Universal server energy consumption dynamic optimization method and system oriented to multi-task scene

Abstract

The invention provides a universal server energy consumption dynamic optimization method and system oriented to a multi-task scene, and relates to the technical field of universal server energy consumption dynamic optimization under multiple tasks, and the method comprises the steps: collecting historical energy consumption data, and removing abnormal historical energy consumption data to obtain effective energy consumption data; performing time sequence analysis on the effective energy consumption data to obtain reference energy consumption values of the task under different loads, and obtaining a first energy consumption distribution scheme in combination with a dynamic distribution algorithm; calculating the difference between the real-time energy consumption and the reference energy consumption, and adjusting the first energy consumption distribution scheme to obtain a second energy consumption distribution scheme; meanwhile, airflow velocity data are collected and associated with hardware heat dissipation efficiency to generate an association table, and finally the second energy consumption distribution scheme is compensated according to the association table, so that energy consumption collaborative optimization is realized, and dynamic collaborative optimization of the energy consumption of the universal server in a multi-task scene can be realized.

Description

Technical Field

[0001] This application relates to the technical field of servers, and in particular to a general-purpose server energy consumption dynamic optimization method and system for multi-tasking scenarios. Background Technology

[0002] In scenarios such as cloud computing and big data processing, general-purpose servers often need to simultaneously handle multiple types of tasks, such as document processing, data computing, and video rendering. The workload of different tasks changes dynamically over time, and their energy consumption requirements also vary significantly. To avoid excessive server energy consumption leading to hardware overload or excessively low energy consumption causing resource waste, there is an urgent need for a dynamic energy consumption optimization method that can adapt to the dynamic characteristics of multiple tasks and take into account both task requirements and hardware carrying capacity, so as to achieve efficient management of server energy consumption in multi-task scenarios.

[0003] Currently, for server energy consumption optimization, existing technologies mostly collect historical energy consumption data of the entire server and set energy consumption allocation schemes based on fixed thresholds or the load characteristics of a single task. In terms of hardware heat dissipation adaptation, a unified heat dissipation efficiency standard is usually used to constrain the overall energy consumption of the server, or the energy consumption allocation is adjusted only for the heat dissipation of a single hardware component in order to balance server energy consumption and operational stability.

[0004] However, existing technologies have the following drawbacks: energy consumption allocation cannot accurately match the needs of multiple tasks with the heat dissipation capacity of hardware, making it difficult to achieve coordinated optimization of server energy consumption in multi-task scenarios, and easily leading to problems such as insufficient energy consumption for high-priority tasks or hardware being unable to support the quota due to heat dissipation limitations. Summary of the Invention

[0005] The purpose of this application is to provide a general server energy consumption dynamic optimization method and system for multi-tasking scenarios, so as to solve the problem that energy consumption allocation in the prior art cannot accurately match the needs of multi-tasking.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a general-purpose server energy consumption dynamic optimization method for multi-tasking scenarios, comprising:

[0007] Historical energy consumption data of the server under multi-tasking scenarios is collected, and anomalies are removed from the historical energy consumption data to obtain effective energy consumption data.

[0008] The effective energy consumption data is processed by time series analysis to obtain the baseline energy consumption value of each type of task under different load intensities. A dynamic allocation algorithm is then used to obtain the first energy consumption allocation scheme for each task based on the priority of each task in the server and the corresponding baseline energy consumption value.

[0009] Calculate the difference between the real-time energy consumption value of each task in the server and the corresponding baseline energy consumption value to obtain the energy consumption deviation value of each task. Based on the energy consumption deviation value, adjust the first energy consumption allocation scheme to obtain the second energy consumption allocation scheme of each task.

[0010] Collect airflow velocity data inside the server chassis, and correlate the airflow velocity data with the heat dissipation efficiency of the corresponding server hardware to obtain a correlation table;

[0011] Based on the aforementioned association table, a compensation algorithm is used to compensate for the second energy consumption allocation scheme, so as to achieve collaborative optimization of server energy consumption in multi-task scenarios.

[0012] Optionally, the step of performing time-series analysis on the effective energy consumption data to obtain the baseline energy consumption value for each type of task under different load intensities, and employing a dynamic allocation algorithm to obtain a first energy consumption allocation scheme for each task based on the priority of each task in the server and the corresponding baseline energy consumption value, includes:

[0013] The effective energy consumption data is classified according to task type to obtain classification results. Each type of energy consumption data in the classification results is then divided according to time series to obtain multiple corresponding time periods.

[0014] Statistical processing is performed on the load intensity corresponding to the task in each time period to obtain multiple different load intervals for each type of task, and each load interval represents a load intensity range.

[0015] Calculate the average energy consumption corresponding to each load interval to obtain the baseline energy consumption value of each type of task under different load intensities, and obtain the weight value corresponding to each priority based on the priority of each task in the server.

[0016] A dynamic allocation algorithm is used to calculate the first energy consumption allocation scheme for each task by weighting all the baseline energy consumption values ​​of each type of task with the corresponding weight values.

[0017] Optionally, the dynamic allocation algorithm is used to perform a weighted calculation of all the baseline energy consumption values ​​of each type of task and the corresponding weight values ​​to obtain a first energy consumption allocation scheme for each task, including:

[0018] Establish the relationship between the load range, baseline energy consumption value, and weight value corresponding to each task, and form a mapping table;

[0019] Based on the mapping table, a dynamic allocation algorithm is used to multiply the baseline energy consumption value and weight value corresponding to each task to obtain the third energy consumption allocation scheme for each task under the corresponding load range.

[0020] The third energy consumption allocation scheme is constrained based on a preset numerical range to obtain the first energy consumption allocation scheme for each task.

[0021] Optionally, the step of using a compensation algorithm to compensate the second energy consumption allocation scheme based on the association table to achieve collaborative optimization of server energy consumption in multi-task scenarios includes:

[0022] Based on the resource distribution characteristics of the tasks during runtime, each type of task is matched with the server hardware to obtain the hardware region to which each type of task belongs.

[0023] The heat dissipation efficiency level corresponding to the hardware region to which each task belongs is obtained from the association table, and the heat dissipation efficiency level is mapped to the corresponding compensation interval according to the preset rules.

[0024] Based on the compensation range of the hardware region to which each task belongs, a compensation algorithm is used to compensate the second energy consumption allocation scheme of each task to obtain the initial compensation value of each task.

[0025] All the initial compensation values ​​within the same hardware region are superimposed to obtain the total compensation value for each hardware region, thereby achieving collaborative optimization of server energy consumption in multi-tasking scenarios.

[0026] Optionally, the compensation algorithm is used to compensate the second energy consumption allocation scheme of each task based on the compensation interval of the hardware region to which each task belongs, to obtain the initial compensation value of each task, including:

[0027] Calculate the compensation ratio corresponding to each of the aforementioned compensation intervals;

[0028] Based on the compensation ratio corresponding to each task, a compensation algorithm is used to compensate the second energy consumption allocation scheme of each task to obtain the compensation result of each task.

[0029] Based on the server hardware's capacity threshold, all compensation results are verified to remove outliers exceeding the capacity threshold, thus obtaining the initial compensation value for each task.

[0030] Optionally, adjusting the first energy consumption allocation scheme based on the energy consumption deviation value to obtain a second energy consumption allocation scheme for each task includes:

[0031] The energy consumption deviations of all tasks are summed to obtain the comprehensive deviation value;

[0032] Calculate the ratio of the energy consumption deviation value of each task to the comprehensive deviation value, and obtain the initial adjustment coefficient of each task based on the ratio;

[0033] Based on the server's preset energy consumption, the initial adjustment coefficient is normalized to obtain the target adjustment coefficient.

[0034] Based on the target adjustment coefficient, the first energy consumption allocation scheme is adjusted to obtain the second energy consumption allocation scheme for each task.

[0035] Optionally, the step of removing anomalies from the historical energy consumption data to obtain valid energy consumption data includes:

[0036] The historical energy consumption data is divided according to task type to obtain multiple sets of historical data;

[0037] The historical data in the same group are arranged in chronological order to obtain the time-series energy consumption data corresponding to each task;

[0038] The energy consumption values ​​in the time-series energy consumption data are analyzed to obtain the energy consumption fluctuation range corresponding to each task.

[0039] Each energy consumption value in the time-series energy consumption data is compared with the corresponding energy consumption fluctuation range to identify abnormal energy consumption values.

[0040] Remove the abnormal energy consumption values ​​from the time-series energy consumption data to obtain the intermediate energy consumption data corresponding to each task;

[0041] All the intermediate energy consumption data are summarized to obtain the effective energy consumption data.

[0042] Secondly, this application provides a general-purpose server power consumption dynamic optimization system for multi-tasking scenarios, including:

[0043] The data acquisition module is used to collect historical energy consumption data of the server in multi-tasking scenarios, and to remove anomalies from the historical energy consumption data to obtain effective energy consumption data.

[0044] The analysis module is used to perform time-series analysis on the effective energy consumption data to obtain the baseline energy consumption value of each type of task under different load intensities. A dynamic allocation algorithm is used to obtain the first energy consumption allocation scheme for each task based on the priority of each task in the server and the corresponding baseline energy consumption value.

