Server aging stability test method and device

Abstract

The application discloses a kind of server aging stability test method and device, it is related to server stability test technical field, comprising: obtaining server model, the component model parameter of server is found by server model;Through aging prediction model, the component of this server is aged prediction;According to the performance prediction model of component aging prediction data, the performance of component when the server is aged is predicted;According to the predicted aging performance data, the parameter adjustment data of component is generated, to construct server aging stability test scheme by parameter adjustment data;Server aging stability test scheme is used to test server, solve the technical problem that cannot effectively simulate the stability of each component aging when server runs in the related art, reaches the technical effect that the running stability of server in different time periods is efficiently tested when aging, guarantees the accuracy of simulating server component aging degree.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of server stability testing, in particular to a server aging stability testing method and device. BACKGROUND

[0002] With the expansion of the application scale and the improvement of the functional complexity of computer products, its performance has become the focus of the industry, at this time, the importance of performance testing carried out to verify the system performance is increasingly prominent.

[0003] However, the components of the server are gradually aging during use, which greatly affects the working performance and running stability, and the related technology cannot effectively simulate the stability of the server running when the components are aging, which needs to be solved urgently. SUMMARY

[0004] The present application provides a server aging stability testing method and device to at least solve the technical problem that the related art cannot effectively simulate the stability of the server running when the components are aging.

[0005] The present application provides a server aging stability testing method, comprising the following steps: collecting historical aging data of a plurality of component parts in a target server, and determining whether the historical aging data meets a preset quantity requirement, and in the case that the historical aging data does not meet the quantity requirement, performing an aging simulation experiment on the corresponding component parts to generate experimental aging data, and training a pre-constructed aging prediction model through the experimental aging data and the historical aging data; determining the component part model parameters corresponding to the target server based on the server identifier of the target server, and inputting the component part model parameters into the trained aging prediction model to output the component part aging parameters corresponding to the target server; based on the component part aging parameters and the pre-constructed performance prediction model, predicting the aging component performance parameters of the target server in different time periods, and obtaining the component parameter data of the plurality of component parts, and determining the parameter adjustment requirement of the corresponding component parts according to the component parameter data and the aging component performance parameters, and adjusting the component parameter data through the parameter adjustment requirement to generate corresponding component parameter adjustment data; generating the aging stability testing scheme corresponding to the target server through the component parameter adjustment data, and performing aging stability testing on the target server according to the aging stability testing scheme to obtain the aging stability result of the target server.

[0006] The application further provides a server aging stability test device, comprising: an aging prediction model training module, configured to collect historical aging data of a plurality of component parts in a target server, and determine whether the historical aging data meets a preset quantity requirement, and in the case that the historical aging data does not meet the quantity requirement, perform an aging simulation experiment on the corresponding component parts to generate experimental aging data, and train a pre-constructed aging prediction model through the experimental aging data and the historical aging data; a component aging prediction module, configured to determine component part model parameters corresponding to the target server based on a server identifier of the target server, and input the component part model parameters into the trained aging prediction model to output component aging parameters corresponding to the target server; a component parameter adjustment module, configured to predict aging component performance parameters of the target server in different time periods based on the component aging parameters and a pre-constructed performance prediction model, and obtain component parameter data of the plurality of component parts, and determine parameter adjustment requirements of the corresponding component parts according to the component parameter data and the aging component performance parameters, and adjust the component parameter data through the parameter adjustment requirements to generate corresponding component parameter adjustment data; and a stability test module, configured to generate an aging stability test scheme corresponding to the target server through the component parameter adjustment data, and perform an aging stability test on the target server according to the aging stability test scheme to obtain an aging stability result of the target server.

[0007] The application further provides an electronic device, comprising: a memory configured to store a computer program; and a processor configured to execute the computer program to implement the steps of any of the server aging stability test methods.

[0008] The application further provides a non-volatile computer readable storage medium, wherein the non-volatile computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of any of the server aging stability test methods.

