Data center job scheduling method and system based on temperature prediction

Abstract

The present invention discloses a data center job scheduling method and system based on temperature prediction, including: obtaining relevant parameters of data center cabinets, relevant parameters of servers in the cabinets, resource sizes required for jobs to be scheduled in a job queue, and relevant parameters of cooling equipment; preprocessing the acquired data, and performing feature screening on the preprocessed data; predicting the temperature of the cabinets within a future set time period based on a trained machine learning model and the screened features, and selecting the cabinet with the lowest temperature; performing initial scheduling and optimal scheduling of the jobs to be scheduled in several servers in the cabinet with the lowest temperature, and selecting the best mapping scheme between the servers and the jobs to be scheduled through multiple iterations; and scheduling the jobs to be scheduled according to the best mapping scheme.

Description

Technical Field

[0001] The present invention relates to the technical field of data center operation scheduling, and in particular to a data center operation scheduling method and system based on temperature prediction. Background Art

[0002] The statements in this section merely mention background art related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid development of data centers, the computing performance of data centers has been greatly improved, and the computing scale has been continuously expanded. However, large-scale computing cluster systems consume more and more energy. Data center energy consumption is mainly divided into two parts: computing energy consumption and cooling energy consumption. Computing energy consumption mainly includes the energy consumption generated by server systems, network systems, etc. Cooling energy consumption mainly refers to the energy consumed by cooling facilities such as air conditioners.

[0004] There are two main reasons for high energy consumption in data centers: the first is low resource utilization. Due to different job scheduling strategies, a large number of jobs will be deployed to many servers, causing the servers in the data center to be in a low-load state, resulting in low overall resource utilization in the data center, which increases the computing energy consumption of the data center. The second is the hot spot problem. Currently, data centers mainly use mixed load scheduling. The server cannot know the type and size of the running job, causing hot spots in some cabinets, thereby reducing the air conditioning set temperature and increasing the cooling energy consumption of the data center.

[0005] In response to the above problems, existing energy-saving technologies are mainly considered from two levels: hardware and software. In terms of hardware, dynamic frequency adjustment technology (DVFS) is mainly used. According to the different computing power requirements of the programs running on the chip, the frequency and voltage are dynamically adjusted to achieve energy saving. However, if the scheduler does not have an accurate understanding of the task, the scheduler cannot guarantee that the given frequency can reduce energy consumption. In terms of software, perception and node allocation technology are mainly used. Existing technologies mainly select server nodes with low temperatures to allocate jobs, thereby avoiding the emergence of hot spots. Or choose to allocate jobs to one or more servers to improve resource utilization. However, temperature and resource utilization are not considered together, and a server with low temperature and suitable resources is selected for job scheduling to achieve the goal of minimizing energy consumption in the data center. Summary of the invention

[0006] In order to solve the deficiencies of the prior art, the present invention provides a data center job scheduling method and system based on temperature prediction; the temperature of the data center cabinets is predicted through machine learning, and the cabinet with the lowest temperature is selected. On this basis, the server and the job are mapped through the algorithm proposed by the present invention. It aims to solve the problems of one-sided optimization and low energy-saving efficiency existing in the relevant technical center, and achieve the ultimate goal of reducing the overall energy consumption of the data center.

[0007] In a first aspect, the present invention provides a data center job scheduling method based on temperature prediction;

[0008] The data center job scheduling method based on temperature prediction includes:

[0009] Obtain relevant parameters of the data center cabinet, relevant parameters of the server in the cabinet, the size of resources required for the jobs to be scheduled in the job queue, and relevant parameters of the cooling equipment;

[0010] Preprocess the acquired data and perform feature screening on the preprocessed data; based on the trained machine learning model and the screened features, predict the temperature of the cabinet within a set time period in the future and select the cabinet with the lowest temperature;

[0011] The jobs to be scheduled are initially scheduled and optimally scheduled in several servers in the cabinet with the lowest temperature. After multiple iterations, the best mapping scheme between the servers and the jobs to be scheduled is selected; the jobs to be scheduled are scheduled according to the best mapping scheme.

