A data center resource management method, system and storage medium

Abstract

This invention discloses a data center resource management method, system, and storage medium, relating to the data center field. The method includes: setting a rack energy consumption upper limit for a data center rack; predicting the energy consumption of each server in the rack based on real-time utilization data of key components of each server in the rack, obtaining an estimated energy consumption for each server in the rack; and dynamically adjusting the upper limit of energy consumption control for each server in the rack, provided that the sum of the upper limit of energy consumption control for all servers in the rack equals the upper limit of rack energy consumption, based on the estimated energy consumption of each server in the rack, to limit the energy consumption of each server in the rack. This invention can reduce the energy consumption of data center servers and reduce the power consumption of the data center.

Description

Technical Field

[0001] This invention relates to the field of data centers, and in particular to a management method for data center energy resource management and space resource management. Background Technology

[0002] Data center energy management primarily relies on a power and environmental monitoring and management system, or simply an environmental system. This environmental system is used to monitor and manage various power equipment and environmental variables within critical facilities such as communication equipment rooms, data centers, and communication base stations. However, it has the following drawbacks:

[0003] 1. The environmental monitoring system only monitors the overall energy consumption data in the computer room through sensors and power monitoring equipment, and cannot monitor the real-time energy consumption data of each device in the computer room.

[0004] 2. The operation and service of the environmental system rely on the data center maintenance team, while the use of data center equipment is the responsibility of the business team. Therefore, the maintenance team will not optimize energy consumption without accurate energy consumption data, resulting in serious waste of resources.

[0005] Data center rack space management is primarily achieved through DCIM (Data Center Infrastructure Management) systems. DCIM systems are critical infrastructure within data centers, including IT infrastructure (such as servers, storage, and network switches) and infrastructure components (such as power distribution units (PDUs) and computer room air conditioning units (CRACs). While DCIM systems can collect and display information such as rack volume and energy consumption, they have the following drawbacks:

[0006] 1. The DCIM system can only collect and display data on rack space usage and IT equipment energy consumption, enabling capacity warnings, but it cannot optimize rack space utilization and reduce space redundancy by uniformly optimizing IT equipment energy consumption. Summary of the Invention

[0007] This invention provides a data center resource management method and system, which aims to intelligently optimize the configuration of data center rack power resources and even space resources.

[0008] This invention provides a data center resource management method, comprising: setting a rack energy consumption upper limit for a data center rack; predicting the energy consumption of each server in the rack based on real-time utilization data of key components of each server in the rack, to obtain an estimated energy consumption of each server in the rack; and dynamically adjusting the upper limit of energy consumption control for each server in the rack, based on the estimated energy consumption of each server in the rack, provided that the sum of the upper limit of energy consumption control for all servers in the rack equals the upper limit of energy consumption of the rack, so as to limit the energy consumption of each server in the rack.

[0009] Preferably, the method further includes: determining the number of servers that can be accommodated in the rack according to the rack energy consumption limit and the known server operating power consumption, so as to install servers on the rack according to the number of servers that can be accommodated in the rack.

[0010] Preferably, the real-time utilization data of the key components includes CPU real-time utilization, memory real-time utilization, and hardware real-time utilization. The step of predicting the energy consumption of each server in the rack based on the real-time utilization data of the key components of each server in the rack, to obtain the energy consumption estimate of each server in the rack, includes: for any server in the rack, obtaining the server's CPU real-time utilization, memory real-time utilization, and hardware real-time utilization; inputting the server's CPU real-time utilization, memory real-time utilization, and hardware real-time utilization into the server's advanced energy consumption model; and using the server's advanced energy consumption model to predict the energy consumption of each server, to obtain the energy consumption estimate of each server.

[0011] Preferably, the method further includes: using historical utilization data of key components of each server in the rack to construct an advanced energy consumption model for predicting energy consumption for each server in the rack.

[0012] Preferably, the historical utilization data of the key components includes historical CPU utilization, historical memory utilization, and historical hardware utilization. The step of using the historical utilization data of the key components of each server in the rack to construct an advanced energy consumption model for predicting energy consumption for each server in the rack includes: for any server in the rack, collecting the historical CPU utilization, historical memory utilization, and historical hardware utilization of the server; using the historical CPU utilization, historical memory utilization, and historical hardware utilization of the server, determining the values ​​of each parameter in the multiple linear regression model; and configuring the determined values ​​of each parameter to the multiple linear regression model to obtain the advanced energy consumption model for predicting energy consumption of the server.