[0045] The adjustment module is used to calculate the difference between the real-time energy consumption value of each task in the server and the corresponding baseline energy consumption value to obtain the energy consumption deviation value of each task. Based on the energy consumption deviation value, the first energy consumption allocation scheme is adjusted to obtain the second energy consumption allocation scheme of each task.

[0046] The association module is used to collect airflow speed data inside the server chassis, associate the airflow speed data with the heat dissipation efficiency of the corresponding server hardware, and obtain an association table.

[0047] The compensation module is used to compensate the second energy consumption allocation scheme based on the association table and using a compensation algorithm, so as to achieve collaborative optimization of server energy consumption in multi-task scenarios.

[0048] Thirdly, this application provides an electronic device, comprising:

[0049] Memory, used to store computer programs;

[0050] The processor is configured to implement the steps of the general server energy consumption dynamic optimization method for multi-tasking scenarios as described in the first aspect above when executing the computer program.

[0051] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the general server energy consumption dynamic optimization method for multi-tasking scenarios described in the first aspect above.

[0052] The general server energy consumption dynamic optimization method for multi-tasking scenarios provided in this application collects historical energy consumption data of the server under multi-tasking scenarios, removes anomalies from the historical energy consumption data to obtain effective energy consumption data; performs time-series analysis on the effective energy consumption data to obtain the baseline energy consumption value of each type of task under different load intensities; adopts a dynamic allocation algorithm to obtain a first energy consumption allocation scheme for each task based on the priority of each task in the server and the corresponding baseline energy consumption value; calculates the difference between the real-time energy consumption value of each task in the server and the corresponding baseline energy consumption value to obtain the energy consumption deviation value of each task; adjusts the first energy consumption allocation scheme based on the energy consumption deviation value to obtain a second energy consumption allocation scheme for each task; collects airflow velocity data in the server chassis, correlates the airflow velocity data with the heat dissipation efficiency of the corresponding server hardware to obtain a correlation table; and uses a compensation algorithm to compensate the second energy consumption allocation scheme based on the correlation table to achieve collaborative optimization of server energy consumption in multi-tasking scenarios.

[0053] The technical solution of this application has the following beneficial effects:

[0054] This application ensures the accuracy and reliability of subsequent energy consumption data by removing outliers from historical energy consumption data. It then determines a baseline energy consumption value based on task type and dynamic load intensity, and combines this with a differentiated energy allocation scheme based on task priority, achieving dual precision in load adaptation and priority differentiation. This avoids the problem that a single quota standard cannot meet the differentiated energy consumption needs of multiple tasks. Next, by dynamically adjusting the quota based on the deviation between real-time energy consumption and the baseline energy consumption, the energy allocation can adapt to the real-time running status of tasks, avoiding energy waste or insufficient demand caused by a fixed first energy allocation scheme. Furthermore, by establishing a quantitative correlation of heat dissipation efficiency, the application clarifies the differences in heat dissipation capabilities of different hardware areas, providing data support for the adaptation of subsequent energy allocation schemes to hardware heat dissipation capabilities, breaking the limitations of a "unified heat dissipation standard." Finally, by binding the energy allocation scheme with hardware heat dissipation capabilities, the application ensures that the quota is within the hardware's carrying capacity, while simultaneously achieving coordinated matching of multi-task energy consumption and hardware heat dissipation, thus achieving the optimization goal of efficient and stable server energy consumption in multi-task scenarios.

[0055] Furthermore, this application first classifies the effective energy consumption data according to task type, and then divides each type of data into time periods according to time series; next, it calculates the task load intensity of each time period to divide multiple load intervals; then it calculates the average energy consumption of each load interval to determine the baseline energy consumption value under different load intensities for each type of task, and obtains the weight value corresponding to the task priority; finally, it uses a dynamic allocation algorithm to calculate the first energy consumption allocation scheme for each task by weighting the baseline energy consumption value with the corresponding weight value.

[0056] This application, through detailed operations from task classification, time period division, load interval statistics, benchmark value calculation to weighted average, ensures that the benchmark energy consumption value accurately matches the different load intensities of each type of task, avoiding confusion of energy consumption data across tasks and loads. On the other hand, by combining priority weights with benchmark energy consumption values, the first energy consumption allocation scheme not only conforms to the actual energy consumption needs of tasks but also reflects the differences in task priorities, further improving the accuracy and rationality of the first energy consumption allocation scheme and laying a more reliable foundation for subsequent dynamic quota adjustments. Attached Figure Description

[0057] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0058] Figure 1A flowchart illustrating a general server energy consumption dynamic optimization method for multi-tasking scenarios provided in this application embodiment;

[0059] Figure 2 A schematic diagram illustrating a specific implementation of a general server energy consumption dynamic optimization method for multi-tasking scenarios provided in this application embodiment;

[0060] Figure 3 A schematic diagram of the structure of a general-purpose server energy consumption dynamic optimization system for multi-tasking scenarios provided in this application embodiment;

[0061] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0062] In multi-tasking scenarios such as cloud computing and big data processing, general-purpose servers need to support multiple tasks simultaneously. However, existing energy consumption optimization technologies have significant shortcomings: on the one hand, they do not combine the different priorities of multiple tasks with the dynamic load intensity to set energy consumption allocation schemes. Instead, they often use fixed standards to allocate energy consumption, which may cause high-priority tasks to be affected by insufficient quotas, or low-load tasks to be wasted due to excessive quotas. On the other hand, they do not establish a correlation between hardware-level heat dissipation efficiency and energy consumption allocation schemes. They only use a unified heat dissipation standard to constrain overall energy consumption, which is difficult to adapt to the differences in heat dissipation capabilities of different hardware areas. This can easily lead to problems such as hardware being unable to support quotas due to insufficient heat dissipation, or heat dissipation resources not being fully utilized.

[0063] To address the aforementioned issues, this application proposes a general server energy consumption dynamic optimization method for multi-task scenarios: First, historical energy consumption data is collected and purified, and a first energy consumption allocation scheme is calculated by combining task load intensity and priority; then, the quota is adjusted according to real-time energy consumption deviation, while simultaneously generating a dedicated correlation table based on hardware heat dissipation efficiency, and finally, the quota is compensated based on this table. This scheme determines the first energy consumption allocation scheme through the dual dimensions of "load and priority," dynamically adjusts real-time deviations, and accurately compensates for hardware heat dissipation. It not only solves the problem of inaccurate energy consumption allocation in multi-task scenarios but also achieves coordinated adaptation of energy consumption and hardware heat dissipation, effectively balancing server energy efficiency and operational stability.

[0064] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0065] The core of this application is to provide a general-purpose dynamic optimization method for server energy consumption in multi-tasking scenarios. A flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0066] S101. Collect historical energy consumption data of the server when facing multi-task scenarios, and remove anomalies from the historical energy consumption data to obtain effective energy consumption data.

[0067] In one specific implementation, step S101 includes:

[0068] Step 1011: Divide the historical energy consumption data according to task type to obtain multiple sets of historical data.

[0069] Step 1012: Arrange the historical data in the same group in chronological order to obtain the time-series energy consumption data corresponding to each task.

[0070] Step 1013: Analyze the energy consumption values ​​in the time-series energy consumption data to obtain the energy consumption fluctuation range corresponding to each task.

[0071] Step 1014: Compare each energy consumption value in the time-series energy consumption data with the corresponding energy consumption fluctuation range to identify abnormal energy consumption values.

[0072] Step 1015: Remove the abnormal energy consumption values ​​from the time-series energy consumption data to obtain the intermediate energy consumption data corresponding to each task.

[0073] Step 1016: Summarize all the intermediate energy consumption data to obtain effective energy consumption data.

[0074] In the above scheme, abnormal energy consumption values ​​refer to energy consumption values ​​in time-series energy consumption data that exceed the energy consumption fluctuation range of the corresponding task. These values ​​are usually caused by abnormal factors such as hardware failure, sudden interference, and abnormal task operation, and cannot reflect the energy consumption of normal task operation. They need to be removed from the time-series energy consumption data. Intermediate energy consumption data refers to the time-series data that contains only normal energy consumption values ​​after removing abnormal energy consumption values ​​from the time-series energy consumption data. This data can truly and continuously reflect the energy consumption changes when a certain type of task is running normally, and is the basis for summarizing effective energy consumption data. Effective energy consumption data refers to the multi-task normal energy consumption data set formed by summarizing the intermediate energy consumption data of all tasks. It contains the "time-energy consumption value" information of all tasks running normally, and can comprehensively reflect the normal energy consumption patterns of various tasks in multi-task scenarios. It is the core data basis for subsequently calculating the task baseline energy consumption value and allocating the first energy consumption allocation scheme.

[0075] In this application example, firstly, step 1011 is executed. The first step is to identify all task types contained in the historical energy consumption data and determine the task categories that need to be split. The second step is to use data classification technology to traverse each energy consumption record in the historical energy consumption data, and filter and group the energy consumption values ​​belonging to the same task type according to the task type marked in the record to form the historical data corresponding to that task type. The third step is to repeat the filtering and classification operation until all task types correspond to a set of historical data.