[0012] In a second aspect, the present invention provides a data center job scheduling system based on temperature prediction;

[0013] The data center job scheduling system based on temperature prediction includes:

[0014] A resource monitoring module is configured to: obtain relevant parameters of a data center cabinet, relevant parameters of a server in the cabinet, the size of resources required for a job to be scheduled in a job queue, and relevant parameters of a cooling device;

[0015] The temperature prediction module is configured to: pre-process the acquired data and perform feature screening on the pre-processed data; based on the trained machine learning model and the screened features, predict the temperature of the cabinet within a future set time period and select the cabinet with the lowest temperature;

[0016] The job scheduling module is configured to: initially schedule and optimize the jobs to be scheduled in several servers in the cabinet with the lowest temperature, select the best mapping scheme between the servers and the jobs to be scheduled through multiple iterations; and schedule the jobs to be scheduled according to the best mapping scheme.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] Through the present invention, relevant data of cabinets and servers in the current data center, resource requirements of jobs in the job queue, and relevant parameters of air-conditioning equipment in the data center are collected, and model training is performed on part of the collected data according to the set machine learning method to predict the temperature of the cabinets in the future. The cabinet with the lowest predicted temperature is selected for random scheduling and optimized scheduling, and finally an optimized scheduling plan is obtained. The job is deployed in the cabinet with the lowest average temperature to solve the hot spot problem in the data center and reduce the air conditioning energy consumption of the data center. At the same time, multiple jobs are allocated to one or more servers under the condition of meeting the resource utilization threshold, and the idle servers are shut down and hibernated, which solves the problem of low resource utilization in the data center and reduces the computing energy consumption of the data center, thereby achieving the goal of reducing the energy consumption of the data center as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0020] Figure 1 This is a flow chart of a data center job scheduling method based on temperature prediction provided in Embodiment 1 of the present invention;

[0021] Figure 2 This is an architecture diagram of a data center job scheduling method based on temperature prediction provided in Embodiment 1 of the present invention;

[0022] Figure 3 is a schematic diagram of a resource monitoring module provided in Embodiment 1 of the present invention;

[0023] Figure 4 This is a flowchart of a job scheduling operation provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0024] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0025] It should be noted that the terms used herein are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "include" and "have" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0026] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0027] In this embodiment, all data is obtained in compliance with laws and regulations and based on the user's consent, and is used legally.

[0028] Embodiment 1

[0029] This embodiment provides a data center job scheduling method based on temperature prediction;

[0030] like Figure 1 As shown, the data center job scheduling method based on temperature prediction includes:

[0031] S100: Obtaining relevant parameters of a data center cabinet, relevant parameters of a server in the cabinet, the size of resources required for a job to be scheduled in a job queue, and relevant parameters of a cooling device;

[0032] S200: preprocessing the acquired data, and performing feature screening on the preprocessed data; based on the trained machine learning model and the screened features, predicting the temperature of the cabinet within a future set time period, and selecting the cabinet with the lowest temperature;

[0033] S300: Initially schedule and optimize the jobs to be scheduled in several servers in the cabinet with the lowest temperature, select the best mapping scheme between the servers and the jobs to be scheduled through multiple iterations, and schedule the jobs to be scheduled according to the best mapping scheme.

[0034] Furthermore, the data center cabinet related parameters include: Cabinet Cab n Power The cabinet list is Cab = {Cab1, Cab2, ..., Cab n}.

[0035] Furthermore, the relevant parameters of the server in the cabinet include: server S i Available CPU size in and server Si The amount of memory available in The server list is S = {S1, S2, ..., S n}.

[0036] Furthermore, the resource size required for the jobs to be scheduled in the job queue includes: i Middle homework J j Required CPU size and server S i Homework J j Required memory size The job list is J = {J1, J2, ..., J n}.

[0037] Furthermore, the relevant parameters of the cooling equipment include: air inlet temperature T in , Air conditioning equipment outlet temperature T out , water inlet temperature of air conditioning water cooling system Return water temperature of air conditioning water cooling system Air intake humidity H of air conditioning equipment in and the outlet humidity H of the air conditioning equipment out .

[0038] Furthermore, the preprocessing of the acquired data includes:

[0039] Delete incomplete and empty data, and merge the cleaned data.