[0013] Preferably, the step of dynamically adjusting the upper limit of energy consumption control for each server in the rack, based on the estimated energy consumption of each server in the rack, under the condition that the sum of the upper limit of energy consumption control for all servers in the rack equals the upper limit of rack energy consumption, includes: determining the proportion of the estimated energy consumption of each server in the rack within the upper limit of rack energy consumption based on the estimated energy consumption of each server in the rack; comparing the energy consumption proportion of each server in the rack calculated in this step with the energy consumption proportion calculated in the previous step to determine the change in the energy consumption proportion of each server in the rack; and dynamically adjusting the upper limit of energy consumption control for each server in the rack, under the condition that the sum of the upper limit of energy consumption control for all servers in the rack equals the upper limit of rack energy consumption, based on the change in the energy consumption proportion of each server in the rack.

[0014] Preferably, the step of dynamically adjusting the upper limit of energy consumption control for each server in the rack based on the change in the energy consumption percentage of each server in the rack, provided that the sum of the upper limit of energy consumption control for all servers in the rack equals the upper limit of energy consumption control for the rack, includes: maintaining the upper limit of energy consumption control for each server in the rack when there are no servers in the rack whose energy consumption percentage has changed; and adjusting the upper limit of energy consumption control for each server in the rack when there are both servers with increasing energy consumption percentages and servers with decreasing energy consumption percentages in the rack, provided that the sum of the upper limit of energy consumption control for all servers in the rack equals the upper limit of energy consumption control for the rack, adjusting the upper limit of energy consumption percentages. The upper limit of energy consumption control for servers with larger energy consumption is increased, while the upper limit of energy consumption control for servers with smaller energy consumption is decreased. When there are servers with larger energy consumption but no servers with smaller energy consumption in the rack, or when there are servers with smaller energy consumption but no servers with larger energy consumption, under the condition that the sum of the upper limits of energy consumption control for all servers in the rack is equal to the upper limit of energy consumption of the rack, the upper limit of energy consumption control for each server in the rack is configured according to the proportion of the estimated energy consumption of each server in the rack in the sum of the estimated energy consumption of all servers in the rack.

[0015] The present invention also provides a data center resource management system, the system including a memory, a processor, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the data center resource management method described above.

[0016] The present invention also provides a storage medium storing a program, which, when executed by a processor, implements the steps of the above-described data center resource management method.

[0017] This invention provides a way to reduce the energy consumption of data center servers and reduce the power consumption of the data center by dynamically adjusting the upper limit of energy consumption control for each server in the rack. In addition, based on the upper limit of rack energy consumption and the historical energy consumption data of the servers actually running in the rack, more servers can be installed on the rack while reasonably controlling the server energy consumption, thereby improving the rack space utilization and reducing rack space redundancy. Attached Figure Description

[0018] Figure 1 This is a flowchart of a data center resource management method provided by the present invention;

[0019] Figure 2 This is another flowchart of the data center resource management method provided by the present invention;

[0020] Figure 3 This is a flowchart illustrating the construction and use of the advanced energy consumption / power consumption model for a single server provided by this invention;

[0021] Figure 4 This is a flowchart of dynamic optimization of server energy consumption within a rack, provided by the present invention. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0023] Example 1

[0024] See Figure 1 The data center resource management method provided by this invention may include the following steps:

[0025] Step S101: Set the rack power consumption limit for the data center racks.

[0026] Based on the actual energy consumption of the servers operating within the rack, a rack energy consumption cap is set with the aim of reducing energy consumption. For example, if three servers are actually running on the rack, and each server typically consumes 350W-400W, then the rack energy consumption cap can be set to 1000W. Subsequent optimization steps will then ensure that the energy consumption of all servers within the rack does not exceed this cap, thereby reducing server energy consumption.

[0027] In addition, if the load power of the actual servers running in the rack changes significantly during the period, different rack energy consumption limits can be set according to the periodic pattern in order to better reduce energy consumption. For example, the peak period is from 8:00 to 20:00 every day, and the rack energy consumption limit is set according to the actual energy consumption of the servers running during the peak period. The off-peak period is from 20:01 to 7:59 the next day, and the rack energy consumption limit is set according to the actual energy consumption of the servers running during the off-peak period.

[0028] Step S102: Based on the real-time utilization data of the key components of each server in the rack, perform energy consumption prediction for each server in the rack to obtain the energy consumption estimate of each server in the rack.

