Server computing power idle degree prediction method and device, equipment and storage medium

Abstract

This application discloses a method, apparatus, device, and storage medium for predicting server computing power idleness, relating to the field of server monitoring technology. The method for predicting server computing power idleness in this application first acquires server status indicator data for a target historical period; wherein, the status indicator data includes CPU utilization, memory utilization, GPU utilization, network bandwidth information, network input / output data, and disk status information; then, based on the server status indicator data for the target historical period, the indicator idleness degree for the target historical period is calculated; subsequently, based on the indicator idleness degree of the target historical period and a preset indicator idleness weight, the server computing power idleness degree for the target historical period is calculated; finally, based on a prediction model, the server computing power idleness degree for a target future period is predicted according to the server computing power idleness degree of the target historical period. This application can achieve accurate prediction of server computing power idleness.

Description

Technical Field

[0001] This application relates to the field of server monitoring technology, and in particular to methods, apparatus, devices and storage media for predicting server computing power idleness. Background Technology

[0002] With the rapid development of technologies such as artificial intelligence and large-scale models, computing power, as the core productivity of the data economy era, is experiencing explosive growth. To meet the growing demand for computing power, the scale of data center servers continues to expand, but at the same time, the average utilization rate of servers is very low, resulting in a large amount of idle computing resources.

[0003] To identify idle server resources, some technologies rely on manual checks of server status. While this can identify idle resources to some extent, it is inefficient and susceptible to subjective factors. Other technologies use historical load pre-configuration of resources, which, although requiring no manual intervention and achieving basic resource allocation, does not consider actual server performance differences and cannot handle sudden surges in business traffic. Furthermore, existing technologies only provide single-dimensional monitoring, leading to misjudgments of idle server resources. Therefore, accurately predicting server computing power idleness has become a pressing issue. Summary of the Invention

[0004] This application provides a method, apparatus, device, and storage medium for predicting server computing power idleness, in order to at least solve the problem of inaccurate prediction of server computing power idleness in related technologies.

[0005] This application provides a method for predicting server computing power idleness, including: Obtain server status indicator data for the target historical period; the status indicator data includes CPU utilization, memory utilization, graphics processor utilization, network bandwidth information, network input / output data, and disk status information; Calculate the idle rate of the indicators for the target historical period based on the status indicator data of the server. The server computing power idleness for the target historical period is calculated based on the indicator idleness and the preset indicator idleness weights. Based on the prediction model, the server computing power idleness in the target future period is predicted according to the server computing power idleness in the target historical period.

[0006] This application also provides a server computing power idleness prediction device, including: The acquisition module is used to acquire the status indicator data of the server for a target historical period. The status indicator data includes CPU utilization, memory utilization, graphics processor utilization, network bandwidth, network input / output data, and disk read / write. The indicator calculation module is used to calculate the indicator idleness of the target historical period based on the status indicator data of the server for the target historical period. The idleness calculation module is used to calculate the server computing power idleness for a target historical period based on the indicator idleness and preset indicator idleness weights. The prediction module is used to predict the server computing power idleness in the future based on the server computing power idleness in the target historical period, using a prediction model.

[0007] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for implementing the steps of any of the above-described server computing power idleness prediction methods when executing the computer program.

[0008] This application also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the steps of any of the above-described server computing power idleness prediction methods.

[0009] In some embodiments of this application, the technical solutions involve acquiring server status indicator data for a target historical period, including CPU utilization, memory utilization, GPU utilization, network bandwidth information, network input / output data, and disk status information. Based on this data, the idle rate of the server during the target historical period is calculated. Then, based on the idle rate and a preset idle rate weight, the server computing power idle rate for the target historical period is calculated. Finally, based on a prediction model, the server computing power idle rate for a future period is predicted. This approach allows for a comprehensive capture of the actual state of server computing power operation through multi-dimensional server status indicator data. By calculating and weighting the idle rate, an accurate representation of the overall server computing power utilization is obtained. Finally, a prediction model is used to predict the computing power idle rate for future periods based on historical idle rates. This solves the problem of inaccurate prediction of server computing power idle rates in some existing technologies. Attached Figure Description

[0010] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments 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.