[0009] The application further provides a computer program product, comprising a computer program, and the computer program is executed by a processor to implement the steps of any of the server aging stability test methods.

[0010] By the present application, the component model parameters corresponding to the target server can be determined based on the server identifier of the target server; the component model parameters are input into the trained aging prediction model to output the component aging parameters corresponding to the target server; based on the component aging parameters and the pre-constructed performance prediction model, the aging component performance parameters of the target server in different time periods are predicted, the component parameter data of the plurality of components are obtained, and the parameter adjustment requirements of the corresponding components are determined according to the component parameter data and the aging component performance parameters, and the component parameter data are adjusted through the parameter adjustment requirements to generate corresponding component parameter adjustment data; the aging stability test scheme corresponding to the target server is generated through the component parameter adjustment data, and the aging stability test of the target server is performed according to the aging stability test scheme to obtain the aging stability result of the target server, so that the technical problem that the server running stability of each component aging cannot be effectively simulated in the related art can be solved, and the technical effects that the running stability of the server in different time periods when aging is efficiently tested, the accuracy of simulating the aging degree of each component of the server is ensured, and the effectiveness of the server aging stability test result is ensured are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present application, the drawings needed in the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0012] Figure 1 A flowchart of a server aging stability test method according to an embodiment of the present application is provided.

[0013] Figure 2 An execution logic schematic diagram of a server aging stability test method according to an embodiment of the present application is provided.

[0014] Figure 3 An execution logic schematic diagram of a server aging stability test method according to an embodiment of the present application is provided.

[0015] Figure 4 An execution logic schematic diagram of a server component aging prediction according to an embodiment of the present application is provided.

[0016] Figure 5 An example diagram of a server aging stability test device according to an embodiment of the present application is provided.

[0017] Among them, the 10-server aging stability test device, the 100-aging prediction model training module, the 200-component aging prediction module, the 300-component parameter adjustment module, and the 400-stability test module. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0019] It should be noted that in the description of the present application, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. The terms "first", "second" and the like in the present application are used to distinguish similar objects, not to describe a specific order or sequence.

[0020] In order 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 drawings and specific embodiments.

[0021] In conjunction with the specific application environment architecture or specific hardware architecture on which the server aging stability test method is executed, the specific application environment architecture or specific hardware architecture is described here.

[0022] The embodiments of the present application provide a server aging stability test method.

[0023] As shown in Figure 1 The server aging stability test method includes the following steps:

[0024] In step S101, historical aging data of a plurality of component parts in the target server is collected, and it is determined whether the historical aging data meets a preset quantity requirement. If the historical aging data does not meet the quantity requirement, aging simulation experiments are performed on the corresponding component parts to generate experimental aging data, and the pre-constructed aging prediction model is trained by the experimental aging data and the historical aging data.

[0025] The embodiment of the present application can first collect historical aging data of each component of the target server, and detect whether the data meets the preset quantity requirement; if not, aging simulation experiments are carried out for the corresponding components to generate experimental aging data, and the pre-constructed aging prediction model is trained in combination with the historical aging data.

[0026] Therefore, the embodiment of the present application can supplement insufficient aging data to ensure sufficient model training data, improve the accuracy of the aging prediction model, help accurately predict the aging condition of the server components, and ensure stable operation of the server.

[0027] Optionally, in an embodiment of the present application, before training the pre-constructed aging prediction model by the experimental aging data and the historical aging data, it further includes: extracting data features corresponding to the aging prediction training sample set, wherein the data features include data dimension type, data time sequence correlation and data feature complexity; determining the corresponding convolutional neural network structure requirement according to the data features, and screening a plurality of candidate convolutional neural network structures meeting the convolutional neural network structure requirement from a pre-set convolutional neural network structure library; inputting the aging prediction training sample set into the plurality of candidate convolutional neural network structures respectively to train and test the performance of the plurality of candidate convolutional neural network structures, and generating corresponding preliminary performance test results; based on the preliminary performance test results, selecting a target candidate convolutional neural network structure meeting the pre-set performance requirement to construct the aging prediction model by using the target candidate convolutional neural network structure.