[0076] Next, step 1012 is executed. The first step is to obtain the historical data of a certain task type obtained in step 1011 and confirm the recording time information attached to each energy consumption value in the data. The second step is to use time sorting technology, with the recording time as the sorting basis, and then rearrange all the energy consumption values ​​in the historical data in chronological order. During the sorting process, it is ensured that each energy consumption value corresponds to its corresponding recording time, forming time-series energy consumption data containing the correspondence between "time and energy consumption value". This operation is repeated to generate corresponding time-series energy consumption data for the historical data of all task types.

[0077] Then, step 1013 is executed. The first step is to obtain the time-series energy consumption data of a certain task obtained in step 1012 and extract all energy consumption values ​​from the data. The second step is to process the extracted energy consumption values ​​using statistical analysis techniques. First, the average value of all energy consumption values ​​is calculated, and then the standard deviation of the energy consumption values ​​is calculated. Subsequently, according to statistical rules, such as the average value ± 2 times the standard deviation, the lower limit and upper limit of the energy consumption fluctuation range of the task are determined by combining the calculated average value and standard deviation, forming the energy consumption fluctuation range corresponding to the task. This statistical analysis operation is repeated to obtain the energy consumption fluctuation range of each task.

[0078] Next, step 1014 is executed. The first step is to obtain the time-series energy consumption data of a certain task obtained in step 1012 and the energy consumption fluctuation range of the task determined in step 1013. The second step is to use data comparison technology to traverse each energy consumption value in the time-series energy consumption data, compare each energy consumption value with the energy consumption fluctuation range, and determine whether the energy consumption value is within the range. If the energy consumption value is less than the lower limit of the range or greater than the upper limit of the range, the energy consumption value is marked as an abnormal energy consumption value. After the traversal is completed, all marked abnormal energy consumption values ​​are collected to form an abnormal energy consumption value list for the task. The comparison and marking operation is repeated to obtain the abnormal energy consumption value list for each task.

[0079] Subsequently, step 1015 is executed. The first step is to obtain the time-series energy consumption data of a certain task obtained in step 1012 and the list of abnormal energy consumption values ​​of the task obtained in step 1014. The second step is to use data filtering technology to traverse each record in the time-series energy consumption data, and locate the records in the time-series energy consumption data that belong to abnormal energy consumption values ​​according to the values ​​in the list of abnormal energy consumption values ​​and the corresponding timestamps. The located abnormal records are deleted from the time-series energy consumption data, and the remaining normal energy consumption records are retained. After the deletion is completed, the time order of the remaining normal energy consumption records is kept unchanged to form the intermediate energy consumption data corresponding to the task. This filtering and deletion operation is repeated to obtain the intermediate energy consumption data of each task.

[0080] Finally, step 1016 is executed. First, the intermediate energy consumption data of all tasks obtained in step 1015 is obtained, and the "task type, time, and energy consumption value" information contained in each intermediate energy consumption data is confirmed. Then, data aggregation technology is used to integrate the intermediate energy consumption data of different tasks at the same time point, with time as the correlation dimension, to ensure that the energy consumption information of various tasks at the same time point corresponds one-to-one. The intermediate energy consumption data of all tasks at all time points are integrated in chronological order to form a set containing the normal energy consumption information of all tasks. This set is the effective energy consumption data.

[0081] In this embodiment of the invention, historical energy consumption data is split according to task type, thus separating energy consumption data for different tasks. This avoids the mixing of energy consumption data from different tasks, which would prevent accurate analysis of the energy consumption characteristics of a single task. Furthermore, by analyzing time-series energy consumption data to determine the energy consumption fluctuation range, the problem of misjudging or missing outliers due to a lack of unified standards is avoided, ensuring that subsequent anomaly removal operations can accurately locate values ​​that deviate from the normal energy consumption pattern of the task. Subsequently, by comparing time-series energy consumption values ​​with the energy consumption fluctuation range, abnormal energy consumption values ​​can be identified, accurately locating energy consumption data that does not reflect the normal operation of the task, preventing abnormal data from being mixed into the subsequent analysis process. Finally, effective energy consumption data is obtained, avoiding the problem of scattered single-task purification data that would prevent unified analysis of the energy consumption patterns of multiple tasks.

[0082] S102. Perform time-series analysis on the effective energy consumption data to obtain the baseline energy consumption value of each type of task under different load intensities. Use a dynamic allocation algorithm to obtain the first energy consumption allocation scheme for each task based on the priority of each task in the server and the corresponding baseline energy consumption value.

[0083] In one specific implementation, step S102 includes:

[0084] Step 1021: Classify the effective energy consumption data according to task type to obtain classification results. Divide each type of energy consumption data in the classification results into time series to obtain multiple corresponding time periods.

[0085] Step 1022: Perform statistical processing on the load intensity corresponding to the task in each time period to obtain multiple different load intervals corresponding to each type of task, with each load interval representing a load intensity range.

[0086] Step 1023: Calculate the average energy consumption corresponding to each load interval to obtain the baseline energy consumption value of each type of task under different load intensities, and obtain the weight value corresponding to each priority based on the priority of each task in the server.

[0087] Step 1024: Using a dynamic allocation algorithm, the baseline energy consumption values ​​of each type of task are weighted and calculated with the corresponding weight values ​​to obtain the first energy consumption allocation scheme for each task.

[0088] Step 1024 may specifically include the following steps: establishing the correlation between the load range, baseline energy consumption value, and weight value corresponding to each task to form a mapping table; based on the mapping table, using a dynamic allocation algorithm, performing product operations on the baseline energy consumption value and weight value corresponding to each task to obtain a third energy consumption allocation scheme for each task under the corresponding load range; constraining the third energy consumption allocation scheme based on a preset value range to obtain a first energy consumption allocation scheme for each task.

[0089] In the above scheme, load intensity refers to the level of activity of a task within a certain period of time, which can be quantified by the amount of business processed by the task, reflecting the degree of server resource demand of the task during that period; load interval refers to the range formed by classifying the load intensity of the same task in multiple time periods according to the numerical value, where each load interval corresponds to a continuous range of load intensity values; baseline energy consumption value refers to the average energy consumption data of all time periods for each type of task within a certain load interval, which can reflect the average energy consumption level of the task operating normally under that load intensity; preset energy consumption quota range refers to the upper and lower limits of the energy consumption allocation scheme set according to the server hardware carrying capacity, and the first energy consumption allocation scheme refers to the basic energy consumption quota obtained after constraints that can be directly used for tasks, which is the starting point for subsequent dynamic adjustment of quotas.

[0090] In this application example, firstly, step 1021 is executed. The first step uses data classification technology to classify the effective energy consumption data obtained in S101 according to task type, extracts and integrates energy consumption records belonging to the same task type, and obtains classification results. The second step is based on the time pattern of task operation, sets a fixed time period, and uses time segmentation technology to divide each type of energy consumption data according to the time sequence. It is necessary to ensure that the divided time periods are continuous and cover the entire time range of the energy consumption data of that type. Finally, each type of energy consumption data corresponds to multiple continuous time periods.

[0091] Next, step 1022 is executed, using load statistics technology to count the business volume corresponding to the task in each time period for each type of task, thereby quantifying the load intensity of each time period; then, after the load intensity statistics of all time periods are completed, the distribution pattern of all load intensity values ​​of the same task is analyzed, and interval division technology is used to group load intensities with similar values ​​into the same range, forming multiple load intervals corresponding to the task, where each load interval corresponds to a continuous range of load intensity.

[0092] Then, execute step 1023. The first step is to extract the energy consumption data corresponding to each load interval from the energy consumption data of each type of task in each time period, and calculate the average value of all energy consumption data in each load interval. This average value is the baseline energy consumption value of the task under the corresponding load intensity. The second step is to set the priority of each task according to the server business requirements, and assign corresponding weight values ​​according to the priority, where the higher the priority, the greater the weight value.

[0093] Next, step 1024 is executed. The first step is to establish the relationship between the obtained task load ranges, baseline energy consumption values, and weight values ​​to form a mapping table, so as to clarify the baseline energy consumption values ​​and weight values ​​corresponding to different load ranges of the same task. The second step is to use a dynamic allocation algorithm based on the mapping table to multiply the baseline energy consumption value and the corresponding weight value of each task in each load range to obtain the third energy consumption allocation scheme of each task in the corresponding load range. The third step is to set a preset energy consumption limit range according to the maximum carrying capacity and minimum operating requirements of the server hardware, and constrain the third energy consumption allocation scheme. If the third energy consumption allocation scheme exceeds the range, it is adjusted to the range boundary, and finally the first energy consumption allocation scheme of each task is obtained.