[0011] Figure 1 A flowchart illustrating a server computing power idleness prediction method provided for some embodiments of this application; Figure 2 GRU model structure diagrams provided for some embodiments of this application; Figure 3 This is a schematic diagram of the overall architecture for the idleness assessment provided in this application; Figure 4 A schematic diagram illustrating the idleness assessment calculation steps provided for this application; Figure 5 A schematic diagram of the data preprocessing components provided for this application; Figure 6 A schematic diagram of a server computing power idleness prediction device provided for some embodiments of this application; Figure 7 A schematic diagram of the modules of an electronic device provided for some embodiments of this application. Detailed Implementation

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0013] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0014] 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.

[0015] Among related technologies, some rely on manual checks of server status to identify idle resources. While this can identify idle resources to some extent, it is inefficient and susceptible to subjective factors, failing to achieve efficient checks on large-scale servers. Other technologies use historical load pre-configuration of resources, achieving basic resource allocation without manual intervention, but they do not consider actual server performance differences and cannot handle sudden surges in business traffic. Furthermore, existing technologies only provide single-dimensional monitoring, leading to misjudgments of server idle resources. Therefore, accurately predicting server computing power idleness has become a pressing issue.

[0016] In view of this, this application provides a method for predicting server computing power idleness, which can solve the above problems. The server computing power idleness prediction method can be applied to servers. (See also...) Figure 1 This is a flowchart illustrating a server computing power idleness prediction method provided in some embodiments of this application. Figure 1 In this paper, the method for predicting server computing power idleness includes the following steps: Step S101: Obtain the status indicator data of the server for the target historical period; wherein, the status indicator data includes CPU utilization, memory utilization, graphics processor utilization, network bandwidth information, network input / output data and disk status information.

[0017] Specifically, the target historical period refers to the preset historical time range used for subsequent calculation of computing power idleness and training of prediction models. It is usually set to be no less than 15 days to comprehensively reflect the historical operating patterns of server computing power.

[0018] Specifically, status indicator data refers to the collection of various data generated by the server during its operation within a target historical period, which can characterize the operating status of the server's hardware resources.

[0019] Specifically, CPU (Central Processing Unit) utilization refers to the average operating load percentage of the CPUs on each node in a server cluster during a target historical period, reflecting the actual utilization of CPU resources.

[0020] Specifically, memory utilization rate refers to the average percentage of memory resources occupied by each node in the server cluster during a target historical period, which is used to reflect the actual utilization of memory resources.

[0021] Specifically, GPU (Graphics Processing Unit) utilization refers to the average percentage of GPU computing cores on each node in a server cluster during a target historical period, reflecting the actual utilization of GPU resources.

[0022] Specifically, network bandwidth information refers to relevant data on the network links of each node in the server cluster, including but not limited to total network bandwidth and actual bandwidth usage, which is used to reflect the actual utilization of network resources.

[0023] Specifically, network input / output data, or network I / O (Input / Output) data for short, refers to the amount of data received and sent by each node in the server cluster through the network within a target historical time period. It can also be used to derive disk I / O related time parameters.

[0024] Specifically, disk status information refers to the operating status data of the disks on each node in the server cluster, including but not limited to total disk capacity, unused capacity, disk I / O idle time, and total disk running time.

[0025] Understandably, the first step is to extract multi-dimensional computing power resource indicators from the server. The server idleness assessment tool obtains multi-dimensional computing power indicator status information from the BMC (Baseboard Management Controller) by sending IPMI (Intelligent Platform Management Interface) commands, without occupying server host resources. This status indicator data includes CPU utilization, memory utilization, graphics processor utilization, network bandwidth information, network input / output data, and disk status information.

[0026] Step S102: Calculate the idle rate of the indicators for the target historical period based on the status indicator data of the server for the target historical period.

[0027] Specifically, the idle rate of an indicator refers to the degree of idleness of each individual hardware resource of a server during a target historical period. Each indicator corresponds to a single hardware resource and can accurately reflect the unutilized status of that hardware resource during the target historical period.

[0028] Understandably, the status metrics data for each type of hardware resource are converted into quantifiable metrics such as idle rate.