[0028] It should be noted that the embodiment of the present application first extracts a plurality of data features from the aging prediction training sample set (including runtime length, fault record, performance attenuation data of different server component models, etc.), and the specific data features are as follows:

[0029] 1. Data dimension type: distinguishing component basic parameters (such as CPU core number, hard disk capacity), time sequence performance data (such as memory read / write speed monthly change) and the like;

[0030] 2. Data time sequence correlation: analyzing the change law of performance indicators with time (such as the correlation between the number of hard disk bad tracks and the use time);

[0031] 3. Data feature complexity: judging whether the feature has a nonlinear relationship (such as the complex mapping of CPU temperature and load rate).

[0032] Secondly, the embodiment of the present application can determine the requirement of the convolutional neural network (CNN) structure according to the data characteristics. If the time sequence correlation is strong, the time sequence convolution layer needs to be increased. If the feature complexity is high, the network depth needs to be improved. Then, the embodiment of the present application can select 3-5 candidate structures (such as a customized CNN containing 3 layers of time sequence convolution layers) that meet the requirements from the preset CNN structure library (containing AlexNet and other structures adapted to different scenes).

[0033] Thirdly, the embodiment of the present application can divide the training sample set into a training set and a test set according to 8:2, respectively input each candidate structure, train the model (iteratively adjust the convolution kernel size, the number of nodes of the full connection layer and other parameters) by using the training set, test the performance of the model such as the aging prediction accuracy and the error rate by using the test set, and thus generate preliminary performance test results (such as the accuracy of A candidate structure is 92% and the error rate is 5%).

[0034] After that, the embodiment of the present application can compare the preliminary results to select a target candidate structure that meets the preset performance requirements (such as the accuracy is ≥90% and the error rate is ≤8%), and construct the final aging prediction model based on this.

[0035] Therefore, the embodiment of the present application customizes and screens the CNN structure based on the sample data characteristics, thereby guaranteeing the model adaptability and prediction accuracy, and providing reliable model support for server component aging prediction.

[0036] Optionally, in an embodiment of the present application, in the process of determining the requirement of the corresponding convolutional neural network structure according to the data characteristics, and screening multiple candidate convolutional neural network structures that meet the requirement of the convolutional neural network structure from the preset convolutional neural network structure library, the convolutional neural network structure can also be customized in combination with the server component correlation, and the specific process is as follows:

[0037] Step 1: The time sequence data of each component of the server is associated and layered. The components are divided into a strong correlation group (entropy value ≥0.7), a weak correlation group (0.3<entropy value<0.7) and an unrelated group (entropy value ≤0.3) by calculating the mutual information entropy value, so as to obtain the corresponding correlation layered results, and provide grouping basis for branch structure design;

[0038] Step 2: Based on the correlation layered results, the basic branch structure is constructed. The strong correlation group shares the feature extraction backbone and connects the time sequence attention module in series, the weak correlation group adopts independent branches, and the unrelated group is configured with completely independent branches, so as to obtain the basic branch structure, and provide framework support for the design of dynamic fusion mechanism.

[0039] Step 3: embedding an adaptive fusion unit in the basic branch structure, which dynamically adjusts the cross-fusion weight of branch features according to the real-time calculated component correlation degree (updated every 30 seconds), converts the correlation degree into a 0-1 weight value through a sigmoid function, and increases the cross-fusion weight when the correlation degree increases, thereby obtaining a CNN structure with a dynamic fusion mechanism, which is used as the basic architecture for model training to improve the adaptability to changes in component correlation.

[0040] Therefore, the embodiment of the application can accurately capture the server component correlation characteristics, thereby dynamically adapting to correlation changes, improving the feature extraction capability of the CNN for complex correlation scenarios, and enhancing the aging prediction accuracy.