[0094] In this embodiment of the invention, by obtaining effective energy consumption data, the impact of data contamination on load analysis is avoided; by dividing continuous time periods to provide time units for load intensity statistics, a foundation for accurate load intervals can be laid; then, by statistically analyzing the load intensity of each time period and dividing it into intervals, scattered data is transformed into regular ranges, which facilitates the calculation of corresponding baseline energy consumption values ​​and provides a basis for load differences in energy consumption allocation schemes; then, by calculating the average energy consumption of each interval to obtain the baseline energy consumption value, combined with task priority weights, both actual needs and importance can be taken into account; finally, by establishing a mapping table to ensure the accuracy of data association, and by constraining quotas according to hardware capabilities, overload or task anomalies are avoided, ultimately resulting in a reasonable first energy consumption allocation scheme, providing a reliable starting point for dynamic adjustments.

[0095] S103. Calculate the difference between the real-time energy consumption value of each task in the server and the corresponding baseline energy consumption value to obtain the energy consumption deviation value of each task. Based on the energy consumption deviation value, adjust the first energy consumption allocation scheme to obtain the second energy consumption allocation scheme of each task.

[0096] In one specific implementation, step S103 includes:

[0097] Step 1031: Sum the energy consumption deviation values ​​of all tasks to obtain the comprehensive deviation value.

[0098] Step 1032: Calculate the ratio of the energy consumption deviation value of each task to the comprehensive deviation value, and obtain the initial adjustment coefficient of each task based on the ratio.

[0099] Step 1033: Based on the server's preset energy consumption, normalize the initial adjustment coefficient to obtain the target adjustment coefficient.

[0100] Step 1034: Based on the target adjustment coefficient, adjust the first energy consumption allocation scheme to obtain the second energy consumption allocation scheme for each task.

[0101] The adjustment of the first energy consumption allocation scheme based on the target adjustment coefficient requires first clarifying the adjustment logic represented by the sign and magnitude of the target adjustment coefficient. For example, when the coefficient is positive, the energy consumption quota needs to be increased based on the first energy consumption allocation scheme, and when the coefficient is negative, the energy consumption quota needs to be reduced. The larger the absolute value of the coefficient, the larger the adjustment range. Then, based on the first energy consumption allocation scheme, the quota adjustment amount is calculated using the formula "first energy consumption allocation scheme × target adjustment coefficient". The first energy consumption allocation scheme and the quota adjustment amount are then added together to obtain the preliminary second energy consumption allocation scheme. Finally, the preliminary second energy consumption allocation scheme is reasonably modified in combination with the energy consumption range required for task operation. At the same time, it is verified whether the sum of the second energy consumption allocation schemes of all tasks after modification meets the server's preset energy consumption to ensure overall energy consumption compliance. Finally, the second energy consumption allocation scheme that each task can actually execute is obtained.

[0102] In the above scheme, the energy consumption deviation value refers to the difference between the current real-time energy consumption value of each task in the server and the baseline energy consumption value under the corresponding load intensity of the task. A positive number indicates that the real-time energy consumption is higher than the baseline, and a negative number indicates that the real-time energy consumption is lower than the baseline. The comprehensive deviation value is the value obtained by summing the energy consumption deviation values ​​of all tasks, and is used to measure the overall energy consumption deviation of the server when multiple tasks are running. The initial adjustment coefficient is the ratio of the energy consumption deviation value of each task to the comprehensive deviation value. This coefficient reflects the proportion of the energy consumption deviation of a single task in the total energy consumption deviation of the server. A positive number indicates that the energy consumption of the task is too high and requires more adjustment, while a negative number indicates that the energy consumption of the task is too low and can share some of the adjustment.

[0103] The preset energy consumption refers to the overall energy consumption limit of the server set in advance based on the server hardware capacity and business energy consumption requirements. The target adjustment coefficient is the coefficient obtained after normalizing the initial adjustment coefficient. The normalization process ensures that the total quota of each task after adjustment does not exceed the preset energy consumption, so that the coefficient can be directly used to adjust the first energy consumption allocation scheme, taking into account both the overall energy consumption constraints and the individual task requirements. The second energy consumption allocation scheme is the final energy consumption quota obtained after adjusting the first energy consumption allocation scheme based on the target adjustment coefficient. This quota not only corrects the deviation between the first energy consumption allocation scheme and the real-time energy consumption, but also conforms to the server's preset energy consumption constraints. It can accurately match the actual operation requirements of the tasks with the overall energy consumption limit of the server, and is the final basis for the server to allocate energy resources to each task in the future.

[0104] In this application example, firstly, step 1031 is executed, which uses real-time energy consumption monitoring technology to collect the real-time energy consumption values ​​of each task in the server, and combines them with the baseline energy consumption values ​​of each task under the corresponding load intensity obtained in step S102. The difference between the real-time energy consumption value and the baseline energy consumption value of each task is calculated by subtraction operation to obtain the energy consumption deviation value of each task. Then, the energy consumption deviation values ​​of all tasks are accumulated by summation algorithm to obtain the comprehensive deviation value.

[0105] Next, step 1032 is executed to obtain the energy consumption deviation value and the overall deviation value of each task, and to calculate the ratio of the energy consumption deviation value to the overall deviation value of each task to obtain the initial adjustment coefficient of each task. If the overall deviation value is 0, it means that the overall energy consumption of the server has no deviation, and the initial adjustment coefficient of each task is set to 0.

[0106] Then, step 1033 is executed to determine the server's preset energy consumption. This value is preset based on the server's hardware capacity and business requirements. The first step is to obtain the first energy consumption allocation scheme for each task obtained in step S102 and calculate the sum of all first energy consumption allocation schemes. The second step is to process the obtained initial adjustment coefficients using a normalization algorithm, according to the normalization formula. ,in, The adjustment factor for the target of the i-th task. Let be the initial adjustment coefficient for the i-th task. Let S be the preset energy consumption, and S be the total energy consumption of the first energy allocation scheme. The sum of the products of the first energy consumption allocation scheme for each task and the corresponding initial adjustment coefficient is used to calculate the target adjustment coefficient.

[0107] Finally, step 1034 is executed to obtain the adjustment coefficients for each task objective and the first energy consumption allocation scheme for each task obtained in step S102, and the adjustment formula is applied. Calculate the second energy consumption allocation scheme for each task, where, This is the second energy consumption allocation scheme for the i-th task. This is the first energy consumption allocation scheme for the i-th task. The target adjustment coefficient for the i-th task is used. If the calculated second energy consumption allocation scheme exceeds the reasonable range required for task operation, such as being negative or higher than the maximum energy consumption supported by the hardware, it is corrected to a reasonable range. Then, it is verified whether the sum of the second energy consumption allocation schemes for all tasks is equal to the preset energy consumption. After confirming that there is no error, the final second energy consumption allocation scheme is obtained.

[0108] In practical applications, firstly, continuing with scenarios S101 and S102, the server is currently running document processing, data computation, and video rendering tasks. The server's energy consumption monitoring module collects the real-time energy consumption values ​​for each task: the document processing task is in a low-load range; for example, the baseline energy consumption value obtained in S102 is 32W, and the real-time energy consumption value is 35W, resulting in an energy consumption deviation of 35 minus 32 equaling 3W. The data computation task is in a medium-load range; the baseline energy consumption value obtained in S102 is 60W, and the real-time energy consumption value is 57W, resulting in an energy consumption deviation of 57 minus 60 equaling -3W. The video rendering task is in a low-load range; the baseline energy consumption value obtained in S102 is 70W, and the real-time energy consumption value is 72W, resulting in an energy consumption deviation of 72 minus 70 equaling 2W. Then, a summation algorithm is used to add the energy consumption deviation values ​​of the three tasks, resulting in a comprehensive deviation value of 3 plus -3 plus 2 equaling 2W.

[0109] Next, the energy consumption deviation values ​​for each task were obtained as follows: document processing task 3W, data calculation task -3W, video rendering task 2W, and the overall deviation value 2W. The initial adjustment coefficients were calculated using division operations, resulting in the initial adjustment coefficients of 3 divided by 2 equaling 1.5 for document processing task, -3 divided by 2 equaling -1.5 for data calculation task, and 2 divided by 2 equaling 1.0 for video rendering task.

[0110] Then, the total server energy consumption T is set to 200W. The first energy allocation schemes for each task are obtained as follows: document processing task 32W, data calculation task 72W, and video rendering task 56W. The total energy allocation scheme S is calculated as 32 + 72 + 56 = 160W. Subsequently, the initial adjustment coefficients are obtained as follows: document processing task 1.5, data calculation task -1.5, and video rendering task 1.0. The sum of the products of the first energy allocation scheme and the corresponding initial adjustment coefficient for each task is calculated. ,get =-4; then, according to the normalization formula, the target adjustment coefficients are obtained: the target adjustment coefficient for document processing task is 1.5 multiplied by -10, which equals -15; the target adjustment coefficient for data calculation task is -1.5 multiplied by -10, which equals 15; and the target adjustment coefficient for video rendering task is 1.0 multiplied by -10, which equals -10.