[0029] Step S103: Calculate the server computing power idleness for the target historical period based on the indicator idleness and preset indicator idleness weights.

[0030] Specifically, the preset indicator idleness weight refers to the pre-set coefficient used to measure the degree of influence of each indicator idleness on the overall server computing power idleness, and the sum of all weight coefficients is 1.

[0031] Understandably, by comprehensively integrating hardware indicator idleness and combining it with preset indicator idleness weights, a weighted summation calculation method is used to convert the idleness of hardware resources into a quantitative value that can comprehensively reflect the overall idleness of server computing power, namely, server computing power idleness.

[0032] Step S104: Based on the prediction model, predict the server computing power idleness in the target future period according to the server computing power idleness in the target historical period.

[0033] Optionally, the prediction model refers to a model used to capture the temporal variation pattern of server computing power idleness and to predict the idleness in future periods. This application adopts the Gated Recurrent Unit (GRU) model, which can effectively mine the time dependency relationship in historical data. Through the synergistic effect of update gate and reset gate, it can accurately capture the pattern of server computing power idleness changing over time.

[0034] Optionally, the target future time period refers to the future time range within which the server's computing power idleness needs to be predicted. Its duration can be flexibly set according to actual business needs, and can be as long as ten minutes.

[0035] It is understandable that by taking the server computing power idleness of the target historical period as input, the prediction model extracts, fuses and processes the time series features to predict the server computing power idleness of the target future period.

[0036] In summary, the technical solutions of some embodiments of this application involve acquiring server status indicator data for a target historical period, including CPU utilization, memory utilization, GPU utilization, network bandwidth information, network input / output data, and disk status information; calculating the idle rate of the server indicators for the target historical period based on the server status indicator data; calculating the server computing power idle rate for the target historical period based on the idle rate of the server indicators and preset idle rate weights; and predicting the server computing power idle rate for a target future period based on the server computing power idle rate for the target historical period using a prediction model. This allows for a comprehensive capture of the actual state of server computing power operation through multi-dimensional server status indicator data. By calculating and weighting the idle rate, an accurate representation of the overall server computing power utilization is obtained. Furthermore, a prediction model is used to predict the computing power idle rate for future periods based on historical computing power idle rates. This solves the problem of inaccurate prediction of server computing power idle rates in some existing technologies.

[0037] In some embodiments, before performing step S102, which calculates the idle rate of the target historical period based on the status indicator data of the server for the target historical period, the method of this application further includes: Step a1: Identify abnormal and missing data in the status indicator data of the target historical period of the server.

[0038] Step a2 involves correcting and supplementing abnormal and missing data to obtain complete status indicator data for the target historical period of the server.

[0039] Specifically, abnormal data refers to data in the server's status indicators for a target historical period that deviates from the normal range and does not conform to the server's actual operating rules.

[0040] Specifically, missing data refers to missing data for a portion of the server's target historical status indicator data, or missing data for a portion of the indicators.

[0041] Optionally, statistical methods can be used to screen and identify the status indicator data. For example, by calculating the mean and standard deviation of each indicator data, data that deviates from the mean by 3 times the standard deviation can be identified as abnormal data. By traversing the status indicator data of the entire target historical period, missing data that has not been collected or is empty can be identified. For the identified abnormal data and missing data, corresponding methods can be used to correct them to obtain complete status indicator data of the target historical period of the server.

[0042] In the above embodiments, the preprocessing logic of status indicator data is improved by identifying and correcting abnormal and missing data, ensuring the integrity and accuracy of the status indicator data, and providing reliable data support for the subsequent calculation of server computing power idleness and the prediction of idleness in future periods.

[0043] In some embodiments, step S102, calculating the idle rate of the indicator for the target historical period based on the status indicator data of the server for the target historical period, includes: Step S1021: Calculate the CPU idle rate for the target historical period based on the CPU utilization rate of the server for the target historical period.

[0044] Step S1022: Calculate the memory idle rate for the target historical period based on the memory usage rate of the server for the target historical period.

[0045] Step S1023: Calculate the graphics processor idle rate for the target historical period based on the graphics processor utilization rate of the server during the target historical period.