[0041] Optionally, in an embodiment of the application, historical aging data of a plurality of component parts in a target server is collected, and it is determined whether the historical aging data meets a preset quantity requirement, and in the case that the historical aging data does not meet the quantity requirement, an aging simulation experiment is performed on the corresponding component part to generate experimental aging data, and the pre-constructed aging prediction model is trained based on the experimental aging data and the historical aging data, including: obtaining historical aging data of a plurality of component parts in a target server, and determining the data quantity corresponding to the historical aging data, and determining whether the data quantity of the historical aging data meets a preset quantity requirement; when the data quantity of the historical aging data meets the quantity requirement, a corresponding aging prediction training sample set is constructed based on the historical aging data, and the aging prediction model is trained based on the aging prediction training sample set; when the data quantity of the historical aging data does not meet the quantity requirement, the target component part that does not meet the quantity requirement is determined, and an aging simulation experiment is performed on the target component part, and experimental equipment data corresponding to the target component part is recorded, and experimental aging data of the target component part is generated based on the experimental equipment data, and a corresponding aging prediction training sample set is constructed based on the experimental aging data and the historical aging data, and the aging prediction model is trained based on the aging prediction training sample set.

[0042] In actual execution, the embodiment of the application can first collect historical aging data of a plurality of component parts such as CPU, memory, hard disk, etc. in a target server, including the use time of each component, the performance decay curve (such as CPU frequency drop amplitude, hard disk read-write speed decay rate), fault occurrence time, etc. data, and the data quantity of each component historical aging data is counted.

[0043] Secondly, the embodiment of the present application can compare the statistical data quantity with the preset quantity requirement (such as at least 1000 valid data for each component), if the historical data of a component (such as a new model NVMe (Non Volatile Memory Express, Non-volatile memory host controller interface specification) hard disk) is only 300, which does not meet the requirement, then the aging simulation experiment is carried out to generate the experimental aging data (such as the performance degradation data of 3 years of simulation) of the component.

[0044] If the historical data of a component (such as a traditional SATA (Serial Advanced Technology Attachment, serial hard disk)) reaches 1200, which meets the quantity requirement, then the historical data is directly used to construct the aging prediction training sample set; the experimental aging data and the existing historical data are combined and supplemented to the sample set, and finally the aging prediction training sample set covering all components and sufficient data quantity is formed, which is used to train the aging prediction model.

[0045] Therefore, the embodiment of the present application can carry out aging simulation experiment on a component of the server when the aging data of the component is less, record the equipment data of the component during the experiment, and generate the aging data of the component according to the obtained experimental data, so as to expand the aging prediction training sample set, thereby improving the prediction accuracy of the aging prediction model for the aging degree of each component of the server, and ensuring the effectiveness of the server aging stability test result.

[0046] Optionally, in an embodiment of the present application, the historical aging data of the plurality of component parts in the target server is obtained, comprising: collecting the original historical aging data of the plurality of component parts in the target server, wherein the original historical aging data includes component running time data, performance degradation data, environmental influence data and fault record data; performing data preprocessing operation on the original historical aging data to obtain corresponding standard historical aging data; performing reliability verification on the standard historical aging data based on the preset data reliability verification rule to obtain corresponding reliability verification result, and screening the historical aging data meeting the preset reliability requirement from the standard historical aging data according to the reliability verification result.

[0047] It should be noted that the embodiment of the present application can first collect the original historical aging data of each component part (such as CPU, memory, hard disk, etc.) of the target server, which is described as follows:

[0048] 1. Component running time data: cumulative boot time, continuous running time, etc.

[0049] 2. Performance degradation data: CPU frequency reduction, memory read / write speed attenuation rate, hard disk IOPS (Input Output Operations Per Second) reduction value, etc.

[0050] 3. Environmental impact data: operating environment temperature, humidity, voltage fluctuation record, etc.

[0051] 4. Fault record data: component error time, fault type, repair and replacement record, etc.

[0052] Secondly, the embodiments of the present application can preprocess the original historical aging data; for example, the embodiments of the present application can obtain standard historical aging data by eliminating repeated data (such as repeated temperature records at the same time point), filling missing values (using adjacent period mean to supplement short-time missing performance data), and unifying data format (converting time data in different units to hours).