[0040] Exemplarily, the preprocessing of the acquired data includes:

[0041] Filter out data with a value of 0 and incomplete data, and merge each field with time as the primary key.

[0042] Furthermore, the feature screening of the preprocessed data includes:

[0043] For the preprocessed data, select the characteristic values ​​related to the cabinet temperature. The specific characteristic values ​​include: Cabinet n Power Air conditioning equipment inlet temperature T in , Air conditioning equipment outlet temperature T out , water inlet temperature of air conditioning water cooling system Return water temperature of air conditioning water cooling system Air intake humidity H of air conditioning equipment in and the outlet humidity H of the air conditioning equipment out .

[0044] Select the characteristic values ​​related to the cabinet temperature. The specific characteristic values ​​are shown in Table 1:

[0045] Table 1 Characteristic values ​​related to cabinet temperature

[0046] time Cabinet power Return air humidity Return air temperature Outlet water temperature Inlet water temperature Supply air humidity Supply air temperature 2021 / 12 / 1 0:00:00 4.36 32.83 21.02 18.25 16.6 42.96 18.23 2021 / 12 / 1 0:05:00 4.78 23.14 31.2 20.3 16.5 39.57 19.6 2021 / 12 / 1 0:10:00 4.76 22.94 31.23 20.27 16.5 39.54 19.6

[0047] Furthermore, based on the trained machine learning model and the screened features, the temperature of the cabinets within a future set time period is predicted, and the cabinet with the lowest temperature is selected, wherein the trained machine learning model; the specific training process includes:

[0048] Build machine learning prediction models;

[0049] Constructing a training set; the training set includes: numerical values ​​of several cabinets, and the numerical value of each cabinet includes: several characteristic values ​​within a historical time period of the cabinet temperature at a known current time point;

[0050] The feature values ​​in the training set include: Cabinet n Power Air conditioning equipment inlet temperature T in , Air conditioning equipment outlet temperature T out , water inlet temperature of air conditioning water cooling system Return water temperature of air conditioning water cooling system Air intake humidity H of air conditioning equipment in and the outlet humidity H of the air conditioning equipment out ;

[0051] The training set is input into the machine learning model to train the model. When the loss function value of the model no longer decreases, or the number of iterations reaches the set number, the training is stopped to obtain the trained machine learning model.

[0052] The machine learning prediction model uses a random forest model.

[0053] Model training,The machine learning model is used to train the above feature values ​​to predict the temperature of the cabinets in the data center in the future. The predicted behavior is expressed as shown in formula (1):

[0054]

[0055] Furthermore, the cabinet with the lowest temperature is selected The formula used is shown in formula (2):

[0056]

[0057] in, Indicates Cabinet n The predicted temperature; Indicates the cabinet with the lowest predicted temperature. n .

[0058] Furthermore, the jobs to be scheduled are initially scheduled and optimally scheduled in several servers in the cabinet with the lowest temperature, and the best mapping scheme between the servers and the jobs to be scheduled is selected through multiple iterations; the jobs to be scheduled are scheduled according to the best mapping scheme; the specific process includes:

[0059] Initial scheduling: Set the job J in the job queue to {J1, J2, …, J n} and the server S in the cabinet with the lowest temperature = {S1,S2,…,S n}Perform random mapping;

[0060] Optimize scheduling: Perform further optimization based on the results of the initial scheduling and calculate the remaining resource size in each server S And in ascending order, determine whether the server with the smallest remaining resources can meet the job requirements of the server with the largest remaining resources and be within the set resource utilization threshold k according to the sorting order. If it can be met, reallocate the job, otherwise continue traversing.