[0046] Step S1024: Calculate the network idle rate for the target historical period based on the network bandwidth information of the server for the target historical period.

[0047] Step S1025: Calculate the disk idle rate for the target historical period based on the disk status information of the server for the target historical period; the idle rate indicator includes CPU idle rate, memory idle rate, graphics processor idle rate, network idle rate, and disk idle rate.

[0048] Specifically, the CPU idleness refers to the CPU idleness, which directly reflects the idle status of CPU resources. The average CPU utilization and total CPU capacity can be calculated from the CPU utilization of the server during the target historical period. The formula for calculating the CPU idleness is shown in expression (1).

[0049] (1) in, Indicates CPU idle time. , This indicates the percentage of CPU usage. The average CPU utilization rate indicates the average CPU utilization rate of each node in the server cluster during the target historical period. The total CPU capacity indicates the total CPU capacity of each node in the server cluster.

[0050] Specifically, memory idleness is an indicator that quantifies the unused portion of memory resources. Memory utilization rate determines available memory and total memory; available memory includes active memory and dynamically releaseable cache memory. The formula for calculating memory idleness is shown in expression (2).

[0051] (2) in, Indicates memory availability. , This indicates the percentage of memory used. Available memory represents the available memory of each node in the server cluster during the target historical period, while total memory represents the total memory of each node in the server cluster.

[0052] Specifically, GPU idleness refers to the idleness of GPU, which is an indicator that quantifies the unused portion of GPU resources. The average utilization rate of GPU computing cores and the total number of GPU cores can be calculated from the GPU utilization rate of the server target historical period. The formula for calculating GPU idleness is shown in expression (3).

[0053] (3) in, Indicates GPU idle time. , This indicates the percentage of GPU computing cores used. The average utilization rate of GPU computing cores indicates the average utilization rate of GPU computing cores on each node in the server cluster during the target historical period. The total number of GPU cores indicates the total number of GPU cores on each node in the server cluster.

[0054] Specifically, network idleness is an indicator that quantifies the unused portion of network resources, and the calculation formula for network idleness is shown in expression (4).

[0055] (4) in, Indicates network idleness. , This indicates the percentage of bandwidth used. Actual bandwidth usage indicates the actual network bandwidth usage of each node in the server cluster during the target historical period. Total network bandwidth indicates the total network bandwidth of each node in the server cluster during the target historical period.

[0056] Specifically, disk idleness is an indicator that quantifies the unused portion of disk resources.

[0057] Understandably, based on the corresponding status indicator data, the usage status of hardware resources is transformed into a quantifiable idle level through corresponding calculation formulas.

[0058] In the above embodiments, the idleness of the corresponding indicators is calculated for five hardware resources: CPU, memory, GPU, network, and disk. This improves the calculation logic of the idleness indicators, ensures the accuracy of the idleness calculation for each hardware resource, and provides reliable support for the subsequent calculation of server computing power idleness.

[0059] In some embodiments, step S1025, calculating the disk idle rate for the target historical period based on the disk status information of the server for the target historical period, includes: Step b1, based on the following formula, determine the disk idle rate for the target historical period:

[0060] in, Indicates disk idle time. , Indicates the percentage of remaining capacity. , This indicates the percentage of I / O idle time. Unused disk capacity and total disk capacity are determined by disk status information, while I / O idle time and total time are determined by network input / output data.

[0061] Understandably, the unused disk capacity and total disk capacity are first extracted from the disk status information, and the percentage of idle capacity is calculated. Then, the I / O idle time and total disk running time are extracted and derived from the network input and output data, and the percentage of I / O idle time is calculated. Then and Multiply each by 100% to convert the ratio to a percentage, and finally add the two percentage results to obtain the disk idle rate for the target historical period.

[0062] In the above embodiments, the calculation logic of disk idleness is improved by using the disk idleness calculation formula for the target historical period, ensuring the accuracy of disk idleness calculation and providing reliable support for the subsequent calculation of server computing power idleness.