[0053] Then, the embodiments of the present application verify according to the preset data reliability verification rules (such as "the performance data continuous collection interval is not more than 1 hour", "the fault record needs to match the repair order") to generate reliability verification results, and filter the data meeting the preset reliability requirements (such as data integrity rate ≥ 95%, abnormal value proportion ≤ 3%) from the standard historical aging data as the final historical aging data.

[0054] Thus, the embodiments of the present application ensure the accuracy and reliability of the historical aging data by multi-dimensional collection and strict processing and verification, thereby providing a high-quality data basis for subsequent model training and improving the aging prediction accuracy.

[0055] Optionally, in an embodiment of the present application, when the data quantity of the historical aging data does not meet the quantity requirement, a target component part that does not meet the quantity requirement is determined, and an aging simulation simulation experiment is performed on the target component part, and experimental equipment data corresponding to the target component part is recorded, including: determining actual operating environment parameters and aging influencing factors of the target component part, to determine a plurality of experimental parameters of the aging simulation simulation experiment according to the actual operating environment parameters and the aging influencing factors, wherein the plurality of experimental parameters include environmental stress parameters, operating load parameters and experimental duration parameters; based on the plurality of experimental parameters, an aging simulation simulation experiment environment of the target component part is built, and the aging simulation simulation experiment is started in the aging simulation simulation experiment environment, and current equipment data of the target component part in the aging simulation simulation experiment process is recorded, wherein the current equipment data includes component operating state data, performance change data and stress response data; the current equipment data is subjected to a data cleaning operation to obtain corresponding current standard equipment data, and a performance degradation trend feature and an aging state feature in the current standard equipment data are extracted to generate experimental aging data of the target component part according to the performance degradation trend feature and the aging state feature.

[0056] In an embodiment of the present application, when the historical aging data of the target component part (such as a new model SSD (Solid State Drive, solid state disk)) is insufficient, the actual operating environment parameters (such as room temperature 25-30℃, humidity 40-60%) and the aging influencing factors (such as read-write frequency, continuous high-load duration) thereof are analyzed first, and a plurality of experimental parameters are determined accordingly, as follows:

[0057] 1. Environmental stress parameters: simulate temperature fluctuation (20-40℃ cycle), voltage instability (±5% deviation);

[0058] 2. Operating load parameters: set different read-write intensity (such as 50% / 100% load alternation), data throughput (100MB / s-1GB / s);

[0059] 3. Experimental duration parameters: set the experimental period equivalent to 3 years of actual use according to the accelerated aging model.

[0060] Based on the above experimental parameters, a simulation environment (such as a high-low temperature box combined with a load generator) is built, and current equipment data is recorded in real time after starting the experiment, as follows:

[0061] 1. Operating state data: real-time temperature, power consumption, fan speed;

[0062] 2. Performance change data: random read-write IOPS attenuation, response time extension value;

[0063] 3. Stress response data: error check frequency under high load, data retransmission frequency.

[0064] After that, the embodiments of the present application clean the collected data (such as eliminating sensor outliers, smoothing transient fluctuations) to obtain current standard device data, and extract corresponding feature data (such as a 2% IOPS decline per month attenuation trend, an error rate surge in high temperature aging state feature) from it, and finally generate experimental aging data consistent with the actual aging law.

[0065] Therefore, the embodiments of the present application accurately simulate the component aging process through simulation experiments, thereby generating high-quality supplementary data, effectively solving the problem of insufficient historical data, and improving the training effect of the aging prediction model.

[0066] In step S102, the component model parameters corresponding to the target server are determined based on the server identifier of the target server, and the component model parameters are input into the trained aging prediction model to output the component aging parameters corresponding to the target server.

[0067] Further, the embodiments of the present application can also obtain the server identifier (i.e. server model) of the server, and find the component model parameters of the server through the server model, thereby providing reliable data guidance and basis for obtaining subsequent component aging parameters.