[0111] Finally, the initial energy consumption allocation scheme Q for each task is obtained as follows: document processing task 32W, data calculation task 72W, and video rendering task 56W, with target adjustment coefficients. The values ​​are -15, 15, and -10 respectively. Then, based on the adjustment formula, the second energy consumption allocation scheme for each task is calculated: the document processing task is 32 plus 32 multiplied by -15 equals 32 minus 480 equals -448W. However, this value is unreasonable, so it is corrected to a minimum operating energy consumption of 25W. The data calculation task is 72 plus 72 multiplied by 15 equals 72 plus 1080 equals 1152W, exceeding the preset energy consumption constraint. Combined with the total quota of 200W, it is corrected to 125W. The video rendering task is 56 plus 56 multiplied by -10 equals 56 minus 560 equals -504W, corrected to a minimum operating energy consumption of 50W. After further calculation and correction, the total quota is 25 plus 125 plus 50 equals 200W, which meets the preset energy consumption.

[0112] In this embodiment of the invention, the difference between actual and baseline energy consumption is accurately captured by calculating the energy consumption deviation value of each task; the comprehensive deviation value is obtained by summing, which grasps the overall picture of energy consumption deviation of multiple tasks and avoids ignoring the overall balance; then, the deviation is converted into a quantitative adjustment ratio by calculating the initial adjustment coefficient, which clarifies the adjustment responsibility of each task; and the initial coefficient is converted into a target adjustment coefficient by normalization in combination with the preset energy consumption, taking into account both individual correction and overall constraints; then, the first energy consumption allocation scheme is adjusted using the target coefficient and the sum is verified to make the quota fit the real-time demand and ensure the compliance of the total energy consumption; the final second energy consumption allocation scheme takes into account both individual accuracy and overall stability, providing a reasonable basis for subsequent hardware heat dissipation collaborative optimization and ensuring that the energy consumption of multiple tasks is efficient and controllable.

[0113] S104. Collect airflow speed data inside the server chassis, and correlate the airflow speed data with the heat dissipation efficiency of the corresponding server hardware to obtain a correlation table.

[0114] In the above scheme, the cooling airflow path refers to the different areas through which the cooling airflow passes from entering to exiting the server chassis. It usually corresponds to the airflow channels of different hardware components and is used to distinguish the cooling environment around different hardware. The airflow speed data refers to the speed values ​​of the airflow in each cooling airflow path collected by monitoring equipment, reflecting the cooling capacity of different paths. The heat dissipation efficiency of the server hardware refers to the efficiency with which each hardware component of the server dissipates the heat it generates. It is usually positively correlated with the surrounding airflow speed; the faster the speed, the higher the heat dissipation efficiency. The correlation table is a table that records the correspondence between the airflow speed data of each cooling airflow path and the corresponding hardware heat dissipation efficiency. It is used to intuitively present the impact of airflow speed on hardware heat dissipation efficiency and provide a basis for hardware heat dissipation for subsequent energy consumption compensation.

[0115] In this application example, firstly, airflow speed monitoring technology is used to deploy monitoring devices, such as miniature wind speed sensors, at key locations inside the server chassis to ensure that each path corresponds to a monitoring point and collect airflow speed data for each path under different server operating loads.

[0116] Next, using heat dissipation efficiency testing technology, the heat dissipation efficiency of the corresponding hardware components was tested at different airflow velocities collected along each path, establishing the correlation between airflow velocity and heat dissipation efficiency.

[0117] Finally, using data association technology, each cooling airflow path, the airflow speed data of that path, the corresponding hardware component, and the heat dissipation efficiency at that speed are matched one by one to form a structured server hardware-level association table.

[0118] In practical applications, firstly, continuing the server scenario of S101-S103, we obtain the server's document processing task, data computing task, and video rendering task, and obtain the second energy consumption allocation scheme for each task: document processing task 25W, data computing task 125W, and video rendering task 50W. Then, we plan three main cooling airflow paths in the server chassis, which correspond to the CPU area path, memory area path, and hard disk area path, respectively. Each path supports the hardware operation of different tasks: the CPU area path mainly supports data computing tasks, i.e., high energy consumption; the memory area path supports document processing tasks and data computing tasks; and the hard disk area path supports video rendering tasks. Furthermore, miniature wind speed sensors are installed at the midpoint of each path to collect airflow speed data for each path under low, medium, and high server loads: Under low load, primarily for document processing tasks, the airflow speed is 1.9 m / s for the CPU area path, 1.7 m / s for the memory area path, and 1.4 m / s for the hard disk area path; Under medium load, primarily for data calculation and document processing tasks, the airflow speed is 2.4 m / s for the CPU area path, 2.1 m / s for the memory area path, and 1.8 m / s for the hard disk area path; Under high load, primarily for video rendering and data calculation tasks, the airflow speed is 2.9 m / s for the CPU area path, 2.5 m / s for the memory area path, and 2.2 m / s for the hard disk area path.

[0119] Next, a heat dissipation efficiency tester was used to test the heat dissipation efficiency of the corresponding hardware at various airflow speeds. The heat dissipation efficiency was calculated using the formula: "Heat dissipation efficiency = (Actual heat dissipation of hardware ÷ Theoretical maximum heat dissipation of hardware) × 100%". The actual heat dissipation was directly read by the tester, while the theoretical maximum heat dissipation was based on preset hardware parameters: CPU theoretical maximum heat dissipation 100W, memory 60W, and hard drive 40W. The tests showed that: at a CPU area path speed of 1.9 m / s, the actual heat dissipation was 62W, with a calculated heat dissipation efficiency of (62 ÷ 100) × 100% = 62%; at 2.4 m / s, the actual heat dissipation was 73W, with a calculated heat dissipation efficiency of 73%; and at 2.9 m / s, the actual heat dissipation was 83W, with a calculated heat dissipation efficiency of 73%. The heat dissipation efficiency is 83%. Similarly, the actual heat dissipation of the memory area path at 1.7m / s is 34.8W, and the calculated heat dissipation efficiency is (34.8 ÷ 60) × 100% = 58%; at 2.1m / s, the actual heat dissipation is 40.8W, and the calculated heat dissipation efficiency is 68%; at 2.5m / s, the actual heat dissipation is 46.8W, and the calculated heat dissipation efficiency is 78%. Similarly, the actual heat dissipation of the hard disk area path at 1.4m / s is 21.2W, and the calculated heat dissipation efficiency is (21.2 ÷ 40) × 100% = 53%; at 1.8m / s, the actual heat dissipation is 25.2W, and the calculated heat dissipation efficiency is 63%; at 2.2m / s, the actual heat dissipation is 29.2W, and the calculated heat dissipation efficiency is 73%.

[0120] Finally, the above data is organized into a table, which includes four columns: cooling airflow path, airflow velocity (m / s), corresponding hardware, and heat dissipation efficiency, forming a correlation table.

[0121] In this embodiment of the invention, by collecting airflow velocity data of different cooling airflow paths and associating them with the corresponding hardware heat dissipation efficiency, a hardware-level correlation table is formed. This correlation table clearly establishes the correspondence between airflow velocity and hardware heat dissipation capacity, avoiding subsequent energy consumption adjustments from deviating from the actual hardware heat dissipation situation. At the same time, the correlation table provides accurate hardware heat dissipation basis for subsequent energy consumption compensation, ensuring that the compensated second energy consumption allocation scheme not only meets the task operation requirements but also remains within the acceptable range of hardware heat dissipation capacity, avoiding hardware failure or energy waste due to insufficient heat dissipation. This provides key hardware heat dissipation data support for achieving collaborative optimization of server energy consumption in multi-tasking scenarios.

[0122] S105. Based on the association table, a compensation algorithm is used to compensate the second energy consumption allocation scheme to achieve collaborative optimization of server energy consumption in multi-task scenarios.

[0123] In one specific implementation, step S105 includes:

[0124] Step 1051: Based on the resource distribution characteristics occupied by the task during runtime, match each type of task with the server hardware to obtain the hardware region to which each type of task belongs.

[0125] The hardware area refers to the physical area within the server chassis that is divided according to the hardware layout and cooling airflow path.

[0126] Step 1052: Obtain the heat dissipation efficiency level corresponding to the hardware region to which each task belongs from the association table, and map the heat dissipation efficiency level to the corresponding compensation interval according to the preset rules.

[0127] Step 1053: Based on the compensation interval of the hardware region to which each task belongs, the second energy consumption allocation scheme of each task is compensated using a compensation algorithm to obtain the initial compensation value of each task.

[0128] Step 1053 may specifically include the following steps: calculating the compensation ratio corresponding to each compensation interval; based on the compensation ratio corresponding to each task, using a compensation algorithm to perform compensation processing on the second energy consumption allocation scheme of each task to obtain the compensation result of each task; and verifying all the compensation results according to the load threshold of the server hardware to remove abnormal values ​​greater than the load threshold and obtain the initial compensation value of each task.

[0129] Step 1054: Superimpose all the initial compensation values ​​in the same hardware area to obtain the total compensation value corresponding to each hardware area, so as to achieve collaborative optimization of server energy consumption in multi-task scenarios.