[0029] The real-time utilization data of the key components includes CPU real-time utilization, memory real-time utilization, and hardware real-time utilization. For any server actually running in the rack, the server's CPU real-time utilization, memory real-time utilization, and hardware real-time utilization are obtained and input into the server's advanced energy consumption model. Using this model, energy consumption prediction is performed for each server, resulting in an estimated energy consumption for each server.

[0030] The advanced energy consumption model of the server can be a trained multiple linear regression model. The training data can be a set of pre-built general data or at least 14 days of continuous historical data from the server.

[0031] Step S103: Based on the estimated energy consumption of each server in the rack, dynamically adjust the upper limit of energy consumption control for each server in the rack, provided that the sum of the upper limit of energy consumption control for each server in the rack is equal to the upper limit of energy consumption of the rack, so as to limit the energy consumption of each server in the rack.

[0032] Specifically, based on the estimated energy consumption of each server in the rack, the energy consumption ratio of the estimated energy consumption of each server in the rack within the rack's upper limit energy consumption is determined. The calculated energy consumption ratio of each server in the rack is then compared with the previously calculated energy consumption ratio to determine the change in the energy consumption ratio of each server in the rack. In order to dynamically adjust the upper limit energy consumption control value of each server in the rack based on the change in the energy consumption ratio of each server in the rack, provided that the sum of the upper limit energy consumption control values ​​of all servers in the rack equals the upper limit energy consumption control value of the rack.

[0033] This invention reduces the power consumption of data centers by dynamically adjusting the upper limit of energy consumption control for each server in a data center rack, thereby reducing the energy consumption of each server in the rack.

[0034] Example 2

[0035] See Figure 2 The data center resource management method provided by this invention may include the following steps:

[0036] Step S201: Set the rack power consumption limit for the data center racks.

[0037] For example, the upper limit of rack power consumption can be set to a specified value, such as 1000W.

[0038] Step S202: Determine the number of servers that can be accommodated in the rack according to the rack energy consumption limit and the known server operating power consumption, and install servers on the rack according to the number of servers that can be accommodated in the rack.

[0039] The server's operating power consumption can be based on empirical data. For example, the maximum power consumption of a server is 600W, but based on experience, the actual load power consumption of a server rarely reaches 600W, generally being 300W-400W. Therefore, the server's operating power consumption can be any value between 300W and 400W, such as 350W. Currently, when installing servers in a rack, the number of servers that can be installed in the rack is calculated according to the maximum power consumption value given by the server at the factory. Assuming that the maximum power consumption of a rack is 1000W and the maximum power consumption of a server is 600W, then this rack can only install one server, wasting a lot of rack space and reducing rack space utilization. However, this invention, based on the rack's maximum power consumption and the server's operating power consumption, combined with a dynamic power consumption control algorithm, can determine that this rack can accommodate 3 servers, thereby greatly improving rack space utilization.

[0040] The server operating power consumption can also be based on the actual power consumption of the servers actually running in the rack. For example, if the power consumption of one server is 300W-400W, then the server operating power consumption can be any value between 300W-400W, such as 350W. After setting the rack energy consumption upper limit, this invention, based on the rack energy consumption upper limit and the server operating power consumption, and in conjunction with a dynamic energy consumption control algorithm, determines that the rack can accommodate 3 servers. At this point, by adding 2 servers, the actual number of servers running in the rack reaches the accommodating number.

[0041] To ensure that the total energy consumption of all servers in the rack does not exceed the rack's energy consumption limit and to minimize the impact on the service operation of each server in the rack, this invention employs a dynamic energy consumption control algorithm that includes steps S203 and S204.

[0042] Step S203: Based on the real-time utilization data of the key components of each server in the rack, perform energy consumption prediction for each server in the rack to obtain an estimated energy consumption for each server in the rack.

[0043] The real-time utilization data of the key components includes CPU real-time utilization, memory real-time utilization, and hardware real-time utilization. For any server actually running in the rack, the server's CPU real-time utilization, memory real-time utilization, and hardware real-time utilization are obtained and input into the server's advanced energy consumption model. Using this model, energy consumption prediction is performed for each server, resulting in an estimated energy consumption for each server.

[0044] The advanced energy consumption model of the server can be a trained multiple linear regression model. The training data can be a pre-built set of general data or at least 14 days of continuous historical data from the server.

[0045] Step S204: Based on the estimated energy consumption of each server in the rack, and under the condition that the sum of the upper limit values ​​of the energy consumption control of each server in the rack is equal to the upper limit value of the rack energy consumption, dynamically adjust the upper limit value of the energy consumption control of each server in the rack to limit the energy consumption of each server in the rack.