[0068] Optionally, in an embodiment of the present application, determining the component model parameters corresponding to the target server based on the server identifier of the target server comprises: reading the server model of the target server, and inputting the server model into a preset association database to extract the component model parameters corresponding to the plurality of components in the target server.

[0069] As an implementable way, the embodiments of the present application can first read the server model information (such as “PowerEdge R750”) through the management interface (such as IPMI (Intelligent Platform Management Interface, intelligent platform management interface) interface) of the server or operating system command (such as dmidecode command of Linux). The model information can be used as a retrieval keyword and input into a preset association database, which pre-stores the mapping relationship between server models of various brands and components, including CPU (Central Processing Unit, central processor), memory, hard disk, motherboard, power supply and other core components.

[0070] When searching the database, this application embodiment can match all component entries corresponding to the target model and extract the model parameters of each component. For example, for "PowerEdge R750", it can extract detailed parameters such as CPU supported model (e.g., "Intel XeonGold 6338"), memory specifications, and hard disk interface type, and finally generate a structured list of component model parameters.

[0071] Therefore, the embodiments of this application can quickly match and extract component parameters by server model, thereby effectively improving the efficiency and accuracy of parameter acquisition and providing reliable data support for server maintenance and upgrades.

[0072] Furthermore, embodiments of this application can use an aging prediction model to predict the aging of each component of the server based on the component model parameters, thereby generating corresponding prediction data.

[0073] Therefore, the embodiments of this application can use component model parameters to quickly predict the aging of server components, which helps to improve the accuracy of aging trend prediction and provides a reliable basis for server maintenance and replacement.

[0074] In step S103, based on component aging parameters and a pre-built performance prediction model, the performance parameters of the aging components of the target server are predicted for different time periods. Component parameter data of multiple components are obtained, and the parameter adjustment requirements of the corresponding components are determined according to the component parameter data and the performance parameters of the aging components. The component parameter data is adjusted according to the parameter adjustment requirements to generate the corresponding component parameter adjustment data.

[0075] Furthermore, embodiments of this application use a performance prediction model to predict the performance of components during the aging process of the server based on component aging prediction data, and generate component parameter adjustment data based on the performance parameters of the aging components.

[0076] Therefore, by combining component aging parameters with a performance prediction model, this application embodiment can accurately predict the performance of server aging components at different time periods and generate parameter adjustment data, thereby delaying performance degradation and ensuring the stable operation of the server.

[0077] Optionally, in one embodiment of this application, the performance parameters of the aged components of the target server in different time periods are predicted based on the component aging parameters and a pre-built performance prediction model. This includes: collecting component performance data of multiple components and constructing a performance prediction training sample dataset based on historical aging data and component performance data; training the performance prediction model through the performance prediction training sample dataset and inputting the component aging parameters into the trained performance prediction model to predict the performance parameters of the aged components of the target server in different time periods.

[0078] In actual implementation, the embodiment of the application can first collect historical aging data (such as cumulative running time, environmental stress exposure record, aging degree rating) of multiple components such as server CPU, memory, hard disk and corresponding component performance data (such as CPU peak computing power, memory bandwidth, hard disk random read-write IOPS), and ensure that each set of data corresponds one by one (such as the aging data of a hard disk running for 2000 hours, matching the performance data of the read-write speed of 150MB / s in the same period), to build a performance prediction training sample data set.

[0079] After building the performance prediction training sample data set, the embodiment of the application can first preprocess the data, such as eliminating outliers (such as sudden increase and decrease of performance data caused by sensor failure), and inputting the data set into a basic model (such as linear regression, LSTM (‌Long Short Term Memory‌, long short-term memory network) neural network) for training.

[0080] In the model training process, the embodiment of the application can adjust the model parameters (such as the number of hidden layer nodes of LSTM, learning rate) by iteration, and if the performance requirement of error rate ≤5% is met, the final performance prediction model is obtained.

