cooling of the air conditioning terminal, so as to reduce the probability of local hot spots and overcooling in the micromodule, thereby achieving the lowest overall power consumption of the server and air conditioning terminal under a single virtual machine deployment. However, if only real-time deployment optimization is performed, it is easy to cause the dispersion of active servers, that is, the virtual machine deployment that can be accommodated by a single rack server is dispersed to servers in multiple racks. Although the total power consumption of the server has not changed, due to the limited fine-grained control capabilities of the near-field wind wall air conditioning terminal, the dispersion of active servers is likely to cause waste of cooling.

[0104] By performing timed decoupling optimization control on virtual machine deployment and air conditioning system at each given moment, the virtual machines running on each server are first migrated as a whole to facilitate centralized adjustment of active servers, without causing waste of cooling capacity, and to minimize the total power consumption of the server. On this basis, the coil evaporation temperature of the air conditioning system and the fan frequency of each terminal module are optimized to achieve the lowest overall power consumption of the air conditioning system, and to facilitate the timed high matching between server heat dissipation and air conditioning cooling, thereby further improving the utilization efficiency of cooling capacity. However, if only timed decoupling optimization is performed, the frequent migration of virtual machines is likely to affect their service quality, and the frequent adjustment of the compressor operating conditions is likely to cause a significant attenuation of their lifespan. Therefore, the time interval of timed decoupling optimization should not be too short, and a long time interval obviously cannot meet the real-time deployment requirements of virtual machines.

[0105] The method of this embodiment coordinates real-time deployment optimization with timing decoupling optimization. Through the timing coordination of the two, the technical problems of local hot spots and overcooling that are prone to occur in near-field wind wall micromodules during energy-saving scheduling can be solved, so as to improve the overall energy efficiency and operation stability of the near-field wind wall micromodules.

[0106] In the method of the present invention, during the operation of the server, the existing virtual machines are automatically released (closed in real time) after completing the task. If all virtual machines on a server are released, the server is regarded as an idle server and will automatically sleep (real-time power consumption is turned to 0).

[0107] The basic principles, main features and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above specific embodiments, and the descriptions in the above specific embodiments and the specification are only for further illustrating the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed for protection.

Claims

1. A collaborative control optimization method applicable to near-field wind wall micromodules, characterized in that: The following steps are involved: Step 10, obtaining IT system scheduling data and air conditioning system data of the near-field wind wall micromodule within a preset time period, and establishing a real-time power consumption model of the micromodule server, a power consumption model of the air conditioning system, and a rapid prediction model of the server temperature; Step 20, based on the real-time power consumption model of the micromodule server, the power consumption model of the air conditioning system and the rapid prediction model of the server temperature, an optimization method combining real-time combination and timing decoupling is adopted to coordinately control the deployment of virtual machines and the air conditioning system; In step 20, an optimization method combining real-time integration and timing decoupling is used to coordinately control virtual machine deployment and the air conditioning system, including: Step 21, when the cloud platform receives a new virtual machine deployment request, the deployment location of the new virtual machine and the fan frequency of each terminal module are jointly optimized to achieve the lowest overall power consumption of the server and each terminal module under a single virtual machine deployment; Step 22: At each given time, first migrate the virtual machines running on each server as a whole to minimize the total power consumption of the server; then optimize the coil evaporation temperature of the air-conditioning system and the fan frequency of each terminal module to minimize the overall power consumption of the air-conditioning system; The step 21 specifically includes: Step 211, when the cloud platform receives the deployment request of the new virtual machine, it traverses the remaining CPU resources of all active servers according to the CPU resource requirements of the new virtual machine to find a server available for deployment; Step 212: If there are servers available for deployment, the locations of the servers available for deployment are used as optimization variables for the deployment of the new virtual machine, and step 213 is executed; if the remaining CPU resources of all active servers cannot deploy the new virtual machine, a dormant server is randomly switched to an active state, the new virtual machine is deployed, and step 214 is executed; Step 213, with the optimization goal of minimizing the sum of the total real-time power consumption of the server after deployment and the total power consumption of the terminal module, with the optimization constraints that the sum of the resources used by all virtual machines running on each server does not exceed the total rated resources of the server, and the chip temperature of each server does not exceed the preset temperature threshold, a genetic algorithm is used to jointly optimize the deployment position of the new virtual machine and the fan frequency of each terminal module to obtain the optimal real-time deployment position of the virtual machine and the optimal fan frequency of each terminal module; the cloud platform implements the deployment of the new virtual machine according to the optimal real-time deployment position of the virtual machine, and the air conditioning system controls the fan frequency of each terminal module according to the optimal fan frequency of each terminal module; Step 214, with the optimization goal of minimizing the sum of the power consumption of each terminal module after the new virtual machine is deployed, and with the constraint that the temperature of each server chip cannot exceed the preset temperature threshold, a genetic algorithm is used to optimize the fan frequency of each terminal module to obtain the optimal fan frequency of each terminal module; the air-conditioning system controls the fan frequency of each terminal module according to the optimal fan frequency of each terminal module.