[0061] Furthermore, if Figure 4 As shown, the optimization scheduling specifically includes:

[0062] Step 1): Based on the results of the initial scheduling, calculate the S of each server in the current scheduling scheme i The remaining resource size in and will obtain Arrange in ascending order; execute step 2);

[0063] Step 2): Traverse the corresponding server nodes according to the remaining resource size after sorting, and select the server node that is currently ranked first and the server node that is ranked last, and execute step 3);

[0064] Step 3): Determine the remaining resource size in the server node that is currently ranked first Whether the job J in the last ranked server node is satisfied j Required CPU size and Homework J j Required memory size If satisfied, go to step 4); if not satisfied, go to step 7);

[0065] Step 4): Allocate the job in the last ranked server node to the first ranked server node, remove the last ranked server node from the queue, update the server queue position information, and execute step 5);

[0066] Step 5): Determine whether the server ranked first has remaining resources under the condition of satisfying the resource utilization threshold k; if so, continue to step 4); otherwise, execute step 6);

[0067] Step 6): Delete the server node currently ranked first from all server queues, update the sorting information of the server queues, and then return to step 2);

[0068] Step 7): Determine whether the traversal of the server node queue is completed. If not, execute step 6); if it is completed, the final mapping relationship is obtained and the scheduling is completed.

[0069] Furthermore, the step 1) is replaced by:

[0070] According to the results of the initial scheduling, determine each server S i Is the resource utilization greater than or equal to the set resource utilization threshold k? If yes, re-initialize the schedule; if no, calculate the resource utilization of each server S in the current schedule. i The remaining resource size in and will obtain Arrange them in order from smallest to largest.

[0071] Furthermore, each server S in the current scheduling scheme i The remaining resource size in The calculation process is as shown in formula (3):

[0072]

[0073] in, Indicates server S i The total amount of CPU in; Indicates server S i The total amount of memory in .

[0074] Furthermore, the step of selecting the best mapping solution between the server and the job to be scheduled specifically includes:

[0075] According to the optimization scheduling process, the mapping relationship between the optimized server and the job is obtained, and the server S under the current mapping relationship is calculated. i Computational power consumption As shown in formula (4):

[0076]

[0077] in, Indicates server S i The energy consumption of the CPU; Indicates server S i Energy consumption of memory;

[0078] The initial schedule and optimized schedule are iterated several times, and finally the The lowest mapping relationship is taken as the best mapping solution.

[0079] For example, refer to Figure 3 , resource monitoring process; specifically including:

[0080] Step S321: According to the result of random mapping, determine each server S i Whether the resource utilization is greater than or equal to the set resource utilization threshold k.

[0081] Step S322: If yes, re-initialize the scheduling.

[0082] Step S323: If not, calculate the remaining server resources according to the formula And sort in ascending order.

[0083] Step S324: traverse the above order and initialize i, j to i=0, j=n, where i, j represent the subscripts of the servers respectively.

[0084] Step S325: Determine the server S i Remaining Resources Whether it satisfies the server S j The resources required for the task.

[0085] Step S326: If yes, calculate S j The middle homework is assigned to S i Post Server S i resource utilization.

[0086] Step S3261: Continue to determine S i Whether the resource utilization in exceeds the set resource utilization threshold k.

[0087] Step S3262: If yes, set i=i+1 and return to step S324.

[0088] Step S3263: If not, then S j The jobs in S are reassigned to i middle.

[0089] Step S3264: Let j=j-1, and return to step S325.

[0090] Step S327: If not, continue to determine whether the mapping relationship between the server and the job has been traversed.

[0091] Step S328: If not, set i=i+1 and return to step S324.

[0092] Step S329: If yes, the final mapping method is obtained and the process ends.

[0093] Step S330: Final scheduling plan: According to the above two steps, a scheduling plan is finally determined, and the computing energy consumption under the mapping relationship is calculated. After multiple iterations of the above steps, select The smallest mapping relationship is used for scheduling.

[0094] The embodiment of the present invention is intended to protect a proposed data center job scheduling method based on temperature prediction, which has the following effects: by predicting the cabinet temperature based on machine learning according to the monitored data, the temperature of each cabinet in the data center within a period of time in the future can be obtained, and the cabinet with the lowest temperature is selected for job scheduling, which can avoid the emergence of hot spots in the data center, thereby avoiding excessive reduction of air conditioning temperature and reducing the cooling energy consumption of the data center. Performing the two steps of initial scheduling and optimized scheduling in the cabinet with the lowest temperature can avoid the situation where some server resources in the cabinet have low utilization, thereby improving the resource utilization of the server and reducing the computing energy consumption of the data center. Therefore, the goal of reducing the overall energy consumption of the data center is achieved as a whole.