[0063] In some embodiments, step S103, calculating the server computing power idleness for the target historical period based on the indicator idleness and a preset indicator idleness weight, includes: Step S1031: Determine the server computing power idle rate for the target historical period based on the following formula:

[0064] in, This indicates the server's computing power idleness during the target historical period. This represents the weighting coefficient corresponding to the idleness of each server metric. This represents the idleness of each indicator, and n represents the total number of indicators participating in the calculation of server computing power idleness.

[0065] Specifically, the weighting coefficients for each server idleness metric are set based primarily on the impact of each hardware metric on the server's computing power operation and industry prior knowledge. Furthermore, the weighting coefficients can be dynamically adjusted according to different business scenarios to ensure the rationality and adaptability of the weight allocation. Recommended weighting coefficients for different scenarios are shown in Table 1, and can be adjusted according to actual business needs.

[0066] Table 1. Weighting coefficients recommended for different scenarios

[0067] Understandably, the idleness of each indicator and its corresponding weight coefficient are first determined; then, the idleness of each indicator is multiplied by its corresponding weight coefficient to obtain the weighted contribution value of each hardware indicator's idleness to the overall computing power idleness; finally, all weighted contribution values ​​are summed to obtain the server computing power idleness.

[0068] In the above embodiments, the impact of different hardware resources on the overall computing power of the server is comprehensively considered by weighted summation, avoiding misjudgment of the overall computing power idleness by a single hardware indicator, and improving the comprehensiveness and accuracy of the server computing power idleness calculation results.

[0069] See also Figure 2 , Figure 2 The diagram illustrates the GRU model structure for some embodiments of this application. In some embodiments, step S104, based on a prediction model, predicts the server computing power idleness for a future target period according to the server computing power idleness for a target historical period, including: Step S1041: Input the server computing power idleness of the target historical period into the prediction model; wherein, the prediction model is used to capture the temporal dependency of the server computing power idleness.

[0070] Step S1042: Extract the temporal features of server computing power idleness through the prediction model.

[0071] Step S1043: The temporal features of server computing power idleness are fused using the prediction model to output hidden state features.

[0072] Step S1044: Determine the server computing power idleness for the target future time period based on the hidden state characteristics.

[0073] Specifically, time-series characteristics refer to the changing patterns of server computing power idleness over time during a target historical period. They can intuitively reflect the inherent logic of server computing power idleness changing over time and serve as the basis for future predictions.

[0074] Specifically, hidden state features refer to the features output by the prediction model after fusing time-series features. In essence, they represent the time-series patterns of historical computing power idleness and contain all the key information that affects future changes in idleness.

[0075] Optionally, the update gate of the prediction model is the core gating unit in the GRU model that controls the proportion of historical information retention and the proportion of new information reception. The output value range is [0,1], and the formula is shown in expression (5).

[0076] (5) in, Indicates an update to the door; Use the Sigmoid activation function; To update the weight matrix corresponding to the gate, For the corresponding bias term; This is the hidden state from the previous time step, i.e., the model at time t. The temporal features learned at time step 1; Input data for the current moment, which is the time series data of server computing power idleness for the target historical period.

[0077] Optionally, the reset gate of the prediction model is a gating unit in the GRU model that controls the degree of influence of the historical hidden state on the current candidate state, with an output value range of [0,1], as shown in expression (6).

[0078] (6) in, Indicates that the door is being reset; To reset the weight matrix corresponding to the gate, This is the corresponding bias term.

[0079] Optionally, the candidate hidden state is the potential state generated by the historical information after the gate filtering is reset and the current input information, as shown in expression (7).

[0080] (7) in, Indicates the candidate hidden state; This is the weight matrix corresponding to the candidate states. For the corresponding bias term; It is the hyperbolic tangent activation function.

[0081] Optionally, the current hidden state is the temporal feature of the final output of the GRU model, as shown in expression (8).

[0082] (8) in, This indicates the current hidden state.

[0083] Optionally, firstly, the continuous state data is divided into fixed-length samples, with data collected at fixed intervals as one sample point, such as sampling once per minute. Using the past 60 minutes of data as samples, the server computing power idleness for the next 10 minutes is predicted. The training set, validation set, and test set can be divided in a 7:1:2 ratio. Next, a one-layer GRU neural network is constructed, with 32 GRU units per layer. The number of neural network layers and the number of GRU units per layer can be adjusted based on the prediction results. Finally, a fully connected layer and an output layer are added at the end of the network. The fully connected layer fuses features, and the output layer selects a linear activation function to calculate the idleness assessment. The prediction model uses the mean squared error loss function to measure the difference between the predicted idleness score and the actual score. The optimizer uses the Adam optimizer, which adaptively adjusts the learning rate to accelerate model training convergence.