[0081] In online application, the embodiment of the application can input the component aging parameters (such as the aging state data of a CPU expected to run for 3000 hours) of the server into the trained performance prediction model, and the model will output the performance data (such as the CPU computing power expected to decrease by 3% after 1 month) of the component in different time periods (such as after 1 month, after 3 months) based on the aging-performance correlation law in the historical data.

[0082] Thus, the embodiment of the application can train the model based on real historical data to ensure that the model can accurately associate aging and performance changes, thereby improving the accuracy of component aging performance prediction and providing a reliable basis for server maintenance.

[0083] Optionally, in an embodiment of the present application, the performance prediction model is trained by the performance prediction training sample data set, and the component aging parameters are input into the trained performance prediction model to predict the aging component performance parameters of the target server in different time periods, comprising: dividing the performance prediction training sample data set into a training subset and a validation subset of the performance prediction model based on a preset data set division ratio; inputting the training subset into the performance prediction model to iteratively train the performance prediction model, and inputting the validation subset into the performance prediction model after the iterative training to output predicted performance validation data corresponding to the validation subset; comparing the predicted performance validation data and the component performance data in the validation subset to obtain a corresponding comparison result, and adjusting a plurality of model parameters of the performance prediction model after the iterative training according to the comparison result to generate a final trained performance prediction model; determining a prediction time period division rule of the target server, and performing a disassembly operation on the component aging parameters according to the prediction time period division rule to obtain component aging prediction sub-data corresponding to different time periods; inputting the component aging prediction sub-data into the final trained performance prediction model respectively to calculate performance prediction data corresponding to a plurality of constituent components in the target server in different time periods by the final trained performance prediction model, and integrating the performance prediction data corresponding to the plurality of constituent components in different time periods to generate the aging component performance parameters of the target server in different time periods.

[0084] Specifically, the embodiments of the present application can first divide the performance prediction training sample data set into a training subset (for model parameter learning) and a validation subset (for evaluating the generalization ability of the model) according to a data set division ratio of 7:3.

[0085] Secondly, the embodiments of the present application can input the training subset into an initial performance prediction model (such as an LSTM neural network) for iterative training. As a realizable way, the embodiments of the present application can set a training period of 50 rounds, calculate the prediction error after the end of each round, and dynamically adjust the learning rate (such as gradually reducing from 0.01 to 0.001). After the training is completed, the embodiments of the present application can input the validation subset into the model to obtain predicted performance validation data, and further optimize the model parameters (such as increasing the number of LSTM layer neurons and adjusting the regularization coefficient) according to the comparison result by comparing the data with the actual component performance data in the validation subset (such as calculating the mean square error), until the prediction error of the model on the validation set is less than 8%, thereby generating a final trained performance prediction model.

[0086] Thirdly, the embodiments of the present application can determine a prediction time period division rule (such as division according to 1 week, 1 month, and 3 months), and disassemble the component aging parameters into aging prediction sub-data corresponding to the time periods (such as the aging degree data of a certain hard disk after 3 months).

[0087] Afterwards, the embodiment of the present application can input each sub-data into the final model respectively to calculate the performance prediction data of each component such as CPU and memory in different time periods (such as a 2% decrease in memory bandwidth after 1 month), and finally integrate the performance prediction data to form the comprehensive aging component performance parameters of the target server in each time period.

[0088] Therefore, the embodiment of the present application improves the model prediction accuracy by scientifically dividing the data set and optimizing the parameters, so as to accurately output the performance data in each time period and provide accurate and reliable reference data for server maintenance.

[0089] Optionally, in an embodiment of the present application, the component parameter data of the plurality of component parts is acquired, and the parameter adjustment requirement of the corresponding component part is determined according to the component parameter data and the aging component performance parameter, and the component parameter data is adjusted through the parameter adjustment requirement to generate corresponding component parameter adjustment data, including: collecting the component parameter data corresponding to the plurality of component parts, and matching the component parameter data of the plurality of component parts and the aging component performance parameter to obtain a matching data pair corresponding to different component parts; determining the parameter adjustment requirement of the corresponding component part according to the aging component performance parameter in the matching data pair, so as to adjust the component parameter data based on the parameter adjustment requirement to generate corresponding component parameter adjustment data.