[0046] Specifically, based on the estimated energy consumption of each server in the rack, the energy consumption ratio of the estimated energy consumption of each server in the rack within the rack's upper limit energy consumption is determined. The calculated energy consumption ratio of each server in the rack is then compared with the previously calculated energy consumption ratio to determine the change in the energy consumption ratio of each server in the rack. In order to dynamically adjust the upper limit energy consumption control value of each server in the rack based on the change in the energy consumption ratio of each server in the rack, provided that the sum of the upper limit energy consumption control values ​​of all servers in the rack equals the upper limit energy consumption control value of the rack.

[0047] This invention, under the premise of a specified upper limit for rack energy consumption, can increase the server density within a rack based on the actual energy consumption of the servers during operation, optimize rack space, and improve rack space utilization. At the same time, by dynamically adjusting the upper limit of energy consumption control for each server within the data center rack, the energy consumption of each server within the rack is reduced, thereby reducing the power consumption of the data center. This achieves intelligent optimization of power and space configuration, realizes efficient resource management and utilization, and solves the problem that existing technologies cannot achieve dynamic optimization management of data center energy consumption and rack utilization.

[0048] In Embodiments 1 and 2 above, historical utilization data and associated historical power consumption data of key components of each server within the rack can be used to construct an advanced energy consumption model (or advanced power consumption model) for predicting energy consumption for each server. The historical utilization data of the key components includes historical CPU utilization, historical memory utilization, and historical hardware utilization. (See [link to documentation]). Figure 3 The process of building and using the advanced power consumption model for a single server is as follows:

[0049] Step S301: Data acquisition.

[0050] For any server (target server) in the rack, key data of the target server for at least 14 consecutive days is collected through the data center management system. The collected data includes historical CPU utilization, historical memory utilization, historical hardware utilization, and associated historical power consumption data.

[0051] Step S302: Machine learning trains the high-level power consumption model parameter values ​​for the server.

[0052] Using historical power consumption data as the training target, the values ​​of each parameter in the multiple linear regression model are determined by utilizing the server's historical CPU utilization, historical memory utilization, and historical hardware utilization. Specifically, the data is input into a machine learning tool (multiple linear regression model), and the parameter values ​​of the server's advanced power consumption model can be obtained through data training.

[0053] The multiple linear regression model is as follows:

[0054] y=beta0+beta1*x1+beta2*x2+betan*xn+epsilon

[0055] Where: y is the power consumption estimate, x1, x2, xn are the historical CPU utilization, historical memory utilization, and historical hardware utilization, respectively, beta0 is the constant term, beta1, beta2, betaan are the coefficients of the independent variables corresponding to x1, x2, xn, respectively, and epsilon is the error term.

[0056] Before collecting historical data, a pre-built set of general data is used to construct the model. After collecting historical data, the values ​​of each parameter are calculated based on the actual historical data (CPU, memory, and disk utilization). As the amount of historical data increases, the additional historical data can be applied to optimize the model parameters.

[0057] Step S303: Generate an advanced power consumption model for the server.

[0058] The determined values ​​of each parameter are assigned to the multiple linear regression model to obtain an advanced power consumption model for predicting the server's energy consumption.

[0059] Step S304: Input the current collected data of the target server into the model to obtain the power consumption estimate of the target server at the next moment.

[0060] In other words, this invention constructs an advanced energy consumption model algorithm for data center servers. This algorithm can construct the server's energy consumption model using a multiple linear regression model by collecting historical utilization data of key server components (such as CPU, memory, and hard disk), and then use the energy consumption model to estimate the server's energy consumption at the next moment.

[0061] In Embodiments 1 and 2 above, the data center management system can set a single-server energy consumption control limit for each server within a rack based on the rack energy consumption limit. Specifically, the algorithm dynamically modifies the energy consumption control limit for each server within the rack based on the estimated energy consumption of each server at each moment and the rack energy consumption limit at that time, thereby ensuring that the sum of the single-server energy consumption limits within the rack equals the rack energy consumption limit. See also Figure 4 The dynamic optimization process for the maximum energy consumption of each server within the rack is as follows:

[0062] Step S401: Set the rack power consumption limit.

[0063] Step S402: Generate an estimated energy consumption for each server within the rack.

[0064] The energy consumption estimate for each server within the computer rack is based on the advanced power consumption model for each server.

[0065] Step S403: The estimated energy consumption of a single server within the computer rack represents a percentage of the rack's maximum energy consumption.