[0084] Optionally, after each epoch (training round), performance is evaluated using a validation set, and the RMSE (Root Mean Square Error) value of the loss function and evaluation metrics are monitored. Based on the validation results, hyperparameters such as the learning rate, the number of GRU units, and the number of network layers are adjusted to avoid overfitting or underfitting.

[0085] Understandably, new data can be input to fine-tune the top-level weights of the GRU, further ensuring the accuracy and adaptability of the top-level predictions. Prediction accuracy can be continuously optimized by loading pre-trained weights, freezing non-top-level parameters, appropriately setting the loss function and learning rate, and iterative training until convergence, ensuring the model can adapt to data changes and adjustments in business scenarios. For example, the number of neural network layers and the number of GRU units per layer can be dynamically adjusted. If the business scenario has higher requirements for capturing temporal patterns, the number of layers or units can be increased; if it is necessary to improve inference speed, the number of layers or units can be appropriately reduced.

[0086] In the above embodiments, the underlying logic of the GRU model is improved through clear formula and model structure design, and the complete process of temporal feature extraction, hidden state update and result output is clearly explained, which further improves the accuracy and reliability of server computing power idleness prediction.

[0087] In some embodiments, the method of this application further includes: Step c1: Set the server computing power idleness threshold.

[0088] Step c2: If the server computing power idleness in the target future time period is greater than the server computing power idleness threshold, generate an alarm message.

[0089] Specifically, the server computing power idleness threshold is a pre-set critical value used to determine whether the server computing power is in an excessively idle state, and it is the judgment standard for triggering alarm information.

[0090] Understandably, the system first obtains the server's computing power idle rate for the target future time period, and then compares this idle rate value with the server computing power idle rate threshold. If the server computing power idle rate for the target future time period is greater than the server computing power idle rate threshold, it indicates that the server will be in an excessively idle state in the future, and the system automatically generates an alarm message. If the server computing power idle rate for the target future time period is less than or equal to the server computing power idle rate threshold, it indicates that the server's future computing power idle state is within a reasonable range, and no alarm is triggered.

[0091] In the above embodiments, by setting thresholds and triggering alarms, it is ensured that excessive idle computing power can be detected and dealt with in a timely manner, thereby improving the practicality of the server computing power idleness prediction method.

[0092] See also Figure 3 , Figure 3 This is a schematic diagram of the overall architecture for the idleness assessment provided in this application. Figure 3 The process begins with data collection; followed by data preprocessing; then, the GRU model is constructed and trained, and the preprocessed data is input into the GRU model; finally, the idleness level is assessed and an early warning is issued.

[0093] See also Figure 4 , Figure 4 A schematic diagram illustrating the calculation steps for the idleness assessment provided in this application. Figure 4 First, the idle rate assessment tool sends IPMI commands to obtain real-time values ​​of indicators such as CPU utilization, memory utilization, GPU utilization, network bandwidth information, network input / output data, and disk status information. Then, data preprocessing, including data cleaning and data segmentation, is performed. Next, idle rate labels are defined, and idle rate indicators are determined to include CPU idle rate, memory idle rate, GPU idle rate, network idle rate, and disk idle rate. Then, the GRU model is built and trained, and the historical server computing power idle rate calculated through the idle rate labels is input into the GRU model. Finally, the GRU model calculates the idle rate based on real-time data and issues an early warning.

[0094] See also Figure 5 , Figure 5 A schematic diagram of the data preprocessing components provided for this application. Figure 5 In this context, data preprocessing includes data cleaning and data segmentation.