[0095] Embodiment 2

[0096] This embodiment provides a data center job scheduling system based on temperature prediction;

[0097] like Figure 2 As shown, the data center job scheduling system based on temperature prediction includes:

[0098] A resource monitoring module is configured to: obtain relevant parameters of a data center cabinet, relevant parameters of a server in the cabinet, the size of resources required for a job to be scheduled in a job queue, and relevant parameters of a cooling device;

[0099] The temperature prediction module is configured to: pre-process the acquired data and perform feature screening on the pre-processed data; based on the trained machine learning model and the screened features, predict the temperature of the cabinet within a future set time period and select the cabinet with the lowest temperature;

[0100] The job scheduling module is configured to: initially schedule and optimize the jobs to be scheduled in several servers in the cabinet with the lowest temperature, select the best mapping scheme between the servers and the jobs to be scheduled through multiple iterations; and schedule the jobs to be scheduled according to the best mapping scheme.

[0101] It should be noted that the resource monitoring module, temperature prediction module and job scheduling module described above correspond to steps S101 to S103 in Embodiment 1, and the examples and application scenarios implemented by the modules and corresponding steps are the same, but are not limited to the contents disclosed in Embodiment 1 described above. It should be noted that the modules described above as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0102] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0103] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0104] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A data center job scheduling method based on temperature prediction, characterized in that: include: Obtain relevant parameters of the data center cabinet, relevant parameters of the server in the cabinet, the size of resources required for the jobs to be scheduled in the job queue, and relevant parameters of the cooling equipment; Preprocess the acquired data and perform feature screening on the preprocessed data; based on the trained machine learning model and the screened features, predict the temperature of the cabinet within a set time period in the future and select the cabinet with the lowest temperature; Initially schedule and optimize the jobs to be scheduled in several servers in the cabinet with the lowest temperature. After multiple iterations, select the best mapping scheme between the servers and the jobs to be scheduled. Schedule the jobs to be scheduled according to the best mapping scheme. The jobs to be scheduled are initially scheduled and optimally scheduled in several servers in the cabinets with the lowest temperature, and the best mapping solution between the servers and the jobs to be scheduled is selected through multiple iterations; Implement scheduling of jobs to be scheduled according to the best mapping scheme; The specific process includes: Initial scheduling: put jobs in the job queue Servers in the coolest cabinet Perform random mapping; Optimize scheduling: Perform further optimization based on the results of the initial scheduling and calculate the remaining resource size in each server S And in ascending order, according to the sorting order, determine whether the server with the smallest remaining resources can meet the job requirements of the server with the largest remaining resources and is within the set resource utilization threshold k. If it can be met, the job will be reallocated, otherwise continue to traverse; The optimization scheduling specifically includes: Step 1): Based on the results of the initial scheduling, calculate the The remaining resource size in , and the obtained Arrange in ascending order; proceed to step 2); Step 2): Traverse the corresponding server nodes according to the remaining resource size after sorting, and select the server node that is currently ranked first and the server node that is ranked last, and execute step 3); Step 3): Determine the remaining resource size in the server node currently ranked first Whether the job in the last ranked server node is satisfied Required CPU size and homework Required memory size ; If satisfied, proceed to step 4); if not satisfied, proceed to step 7); Step 4): Assign the job in the last ranked server node to the first ranked server node, remove the last ranked server node from the queue, update the server queue position information, and execute step 5); Step 5): Determine whether the server ranked first has remaining resources under the condition of satisfying the resource utilization threshold k; if so, continue to step 4); otherwise, execute step 6); Step 6): Delete the server node currently ranked first from all server queues, update the sorting information of the server queues, and then return to step 2); Step 7): Determine whether the traversal of the server node queue is completed. If not, execute step 6); if it is completed, the final mapping relationship is obtained and the scheduling is completed; The step of selecting the best mapping solution between the server and the job to be scheduled specifically includes: According to the optimization scheduling process, the mapping relationship between the optimized server and the job is obtained, and the server under the current mapping relationship is calculated. Computational power consumption : in, Representation Server The energy consumption of the CPU; Representation Server Energy consumption of memory; The initial schedule and optimized schedule are iterated several times, and finally the The lowest mapping relationship is taken as the best mapping solution.