[0066] Step S404: Compare the energy consumption percentage data of each server in the rack calculated this time with the energy consumption percentage data calculated in the previous time to determine whether the percentage has changed. If it has changed, proceed to step S405 to adjust the upper limit of server energy consumption control according to the estimated value. After completion, proceed to step S402 to continue calculating the energy consumption estimate for the next moment. If it has not changed, do not adjust the upper limit of server energy consumption control, and proceed to step S402 to calculate the energy consumption estimate for the next moment.

[0067] When there are no servers in the rack whose energy consumption ratio changes, the upper limit of energy consumption control for each server in the rack is maintained, that is, the upper limit of server energy consumption is not adjusted.

[0068] When servers within the rack exhibiting varying energy consumption percentages, if both servers with increasing and decreasing energy consumption percentages exist simultaneously, then, provided the sum of the energy consumption control limits for all servers in the rack equals the rack's overall energy consumption limit, the energy consumption control limit for the server with the increasing energy consumption percentage will be increased, and the energy consumption control limit for the server with the decreasing energy consumption percentage will be decreased. This is because if the estimated or decreasing energy consumption percentage indicates a reduction in the server's workload, further reducing the server's energy consumption limit allows for greater energy savings. Conversely, if the estimated or decreasing energy consumption percentage indicates an increase in the server's workload, increasing the server's energy consumption limit allows for more energy resources without impacting server performance.

[0069] When there are servers in the rack whose energy consumption percentage changes, if there are servers with a higher energy consumption percentage but no servers with a lower energy consumption percentage, or vice versa, then, provided that the sum of the energy consumption control limits of all servers in the rack equals the rack's energy consumption limit, an energy consumption control limit is configured for each server in the rack according to its proportion within the sum of the estimated energy consumption of all servers in the rack. Taking a rack energy consumption limit of 1000W as an example, 1000W is allocated to each server in the rack according to its proportion. This ensures that the sum of the energy consumption control limits of all servers in the rack equals the rack's energy consumption limit, and also ensures that the energy consumption distribution among the servers in the rack is relatively reasonable and does not affect business operations.

[0070] In this invention, the upper limit of the server's energy consumption control is a control value. By using the power consumption control interface in the server's out-of-band management function to send an energy consumption control command carrying the control value to the server, the server's energy consumption can be kept at this control value, thereby limiting the server's power consumption.

[0071] Example 3

[0072] The present invention can also provide a data center resource management system, the system including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, it implements the steps of the data center resource management method as described in Embodiment 1 or Embodiment 2.

[0073] For example, given a fixed number of servers in a rack, a rack energy consumption cap can be set based on the actual energy consumption of each server. This cap can reduce server energy consumption. For instance, if a rack has three servers, and their actual operating energy consumption is typically 300-400W, the rack energy consumption cap can be set to 1000W. By estimating the energy consumption of each of the three servers, and ensuring the sum of their energy consumption caps equals the rack energy consumption cap of 1000W, the individual energy consumption caps of each server can be dynamically adjusted to control server energy costs.

[0074] For example, given a known maximum rack energy consumption value A, existing technologies calculate that a rack can accommodate a maximum of 20 servers based on the server's factory maximum power consumption. However, once these 20 servers are running, the actual load on each server is always less than 100% (possibly only 50%), so the sum of the actual energy consumption of all 20 servers is always less than the maximum rack energy consumption value A. This results in high energy costs and significant waste of rack space. The system of this invention can install and run more than 20 servers on a rack, say 25. During actual operation, the system dynamically limits the maximum energy consumption of each server based on its estimated energy consumption, ensuring that the sum of the maximum energy consumption values ​​for all servers never exceeds A. This efficiently utilizes rack space and reduces server energy costs.

[0075] Example 4

[0076] The present invention also provides a storage medium storing a program, which, when executed by a processor, implements the steps of the data center resource management method as described in Embodiment 1 or Embodiment 2.

[0077] The storage medium can be a magnetic medium, an optical medium, a flash memory medium, or other media with storage functions.

[0078] This invention can directly reduce the energy consumption of data center servers and maximize the utilization of rack space, bringing direct economic benefits to data center server providers.

[0079] The preferred embodiments of the present invention have been described above with reference to the accompanying drawings, but this does not limit the scope of the invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and spirit of the present invention should be within the scope of the present invention.