[0095] Corresponding to the server computing power idleness prediction method, this application also provides a server computing power idleness prediction device. (See also...) Figure 6 This is a schematic diagram of a server computing power idleness prediction device provided in some embodiments of this application. Figure 6 The server computing power idleness prediction device includes: The acquisition module 601 is used to acquire the status indicator data of the server for a target historical period; wherein, the status indicator data includes CPU utilization, memory utilization, graphics processor utilization, network bandwidth, network input / output data, and disk read / write. The indicator calculation module 602 is used to calculate the indicator idleness of the target historical period based on the status indicator data of the server for the target historical period. The idleness calculation module 603 is used to calculate the server computing power idleness of the target historical period based on the indicator idleness and the preset indicator idleness weight. The prediction module 604 is used to predict the server computing power idleness in the future period based on the server computing power idleness in the target historical period, using a prediction model.

[0096] In some embodiments, the device further includes: The identification unit is used to identify abnormal and missing data in the status indicator data of the target historical period of the server.

[0097] The preprocessing unit is used to correct and supplement abnormal and missing data to obtain complete status indicator data for the target historical period of the server.

[0098] In some embodiments, the indicator calculation module 602 includes: The CPU idle rate calculation unit is used to calculate the CPU idle rate for a target historical period based on the CPU utilization rate of the server for that period.

[0099] The memory idleness calculation unit is used to calculate the memory idleness of a target historical period based on the memory usage rate of the server during that period.

[0100] The graphics processor idleness calculation unit is used to calculate the graphics processor idleness for a target historical period based on the graphics processor utilization rate of the server for that period.

[0101] The network idleness calculation unit is used to calculate the network idleness of the target historical period based on the network bandwidth information of the server for the target historical period.

[0102] The disk idleness calculation unit is used to calculate the disk idleness for a target historical period based on the disk status information of the server for that period. The idleness index includes CPU idleness, memory idleness, graphics processor idleness, network idleness, and disk idleness.

[0103] In some embodiments, the disk idleness calculation unit includes: The disk idleness calculation subunit is used to determine the disk idleness for a target historical period based on the following formula:

[0104] in, This indicates the percentage of remaining capacity. This indicates the percentage of I / O idle time. The unused disk capacity and total disk capacity are determined by disk status information, while the I / O idle time and total time are determined by network input / output data.

[0105] In some embodiments, the idle time calculation module 603 includes: The idle rate calculation unit is used to determine the disk idle rate for a target historical period based on the following formula:

[0106] in, Indicates disk idle time. , Indicates the percentage of remaining capacity. , This indicates the percentage of I / O idle time. Unused disk capacity and total disk capacity are determined by disk status information, while I / O idle time and total time are determined by network input / output data.

[0107] In some embodiments, the prediction module 604 includes: The input unit is used to input the server computing power idleness of the target historical period into the prediction model; wherein, the prediction model is used to capture the temporal dependency of the server computing power idleness.

[0108] The feature extraction unit is used to extract the temporal features of server computing power idleness through the prediction model.

[0109] The feature fusion unit is used to fuse the temporal features of server computing power idleness through a prediction model and output hidden state features.

[0110] The output unit is used to determine the server computing power idleness of the target in future time periods based on the hidden state characteristics.

[0111] In some embodiments, the device further includes: The threshold setting unit is used to set the server computing power idleness threshold.

[0112] The alarm unit is used to generate alarm information when the server's computing power idleness exceeds the server computing power idleness threshold in a future time period.

[0113] For a description of the features in the embodiment of the server computing power idleness prediction device, please refer to the relevant description of the embodiment of the sample data processing method, which will not be repeated here.

[0114] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0115] See also Figure 7 The embodiments of this application also provide an electronic device, including a memory 10 and a processor 20, wherein the memory 10 stores a computer program and the processor 20 is configured to run the computer program to perform the steps in any of the above embodiments of the server computing power idleness prediction method.

[0116] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above embodiments of the server computing power idleness prediction method when running.

[0117] 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 a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0118] The embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the server computing power idleness prediction method.

[0119] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in any of the above embodiments of the server computing power idleness prediction method.

[0120] 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 implementation should not be considered beyond the scope of this application.

[0121] The foregoing has provided a detailed description of a server computing power idleness prediction method, apparatus, device, and storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only intended to help understand the method and core ideas of this application. 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 the claims of this application.