[0130] In the above scheme, the hardware area refers to the physical area divided according to the hardware layout and cooling airflow path inside the server chassis, where each area contains specific hardware and corresponds to an independent cooling airflow path; the hardware area to which each type of task belongs refers to the physical area where the task's main dependent hardware is located, obtained by matching the task's resource usage characteristics with the hardware situation contained in the hardware area; the compensation ratio is inversely proportional to the heat dissipation efficiency level of the compensation interval, the preset rule refers to the pre-set correspondence between the heat dissipation efficiency level and the compensation interval, the compensation interval refers to the range of compensation ratios set for different heat dissipation efficiency levels, and the compensation ratio refers to the specific compensation percentage determined from the compensation interval, which is inversely proportional to the heat dissipation efficiency level.

[0131] The compensation result refers to the compensated quota calculated using the second energy consumption allocation scheme and the compensation ratio. The carrying threshold refers to the maximum energy consumption limit that the hardware in each hardware area of ​​the server can withstand. This threshold can be obtained by pre-setting based on hardware parameters. The initial compensation value refers to the compensated quota after reasonableness verification, which can ensure that the quota is within the hardware carrying capacity range. The total compensation value refers to the value obtained by superimposing the initial compensation values ​​of all tasks in the same hardware area, which is used to reflect the total energy consumption demand of the hardware area after compensation. The collaborative optimization of server energy consumption in multi-task scenarios refers to superimposing the total compensation values ​​of each hardware area to ensure that the total energy consumption of each area is within the hardware carrying capacity and heat dissipation capacity range, while achieving the matching of multi-task energy consumption with hardware resources and heat dissipation capacity, thereby avoiding local energy overload or waste.

[0132] In this application example, firstly, step 1051 is executed, using hardware resource monitoring technology to collect the resource usage ratios of CPU, memory, and hard disk during the runtime of each type of task, determining the resource distribution characteristics of each task, and identifying the hardware that the task mainly depends on; then, based on the server hardware layout and the cooling airflow path divided in S104, the divided hardware areas and the hardware components they contain are identified, and the correspondence between hardware areas and hardware components is established; subsequently, a resource matching algorithm is used to match the resource distribution characteristics of each task with the hardware areas, and the area where the hardware that the task mainly depends on is located is determined as the hardware area to which the task belongs.

[0133] Next, step 1052 is executed. First, the association table obtained in S104 is retrieved, and the heat dissipation efficiency value of each region under the current server load is extracted according to the hardware region to which each task belongs. Then, according to the preset heat dissipation efficiency level classification standard, the extracted heat dissipation efficiency values ​​are classified into low, medium and high levels. Then, according to the preset rule that the heat dissipation efficiency level is inversely proportional to the compensation interval, the heat dissipation efficiency level of each hardware region is mapped to the corresponding compensation interval.

[0134] Then, step 1053 is executed. According to the rule that "the lower the heat dissipation efficiency level, the higher the compensation ratio", a specific compensation ratio is determined within the compensation range of each hardware area to ensure that areas with weak heat dissipation capabilities receive higher compensation. Subsequently, the second energy consumption allocation scheme of each task obtained in S103 is retrieved, and the compensation result of each task is calculated by using a compensation algorithm combined with a formula. The reasonableness of the compensation result is verified by setting the bearing threshold of each hardware area based on the hardware parameters. If the calculation result is less than the threshold, it is directly used as the initial compensation value. If it is greater than the threshold, it is adjusted to the threshold.

[0135] Finally, step 1054 is executed to organize the initial compensation values ​​of each task and their corresponding hardware regions to identify all tasks within the same hardware region. Then, a summation algorithm is used to add up the initial compensation values ​​of all tasks within the same hardware region to obtain the total compensation value for each hardware region. Subsequently, the total compensation value of each hardware region is compared and verified with the preset carrying threshold of that region. If all values ​​are less than the threshold, the collaborative optimization of server energy consumption in the multi-task scenario is completed to ensure that the energy consumption of each region is balanced and meets hardware constraints.

[0136] In practical applications, continuing with scenarios S101-S104, the server first runs three types of tasks: document processing, data computation, and video rendering. S104 has been divided into three hardware regions: CPU region, memory region, and hard disk region, which respectively contain CPU clusters, memory modules, and hard disk arrays. The resource usage of each task is collected through hardware resource monitoring tools: document processing task uses 60% memory, 20% CPU, and 20% hard disk, indicating that the resource distribution characteristics mainly depend on memory; data computation task uses 70% CPU, 25% memory, and 5% hard disk, indicating that the main dependence is CPU; video rendering task uses 50% hard disk, 30% CPU, and 20% memory, indicating that the main dependence is hard disk. Then, a resource matching algorithm is used to match tasks with hardware regions, indicating that the document processing task mainly depends on memory corresponding to the memory region, the data computation task mainly depends on CPU corresponding to the CPU region, and the video rendering task mainly depends on hard disk corresponding to the hard disk region.

[0137] Next, assuming the server is running at medium load, the association table is retrieved, primarily for data computation and document processing tasks, to obtain the heat dissipation efficiency of the hardware regions to which each task belongs: the memory region belongs to the document processing task, with a heat dissipation efficiency of 68% under load; the CPU region belongs to the data computation task, with a heat dissipation efficiency of 73% under load; and the hard disk region belongs to the video rendering task, with a heat dissipation efficiency of 63% under load. Subsequently, the preset heat dissipation efficiency level classification standard is low level 50%-65%, medium level 66%-80%, and high level 81%-95%. Based on this, the memory region 68% is determined to be medium level, the CPU region 73% is medium level, and the hard disk region 63% is low level. The preset rule is that the low level corresponds to a compensation range of 10%-15%, the medium level corresponds to 5%-10%, and the high level corresponds to 0%-5%. Therefore, the memory region is mapped to a compensation range of 5%-10%, the CPU region to 5%-10%, and the hard disk region to 10%-15%.

[0138] Then, first, obtain the compensation ranges for each hardware region as 5%-10% for the memory region, 5%-10% for the CPU region, and 10%-15% for the hard disk region. Then, determine the compensation ratio according to the inverse proportionality principle: for the memory region, the highest level of heat dissipation efficiency is 68%, so take the median value of 7%; since the CPU region's highest level of heat dissipation efficiency is 73%, which is higher than the memory region, take the lower value of 6%; for the hard disk region, the lowest level of heat dissipation efficiency is 63%, so take the median value of 12%. Next, obtain the second energy consumption allocation scheme for each task as 25W for document processing tasks, 125W for data calculation tasks, and 50W for video rendering tasks, and then apply the formula... Calculate the compensation result, where To compensate for the outcome, For the second energy consumption allocation scheme, r is the compensation ratio, and the document processing task is calculated. =25×(1+7%)=26.75W, data calculation task =125×(1+6%)=132.5W, video rendering task =50×(1+12%)=56W; then the preset load thresholds for each hardware area are 150W for the CPU area, 50W for the memory area, and 80W for the hard disk area. After verification, it was found that all calculation results were less than the corresponding thresholds. Therefore, 26.75W for the document processing task, 132.5W for the data calculation task, and 56W for the video rendering task were used as the initial compensation values ​​for each task.

[0139] Finally, the initial compensation values ​​and their corresponding hardware regions for each task were obtained: document processing task 26.75W belonged to the memory region, data calculation task 132.5W belonged to the CPU region, and video rendering task 56W belonged to the hard disk region. No other tasks belonged to the same region. Then, a summation algorithm was used to superimpose the initial compensation values ​​of the tasks in each hardware region. Specifically, the total compensation value for the memory region was 26.75W, for the CPU region it was 132.5W, and for the hard disk region it was 56W. The load thresholds for each hardware region were then preset as 150W for the CPU region, 50W for the memory region, and 80W for the hard disk region. It was verified that the memory region 26.75W was less than 50W, the CPU region 132.5W was less than 150W, and the hard disk region 56W was less than 80W, indicating that the total compensation value for all regions was within the load range.

[0140] In this embodiment of the invention, by matching task resource distribution characteristics with hardware regions, the physical region where the hardware mainly depends on each type of task is located is clearly identified, avoiding region mismatch when associating subsequent heat dissipation efficiency. At the same time, binding tasks with specific hardware regions provides a precise correlation basis for subsequent energy consumption compensation based on regional heat dissipation efficiency, ensuring that compensation is targeted at the actual heat dissipation capacity of the hardware that the task depends on, avoiding blind compensation that is detached from the actual hardware situation. By extracting heat dissipation efficiency from the association table and classifying it into levels, specific heat dissipation efficiency values ​​are transformed into easily mappable levels, simplifying the subsequent compensation interval matching process. By mapping levels to compensation intervals through preset rules, a correlation between heat dissipation efficiency and compensation amount is established, ensuring that regions with low heat dissipation efficiency can obtain higher compensation space, providing a range basis for subsequent accurate calculation of compensation ratios, and avoiding a disconnect between compensation ratios and heat dissipation efficiency.