Claims

1. A data center resource management method, characterized in that, The method includes: Set upper limits for rack power consumption for data center racks; Based on the real-time utilization data of the key components of each server in the rack, the energy consumption of each server in the rack is predicted to obtain the energy consumption estimate of each server in the rack. Based on the estimated energy consumption of each server in the rack, the energy consumption control upper limit of each server in the rack is dynamically adjusted under the condition that the sum of the upper limit of energy consumption control of each server in the rack is equal to the upper limit of energy consumption of the rack, so as to limit the energy consumption of each server in the rack.

2. The method according to claim 1, characterized in that, The method further includes: Based on the rack energy consumption limit and the known server operating power consumption, determine the number of servers that can be accommodated in the rack, and install servers on the rack according to the number of servers that can be accommodated in the rack.

3. The method according to claim 1, characterized in that, The real-time utilization data of the key components includes CPU real-time utilization, memory real-time utilization, and hardware real-time utilization. The step of predicting the energy consumption of each server in the rack based on the real-time utilization data of the key components of each server in the rack, and obtaining the estimated energy consumption of each server in the rack, includes: For any server in the rack, obtain the server's real-time CPU utilization, real-time memory utilization, and real-time hardware utilization. The server's real-time CPU utilization, real-time memory utilization, and real-time hardware utilization are input into the server's advanced energy consumption model. Using the advanced energy consumption model of the server, energy consumption is predicted for each server to obtain an estimated energy consumption for each server.

4. The method according to claim 3, characterized in that, The method further includes: Using historical utilization data of key components of each server in the rack, an advanced energy consumption model is built for each server in the rack to predict energy consumption.

5. The method according to claim 4, characterized in that, The historical utilization data of the key components includes historical CPU utilization, historical memory utilization, and historical hardware utilization. The step of using historical utilization data of key components of each server in the rack to build an advanced energy consumption model for predicting energy consumption for each server in the rack includes: For any server in the rack, collect the server's historical CPU utilization, historical memory utilization, and historical hardware utilization. Using the historical CPU utilization, historical memory utilization, and historical hardware utilization of the server, the values ​​of each parameter in the multiple linear regression model are determined. The values ​​of the determined parameters are assigned to the multiple linear regression model to obtain an advanced energy consumption model for predicting the server's energy consumption.

6. The method according to any one of claims 1-5, characterized in that, The step of dynamically adjusting the upper limit of energy consumption control for each server in the rack, based on the estimated energy consumption of each server in the rack and provided that the sum of the upper limit of energy consumption control for all servers in the rack equals the upper limit of energy consumption control for the rack, includes: Based on the energy consumption estimate of each server in the rack, determine the proportion of the energy consumption estimate of each server in the rack to the rack's energy consumption ceiling. The energy consumption percentage of each server in the rack is calculated in this instance and compared with the energy consumption percentage calculated in the previous instance to determine the change in the energy consumption percentage of each server in the rack. Based on the changes in the energy consumption ratio of each server in the rack, the energy consumption control upper limit of each server in the rack is dynamically adjusted under the condition that the sum of the energy consumption control upper limit of each server in the rack is equal to the rack energy consumption upper limit.

7. The method according to claim 6, characterized in that, The step of dynamically adjusting the upper limit of energy consumption control for each server in the rack, based on the change in the energy consumption proportion of each server in the rack, under the condition that the sum of the upper limit of energy consumption control for all servers in the rack equals the upper limit of energy consumption control for the rack, includes: When there are no servers in the rack whose energy consumption percentage changes, the energy consumption control upper limit of each server in the rack is maintained. When both servers with a higher energy consumption ratio and servers with a lower energy consumption ratio exist in the rack, under the condition that the sum of the upper limit of energy consumption control of each server in the rack is equal to the upper limit of energy consumption of the rack, the upper limit of energy consumption control of the server with a higher energy consumption ratio is increased, and the upper limit of energy consumption control of the server with a lower energy consumption ratio is decreased. When there are servers in the rack with a higher energy consumption percentage but no servers with a lower energy consumption percentage, or when there are servers in the rack with a lower energy consumption percentage but no servers with a higher energy consumption percentage, under the condition that the sum of the energy consumption control upper limits of all servers in the rack is equal to the rack's energy consumption upper limit, an energy consumption control upper limit is configured for each server in the rack according to the proportion of the energy consumption of each server in the rack in the sum of the estimated energy consumption of all servers in the rack.

8. A data center resource management system, characterized in that, The system includes a memory, a processor, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the data center resource management method as described in any one of claims 1-7.

9. A storage medium, characterized in that, The storage medium stores a program, which, when executed by a processor, implements the steps of the data center resource management method as described in any one of claims 1-7.

Images