Claims

1. A method for predicting server computing power idleness, characterized in that, The method includes: Obtain server status indicator data for a target historical period; wherein, the status indicator data includes CPU utilization, memory utilization, graphics processor utilization, network bandwidth information, network input / output data, and disk status information; Calculate the idle rate of the indicators for the target historical period based on the status indicator data of the server for the target historical period; The server computing power idleness for the target historical period is calculated based on the indicator idleness and the preset indicator idleness weight. Based on the prediction model, the server computing power idleness in the target future period is predicted according to the server computing power idleness in the target historical period.

2. The computing power idleness prediction method according to claim 1, characterized in that, Before calculating the idle rate of the indicators for the target historical period based on the status indicator data of the server for the target historical period, the method further includes: Identify abnormal and missing data in the status indicator data of the target historical period of the server; The abnormal and missing data are corrected and supplemented to obtain complete status indicator data for the target historical period of the server.

3. The computing power idleness prediction method according to claim 1, characterized in that, The step of calculating the idle rate of indicators for the target historical period based on the status indicator data of the server for the target historical period includes: Calculate the CPU idle rate for the target historical period based on the CPU utilization rate of the server during the target historical period; Calculate the memory idle rate for the target historical period based on the memory usage rate of the server during the target historical period; Calculate the graphics processor idle rate for the target historical period based on the graphics processor utilization rate of the server during the target historical period; Calculate the network idle rate for the target historical period based on the network bandwidth information of the server for the target historical period; Calculate the disk idle rate for the target historical period based on the disk status information of the server for that period. The idle rate indicators include CPU idle rate, memory idle rate, graphics processor idle rate, network idle rate, and disk idle rate.

4. The computing power idleness prediction method according to claim 3, characterized in that, The step of calculating the disk idle rate for the target historical period based on the disk status information of the server for the target historical period includes: The disk idle rate for the target historical period is determined based on the following formula: in, Indicates disk idle time. , Indicates the percentage of remaining capacity. , This indicates the percentage of I / O idle time. Unused disk capacity and total disk capacity are determined by disk status information, while I / O idle time and total time are determined by network input / output data.

5. The computing power idleness prediction method according to claim 1, characterized in that, The step of calculating the server computing power idleness for the target historical period based on the indicator idleness and preset indicator idleness weights includes: The server computing power idleness for the target historical period is determined based on the following formula: in, This indicates the server's computing power idleness during the target historical period. This represents the weighting coefficient corresponding to the idleness of each server metric. This represents the idleness of each indicator, and n represents the total number of indicators participating in the calculation of server computing power idleness.

6. The computing power idleness prediction method according to claim 1, characterized in that, The method of predicting server computing power idleness in future periods based on the prediction model and the server computing power idleness in the target historical period includes: The server computing power idleness during the target historical period is input into the prediction model; wherein, the prediction model is used to capture the temporal dependency of the server computing power idleness. The temporal characteristics of the server's computing power idleness are extracted using the prediction model. The prediction model is used to fuse the temporal features of the server's computing power idleness and output the hidden state features. Based on the hidden state characteristics, the server computing power idleness in the target future time period is determined.

7. The computing power idleness prediction method according to claim 1, characterized in that, The method further includes: Set the server computing power idle threshold; An alarm message is generated if the server computing power idleness level is greater than the server computing power idleness threshold during the target future time period.

8. A server computing power idleness prediction device, characterized in that, The device includes: The acquisition module is used to acquire the status indicator data of the server for a target historical period; wherein, the status indicator data includes CPU utilization, memory utilization, graphics processor utilization, network bandwidth, network input / output data, and disk read / write. The indicator calculation module is used to calculate the indicator idleness of the target historical period based on the status indicator data of the server for the target historical period. The idleness calculation module is used to calculate the server computing power idleness of the target historical period based on the indicator idleness and the preset indicator idleness weight. The prediction module is used to predict the server computing power idleness in the future period based on the server computing power idleness in the target historical period, using a prediction model.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the server computing power idleness prediction method 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, wherein when the computer program is executed by a processor, it implements the steps of the server computing power idleness prediction method as described in any one of claims 1 to 7.

Images