[0141] Furthermore, by determining a compensation ratio inversely proportional to heat dissipation efficiency, areas with weak heat dissipation capabilities are ensured to receive more compensation, balancing the energy consumption needs of different areas. The compensation results are calculated using a formula, making the compensation amount quantifiable and precise. Simultaneously, based on hardware maximum capacity threshold verification, the system avoids exceeding hardware limits after compensation, preventing malfunctions. The resulting initial compensation value is both compatible with heat dissipation efficiency and conforms to hardware constraints, providing reliable basic data for subsequent regional total energy consumption aggregation. By aggregating the initial compensation values ​​of the same hardware area, the total energy consumption needs of each area are understood, preventing excessive energy consumption due to task aggregation in local areas. Combined with hardware maximum capacity threshold verification, the system ensures that the total energy consumption of each area remains within a safe range, achieving coordinated matching between multi-task energy consumption and hardware resources and heat dissipation capabilities. This satisfies the energy compensation needs of each task while ensuring the overall stability of the server operation, ultimately achieving efficient optimization of server energy consumption in multi-task scenarios.

[0142] The following is a complete example for steps S101 to S105, such as Figure 2As shown, firstly, historical energy consumption data is generated by collecting energy consumption values ​​for three types of tasks every 10 minutes over the past 72 hours on the server. This 72 hours is then converted to 4320 minutes, with a collection interval of 10 minutes. Theoretically, a single task running continuously should collect 4320 ÷ 10 = 432 data entries. For example, a document processing task running continuously for 72 hours would collect 432 data entries, a data calculation task that was not run for two periods would collect 432 - 2 = 430 data entries, and a video rendering task that was not run for four periods would collect 432 - 4 = 428 data entries. Next, the historical energy consumption data is split according to task type to obtain historical data. Then, the historical data is sorted by time to form time-series energy consumption data. Finally, standard deviation analysis is used to calculate the average energy consumption of the document processing task as 39W and the standard deviation as 4.5W, according to the formula... Calculate the energy consumption fluctuation range, where, Let s be the average energy consumption and s be the standard deviation. Substituting the data, we get 39W ± 2 × 4.5W, which is 30W-48W. Similarly, we calculate the average energy consumption of the task with a standard deviation of 5W and a range of 58W ± 2 × 5W = 50W-68W. The average energy consumption of the video rendering task is 98W with a standard deviation of 6W and a range of 98W ± 2 × 6W = 86W-110W. We then identify and remove values ​​outside the range, such as 28W and 50W for the document processing task. Finally, we summarize the remaining data from the three types of tasks to obtain the effective energy consumption data.

[0143] Next, the effective energy consumption data was first categorized by task type, then divided into 72 time periods per hour, and the load intensity of each time period was statistically analyzed: document processing tasks of 1-3 documents / time period were low load, 4-6 documents / time period were medium load, and 7-10 documents / time period were high load; data calculation tasks of 1-2 tasks / time period were low load, 3-4 tasks / time period were medium load, and 5-6 tasks / time period were high load; video rendering tasks of 1-2 tasks / time period were low load, 3-5 tasks / time period were medium load, and 6-8 tasks / time period were high load. Then, the average energy consumption of each load range was calculated to obtain the baseline energy consumption value: the total energy consumption of the 20 time periods in the low load range for document processing tasks was 640W, with an average of 640 ÷ 20 = 32W; the total energy consumption of the 30 time periods in the medium load range was 1350W, with an average of 1350 ÷ 32W. 0 = 45W; the total of 22 time periods in the high-load range is 1276W, and the average value is 1276 ÷ 22 = 58W; similarly, the data calculation task has a low load of 40W, a medium load of 60W, and a high load of 80W, and the video rendering task has a low load of 70W, a medium load of 90W, and a high load of 110W; then, the data calculation task is set as a high priority with a weight of 1.2, the document processing task as a medium priority with a weight of 1.0, and the video rendering task as a low priority with a weight of 0.8, and a dynamic allocation algorithm is used to calculate the first energy consumption allocation scheme for each type of task according to the formula "first energy consumption allocation scheme = baseline energy consumption value × weight": document processing task low load 32W × 1.0 = 32W, data calculation task low load 40W × 1.2 = 48W, video rendering task low load 70W × 0.8 = 56W.

[0144] Then, the real-time energy consumption values ​​of each task were collected: document processing task at low load 35W, data calculation task at medium load 57W, and video rendering task at low load 72W. Energy consumption deviation values ​​were calculated: document processing task 35W - 32W = 3W, data calculation task 57W - 60W = -3W, video rendering task 72W - 70W = 2W. Based on the energy consumption deviation values ​​of each task, the comprehensive deviation value was calculated to be 3W + (-3W) + 2W = 2W. Subsequently, the initial adjustment coefficients were calculated: document processing task 3 ÷ 2 = 1.5, data calculation task -3 ÷ 2 = -1.5, video rendering task 2 ÷ 2 = 1.0. The server's preset energy consumption was set to 200W. The total energy consumption of the first energy allocation scheme was then calculated to be 32W + 60W × 1.2 + 56W = 160W. Finally, the normalization formula was applied. Calculate the target adjustment factor, where, Adjust the coefficients to the target. Here, T is the initial adjustment coefficient, T is the preset energy consumption, and S is the sum of the first energy consumption allocation scheme. The sum of the products of the first energy consumption allocation scheme and the initial coefficients is used to calculate the document processing task. Data computing tasks Video rendering task and according to the formula Calculate the second energy consumption allocation scheme, where, This is the second energy consumption allocation scheme. The first energy consumption allocation scheme is as follows: document processing task 32 + 32 × (-15) = -448W, which is adjusted to a minimum of 25W; data calculation task 72 + 72 × 15 = 1152W, which is adjusted to 125W; video rendering task 56 + 56 × (-10) = -504W, which is adjusted to 50W.

[0145] Next, miniature anemometers were installed in the middle of the three cooling airflow paths within the server chassis: CPU area, memory area, and hard drive area. Airflow speed data was collected under different loads: Low load: CPU area 1.9 m / s, memory area 1.7 m / s, hard drive area 1.4 m / s; Medium load: CPU area 2.4 m / s, memory area 2.1 m / s, hard drive area 1.8 m / s; High load: CPU area 2.9 m / s, memory area 2.5 m / s, hard drive area 2.2 m / s. A thermal efficiency tester was then used to measure the efficiency using the formula: "Temperature efficiency = (Actual hardware heat dissipation ÷ Theoretical maximum hardware heat dissipation) × 100%". "In the test, the theoretical maximum heat dissipation of the hardware was 100W for the CPU, 60W for the memory, and 40W for the hard drive. The actual heat dissipation of the CPU area at 1.9m / s was 62W, with a heat dissipation efficiency of (62÷100)×100%=62%; the actual heat dissipation of the memory area at 1.7m / s was 34.8W, with a heat dissipation efficiency of (34.8÷60)×100%=58%; and the actual heat dissipation of the hard drive area at 1.4m / s was 21.2W, with a heat dissipation efficiency of (21.2÷40)×100%=53%. Similarly, the heat dissipation efficiency at other speeds was obtained, and the airflow speed data was mapped one-to-one with the corresponding hardware heat dissipation efficiency to form a correlation table."

[0146] Finally, based on the resource usage characteristics of the tasks—namely, document processing tasks using 60% memory, thus relying primarily on memory, and matching the memory region; data computing tasks using 70% CPU, thus relying primarily on CPU, and matching the CPU region; and video rendering tasks using 50% hard disk, thus relying primarily on hard disk, and matching the hard disk region—the heat dissipation efficiency of each region under medium load is extracted from the association table: memory region 68%, CPU region 73%, and hard disk region 63%. Then, according to the standard classification, the levels are: low level 50%-65%, medium level 66%-80%, and high level 81%-95%. Therefore, 68% for the memory region is considered medium level, 73% for the CPU region is considered medium level, and 63% for the hard disk region is considered low level.

[0147] Subsequently, compensation intervals are mapped according to preset rules: low level corresponds to 10%-15%, medium level corresponds to 5%-10%, therefore the memory area is mapped to 5%-10%, the CPU area to 5%-10%, and the hard disk area to 10%-15%. Then, the compensation ratio is determined: the memory area is taken as the median value of 7%, the CPU area is taken as 6% due to its higher heat dissipation efficiency, and the hard disk area is taken as the median value of 12%, and then calculated according to the formula... Calculate the initial compensation value, where, This is the initial compensation value. For the second energy consumption allocation scheme, where r is the compensation ratio, the initial compensation value for document processing tasks is 25×(1+7%)=26.75W, the initial compensation value for data calculation tasks is 125×(1+6%)=132.5W, and the initial compensation value for video rendering tasks is 50×(1+12%)=56W. These initial quotas for each hardware region are then added, resulting in 26.75W for the memory region, 132.5W for the CPU region, and 56W for the hard disk region. Preset load thresholds for each region are 150W for the CPU region, 50W for the memory region, and 80W for the hard disk region. Finally, it is verified that the total compensation value for all regions is less than the threshold, thus achieving collaborative optimization of server energy consumption in multi-tasking scenarios.

[0148] Figure 3 This is a schematic diagram illustrating a specific implementation of a general-purpose server power consumption dynamic optimization system for multi-tasking scenarios provided in this application. (Refer to...) Figure 3 The system may include:

[0149] The data acquisition module 31 is used to collect historical energy consumption data of the server when facing multi-task scenarios, and to remove anomalies from the historical energy consumption data to obtain effective energy consumption data.