2. The data center job scheduling method based on temperature prediction according to claim 1, characterized in that: The step 1) is replaced by: According to the results of the initial scheduling, determine each server Is the resource utilization greater than or equal to the set resource utilization threshold k? If yes, re-initialize the schedule; if no, calculate the resource utilization of each server in the current schedule. The remaining resource size in , and the obtained Arrange them in order from smallest to largest.

3. The data center job scheduling method based on temperature prediction according to claim 2, characterized in that: Each server in the current scheduling scheme The remaining resource size in , the calculation process is: in, Representation Server The total amount of CPU in; Representation Server The total amount of memory in and Represents weight.

4. The data center job scheduling method based on temperature prediction according to claim 1, characterized in that: The feature screening of the preprocessed data includes: For the preprocessed data, select the characteristic values ​​related to the cabinet temperature. The specific characteristic values ​​include: cabinet Power , Air conditioning equipment inlet temperature , Air outlet temperature of air conditioning equipment , water inlet temperature of air conditioning water cooling system , Return water temperature of air conditioning water cooling system , Inlet humidity of air conditioning equipment and air outlet humidity of air conditioners .

5. A data center job scheduling system based on temperature prediction, characterized in that: include: A resource monitoring module is configured to: obtain relevant parameters of a data center cabinet, relevant parameters of a server in the cabinet, the size of resources required for a job to be scheduled in a job queue, and relevant parameters of a cooling device; The temperature prediction module is configured to: pre-process the acquired data and perform feature screening on the pre-processed data; based on the trained machine learning model and the screened features, predict the temperature of the cabinet within a future set time period and select the cabinet with the lowest temperature; The job scheduling module is configured to: initially schedule and optimize the jobs to be scheduled in several servers in the cabinet with the lowest temperature, select the best mapping scheme between the servers and the jobs to be scheduled through multiple iterations; and schedule the jobs to be scheduled according to the best mapping scheme; The jobs to be scheduled are initially scheduled and optimally scheduled in several servers in the cabinets with the lowest temperature, and the best mapping solution between the servers and the jobs to be scheduled is selected through multiple iterations; Implement scheduling of jobs to be scheduled according to the best mapping scheme; The specific process includes: Initial scheduling: put jobs in the job queue Servers in the coolest cabinet Perform random mapping; Optimize scheduling: Perform further optimization based on the results of the initial scheduling and calculate the remaining resource size in each server S And in ascending order, according to the sorting order, determine whether the server with the smallest remaining resources can meet the job requirements of the server with the largest remaining resources and is within the set resource utilization threshold k. If it can be met, the job will be reallocated, otherwise continue to traverse; The optimization scheduling specifically includes: Step 1): Based on the results of the initial scheduling, calculate the The remaining resource size in , and the obtained Arrange in ascending order; proceed to step 2); Step 2): Traverse the corresponding server nodes according to the remaining resource size after sorting, and select the server node that is currently ranked first and the server node that is ranked last, and execute step 3); Step 3): Determine the remaining resource size in the server node currently ranked first Whether the job in the last ranked server node is satisfied Required CPU size and homework Required memory size ; If satisfied, proceed to step 4); if not satisfied, proceed to step 7); Step 4): Assign the job in the last ranked server node to the first ranked server node, remove the last ranked server node from the queue, update the server queue position information, and execute step 5); Step 5): Determine whether the server ranked first has remaining resources under the condition of satisfying the resource utilization threshold k; if so, continue to step 4); otherwise, execute step 6); Step 6): Delete the server node currently ranked first from all server queues, update the sorting information of the server queues, and then return to step 2); Step 7): Determine whether the traversal of the server node queue is completed. If not, execute step 6); if it is completed, the final mapping relationship is obtained and the scheduling is completed; The step of selecting the best mapping solution between the server and the job to be scheduled specifically includes: According to the optimization scheduling process, the mapping relationship between the optimized server and the job is obtained, and the server under the current mapping relationship is calculated. Computational power consumption : in, Representation Server The energy consumption of the CPU; Representation Server Energy consumption of memory; The initial schedule and optimized schedule are iterated several times, and finally the The lowest mapping relationship is taken as the best mapping solution.

Images