2. The collaborative control optimization method according to claim 1, characterized in that: The IT system scheduling data includes the CPU resource requirements of the virtual machine, the mapping relationship between the virtual machine and the server, the CPU utilization rate of each server, the real-time power consumption of each server and the chip temperature of each server; the air-conditioning system data includes the fan frequency of each terminal module, the coil evaporation temperature, the outdoor unit condensing temperature, the load rate and the total power consumption of the air-conditioning system.

3. The collaborative control optimization method according to claim 2, characterized in that: In the step 10, a real-time power consumption model of the micromodule server is established according to the IT system scheduling data; Specifically include: Step 101, filtering out power consumption data of each server under different chip temperatures and CPU usage rates from IT system scheduling data; Step 102, establishing a server static power consumption matrix according to the power consumption data of each server when the CPU usage rate is 0; Step 103, according to the data of the change of the real-time power consumption of each server with the CPU usage rate, a conversion coefficient between the CPU usage rate and the dynamic power consumption of each server is fitted by a multivariate linear regression method to form a conversion coefficient matrix between the CPU usage rate and the dynamic power consumption; Step 104, according to different temperature data of the chip of each server at the same CPU usage rate, a multivariate linear regression method is used to fit the expression of the leakage power consumption of each server to form a server leakage power consumption matrix; Step 105, combining the server static power consumption matrix, the conversion coefficient matrix of CPU usage and dynamic power consumption, and the server leakage power consumption matrix, to obtain the real-time power consumption model of the micromodule server shown in formula (1): Formula (1) In the formula, represents the server real-time power consumption matrix, represents the server static power consumption matrix, Represents the CPU resource requirement matrix of the virtual machine, Represents the mapping matrix between virtual machines and servers, represents the real-time deployment matrix of virtual machines, represents the virtual machine migration matrix, Indicates the CPU usage of the server. Represents the conversion coefficient matrix between CPU utilization and dynamic power consumption, Represents the server leakage power consumption matrix.

4. The collaborative control optimization method according to claim 2, characterized in that: In step 10, based on the air conditioning system data, the air conditioning system power consumption model shown in formula (2) is established: Formula (2) In the formula, Indicates the total power consumption of the air conditioning system. Indicates the power consumption of the external unit. Indicates the coil evaporation temperature, Indicates the outdoor unit condensing temperature. Indicates the load factor, Represents the fan frequency matrix of each terminal module, Represents the power consumption matrix of each terminal module.

5. The collaborative control optimization method according to claim 2, characterized in that: In the step 10, a server temperature rapid prediction model is established based on IT system scheduling data and air conditioning system data; Specifically include: Step 131, filtering out the mapping relationship between virtual machines and servers, the CPU resource requirements of virtual machines, and the chip temperature of each server from the IT system scheduling data, and filtering out the fan frequency and coil evaporation temperature of each terminal module from the air conditioning system data to form a modeling data set; Step 132, based on the modeling data set, the fan frequency of each terminal module, the coil evaporation temperature and the CPU utilization rate of each server are used as input variables of the BP neural network model, and the chip temperature of each server is used as the output variable of the BP neural network model, and the BP neural network model is trained to obtain a server temperature rapid prediction model; wherein, the CPU utilization rate of each server is calculated based on the mapping relationship between the virtual machine and the server and the CPU resource demand of the virtual machine.

6. The collaborative control optimization method according to claim 1, characterized in that: The step 22 specifically includes: Step 221, at a given moment, calculating the minimum number of active servers required based on the sum of the number of all virtual machines running on the server and their CPU resource requirements; Step 222, based on the required minimum number of active servers, select servers as active servers for the next period in the order of rack numbers and server numbers on each rack, until the number of selected servers reaches the minimum number of active servers; Step 223, according to the load balancing strategy, evenly migrate the existing virtual machines to the selected active servers; hibernate the idle servers; Step 224, with the goal of minimizing the total power consumption of the air conditioning system, and with the temperature of each server chip not exceeding a preset temperature threshold as a constraint, a genetic algorithm is used to optimize the coil evaporation temperature and the fan frequency of each terminal module to obtain the optimal coil evaporation temperature and the optimal fan frequency of each terminal module; Step 225 , the air conditioning system regulates the coil evaporation temperature according to the optimal coil evaporation temperature, and controls the fan frequency of each terminal module according to the optimal fan frequency of each terminal module.

Images