[0150] Analysis module 32 is used to perform time-series analysis on the effective energy consumption data to obtain the baseline energy consumption value of each type of task under different load intensities. Using a dynamic allocation algorithm, based on the priority of each task in the server and the corresponding baseline energy consumption value, a first energy consumption allocation scheme for each task is obtained.

[0151] The adjustment module 33 is used to calculate the difference between the real-time energy consumption value of each task in the server and the corresponding baseline energy consumption value to obtain the energy consumption deviation value of each task. Based on the energy consumption deviation value, the first energy consumption allocation scheme is adjusted to obtain the second energy consumption allocation scheme of each task.

[0152] The association module 34 is used to collect airflow speed data inside the server chassis, and associate the airflow speed data with the heat dissipation efficiency of the corresponding server hardware to obtain an association table.

[0153] The compensation module 35 is used to compensate the second energy consumption allocation scheme based on the association table and using a compensation algorithm to achieve collaborative optimization of server energy consumption in multi-task scenarios.

[0154] The general server energy consumption dynamic optimization system for multi-tasking scenarios in this application embodiment is used to implement the aforementioned general server energy consumption dynamic optimization method for multi-tasking scenarios. Therefore, the specific implementation of the general server energy consumption dynamic optimization system for multi-tasking scenarios can be found in the embodiment section of the general server energy consumption dynamic optimization method for multi-tasking scenarios above. The specific implementation can be referred to the description of the corresponding embodiment, and will not be repeated here.

[0155] like Figure 4 As shown, this application also provides an electronic device, including: a memory 41 for storing a computer program; and a processor 42 for executing the computer program to implement the steps of any of the above-described general server energy consumption dynamic optimization methods for multi-tasking scenarios.

[0156] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described general server energy consumption dynamic optimization methods for multi-tasking scenarios.

[0157] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0158] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the general server energy consumption dynamic optimization method for multi-tasking scenarios.

[0159] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0160] The foregoing has provided a detailed description of a general-purpose server energy consumption dynamic optimization method and system for multi-tasking scenarios. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A general-purpose server energy consumption dynamic optimization method for multi-tasking scenarios, characterized in that, include: Historical energy consumption data of the server under multi-tasking scenarios is collected, and anomalies are removed from the historical energy consumption data to obtain effective energy consumption data. The effective energy consumption data is processed by time series analysis to obtain the baseline energy consumption value of each type of task under different load intensities. A dynamic allocation algorithm is then used to obtain the first energy consumption allocation scheme for each task based on the priority of each task in the server and the corresponding baseline energy consumption value. Calculate the difference between the real-time energy consumption value of each task in the server and the corresponding baseline energy consumption value to obtain the energy consumption deviation value of each task. Based on the energy consumption deviation value, adjust the first energy consumption allocation scheme to obtain the second energy consumption allocation scheme of each task. Collect airflow velocity data inside the server chassis, and correlate the airflow velocity data with the heat dissipation efficiency of the corresponding server hardware to obtain a correlation table; Based on the aforementioned association table, a compensation algorithm is used to compensate for the second energy consumption allocation scheme, so as to achieve collaborative optimization of server energy consumption in multi-task scenarios.

2. The method according to claim 1, characterized in that, The effective energy consumption data is processed through time-series analysis to obtain baseline energy consumption values ​​for each type of task under different load intensities. A dynamic allocation algorithm is then used to obtain a first energy consumption allocation scheme for each task based on its priority and the corresponding baseline energy consumption value. This scheme includes: The effective energy consumption data is classified according to task type to obtain classification results. Each type of energy consumption data in the classification results is then divided according to time series to obtain multiple corresponding time periods. Statistical processing is performed on the load intensity corresponding to the task in each time period to obtain multiple different load intervals for each type of task, and each load interval represents a load intensity range. Calculate the average energy consumption corresponding to each load interval to obtain the baseline energy consumption value of each type of task under different load intensities, and obtain the weight value corresponding to each priority based on the priority of each task in the server. A dynamic allocation algorithm is used to calculate the first energy consumption allocation scheme for each task by weighting all the baseline energy consumption values ​​of each type of task with the corresponding weight values.

3. The method according to claim 2, characterized in that, The dynamic allocation algorithm is used to perform a weighted calculation of all the baseline energy consumption values ​​of each type of task and the corresponding weight values ​​to obtain a first energy consumption allocation scheme for each task, including: Establish the relationship between the load range, baseline energy consumption value, and weight value corresponding to each task, and form a mapping table; Based on the mapping table, a dynamic allocation algorithm is used to multiply the baseline energy consumption value and weight value corresponding to each task to obtain the third energy consumption allocation scheme for each task under the corresponding load range. The third energy consumption allocation scheme is constrained based on a preset numerical range to obtain the first energy consumption allocation scheme for each task.

4. The method according to claim 1, characterized in that, The step of using a compensation algorithm to compensate the second energy consumption allocation scheme based on the association table to achieve collaborative optimization of server energy consumption in multi-task scenarios includes: Based on the resource distribution characteristics of the tasks during runtime, each type of task is matched with the server hardware to obtain the hardware region to which each type of task belongs. The heat dissipation efficiency level corresponding to the hardware region to which each task belongs is obtained from the association table, and the heat dissipation efficiency level is mapped to the corresponding compensation interval according to the preset rules. Based on the compensation range of the hardware region to which each task belongs, a compensation algorithm is used to compensate the second energy consumption allocation scheme of each task to obtain the initial compensation value of each task. All the initial compensation values ​​within the same hardware region are superimposed to obtain the total compensation value for each hardware region, thereby achieving collaborative optimization of server energy consumption in multi-tasking scenarios.

5. The method according to claim 4, characterized in that, The compensation interval based on the hardware region to which each task belongs is used to compensate the second energy consumption allocation scheme of each task using a compensation algorithm to obtain the initial compensation value for each task, including: Calculate the compensation ratio corresponding to each of the aforementioned compensation intervals; Based on the compensation ratio corresponding to each task, a compensation algorithm is used to compensate the second energy consumption allocation scheme of each task to obtain the compensation result of each task. Based on the server hardware's capacity threshold, all compensation results are verified to remove outliers exceeding the capacity threshold, thus obtaining the initial compensation value for each task.

6. The method according to claim 1, characterized in that, The step of adjusting the first energy consumption allocation scheme based on the energy consumption deviation value to obtain a second energy consumption allocation scheme for each task includes: The energy consumption deviations of all tasks are summed to obtain the comprehensive deviation value; Calculate the ratio of the energy consumption deviation value of each task to the comprehensive deviation value, and obtain the initial adjustment coefficient of each task based on the ratio; Based on the server's preset energy consumption, the initial adjustment coefficient is normalized to obtain the target adjustment coefficient. Based on the target adjustment coefficient, the first energy consumption allocation scheme is adjusted to obtain the second energy consumption allocation scheme for each task.

7. The method according to claim 1, characterized in that, The process of removing anomalies from the historical energy consumption data to obtain valid energy consumption data includes: The historical energy consumption data is divided according to task type to obtain multiple sets of historical data; The historical data in the same group are arranged in chronological order to obtain the time-series energy consumption data corresponding to each task; The energy consumption values ​​in the time-series energy consumption data are analyzed to obtain the energy consumption fluctuation range corresponding to each task. Each energy consumption value in the time-series energy consumption data is compared with the corresponding energy consumption fluctuation range to identify abnormal energy consumption values. Remove the abnormal energy consumption values ​​from the time-series energy consumption data to obtain the intermediate energy consumption data corresponding to each task; All the intermediate energy consumption data are summarized to obtain the effective energy consumption data.

8. A general-purpose server energy consumption dynamic optimization system for multi-tasking scenarios, characterized in that, include: The data acquisition module is used to collect historical energy consumption data of the server in multi-tasking scenarios, and to remove anomalies from the historical energy consumption data to obtain effective energy consumption data. The analysis module is used to perform time-series analysis on the effective energy consumption data to obtain the baseline energy consumption value of each type of task under different load intensities. A dynamic allocation algorithm is used to obtain the first energy consumption allocation scheme for each task based on the priority of each task in the server and the corresponding baseline energy consumption value. The adjustment module is used to calculate the difference between the real-time energy consumption value of each task in the server and the corresponding baseline energy consumption value to obtain the energy consumption deviation value of each task. Based on the energy consumption deviation value, the first energy consumption allocation scheme is adjusted to obtain the second energy consumption allocation scheme of each task. The association module is used to collect airflow speed data inside the server chassis, associate the airflow speed data with the heat dissipation efficiency of the corresponding server hardware, and obtain an association table. The compensation module is used to compensate the second energy consumption allocation scheme based on the association table and using a compensation algorithm, so as to achieve collaborative optimization of server energy consumption in multi-task scenarios.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the general server energy consumption dynamic optimization method for multi-tasking scenarios as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of a general server energy consumption dynamic optimization method for multi-tasking scenarios as described in any one of claims 1 to 7.

Images