[0090] It should be noted that the embodiment of the present application can first collect the current component parameter data of the plurality of component parts such as CPU, memory and hard disk in the target server, such as CPU frequency setting, memory timing parameter, hard disk cache size, etc., and match the current component parameter data with the previously predicted component aging performance data (such as CPU computing power attenuation rate, memory read-write speed decrease value) one by one to form a component parameter-aging performance matching data pair (for example, "CPU frequency 3.2GHz-computing power attenuation 5%", "memory timing CL20-read-write speed decrease 8%").

[0091] Secondly, the embodiment of the present application can determine the parameter adjustment requirement of each component part according to the aging performance data in the matching data pair. If the performance of a certain component part is not up to the business requirement due to aging (such as the read-write speed of the hard disk is reduced to 100MB / s due to aging, which is lower than the business requirement of 120MB / s), the requirement of compensating for the performance gap through parameter adjustment is explicitly required (such as increasing the hard disk cache from 128MB to 256MB to improve the read-write speed).

[0092] Thirdly, the embodiment of the present application can adjust the component parameter data according to the adjustment requirement, for example, for the case of CPU computing power attenuation of 5%, the frequency is fine-tuned from 3.2GHz to 3.3GHz within the hardware support range; for the memory read-write speed decrease, the memory timing parameter is optimized.

[0093] After the adjustment is completed, the embodiment of the present application can generate component parameter adjustment data containing component name-original parameter-adjusted parameter-expected performance improvement value to ensure that the adjusted parameter can match the component performance with the aging state and meet the business operation requirements.

[0094] Therefore, the embodiment of the present application determines the adjustment direction by matching the parameter and the aging performance, thereby accurately generating the parameter adjustment data, effectively compensating for the performance loss caused by aging, and ensuring that the server component performance meets the business requirements.

[0095] In step S104, the aging stability test scheme corresponding to the target server is generated through the component parameter adjustment data, and the aging stability test of the target server is performed according to the aging stability test scheme to obtain the aging stability result of the target server.

[0096] After that, the embodiment of the present application can construct a server aging stability test scheme according to the generated component parameter adjustment data, so as to perform the aging stability test on the server by using the constructed server aging stability test scheme.

[0097] Therefore, the embodiment of the present application performs targeted aging stability test, thereby accurately verifying whether the parameter adjustment data can adapt to the component aging state, ensuring the stability of the adjusted server performance, and reducing the risk of failure.

[0098] Optionally, in an embodiment of the present application, the aging stability test scheme corresponding to the target server is generated through the component parameter adjustment data, including: determining the test dimensions of the aging stability test corresponding to the target server, and determining the test parameter range under different test dimensions according to the component parameter adjustment data and the test dimensions; based on the test parameter range, establishing the test execution process and the test result collection rule of the aging stability test, and integrating the test execution process and the test result collection rule to generate the aging stability test scheme.

[0099] Specifically, the embodiment of the present application can first determine the core test dimensions of the aging stability test in combination with the application scenarios of the server (such as enterprise-level data storage and high-performance computing) and the component parameter adjustment focus (such as CPU frequency improvement and hard disk cache expansion), which usually include three categories of performance compliance dimensions (verifying whether the performance of the adjusted component meets the business requirements), long-term stability dimensions (testing the stability of the component after parameter adjustment for continuous operation), and environment adaptation dimensions (simulating the adaptability of parameter adjustment under different environmental stresses).

[0100] Secondly, the embodiment of the present application can determine the test parameter range in detail according to the component parameter adjustment data (such as adjusting the original CPU frequency of 3.2 GHz to 3.3 GHz and adjusting the original hard disk cache of 128 MB to 256 MB) and each test dimension, as follows:

[0101] 1. Performance compliance dimension: For CPU, set the test parameter range to "maintain 3.3 GHz, and the computing power should be ≥95% of the original (offset 5% aging loss)"; for hard disk, set "cache 256 MB, read and write speed ≥120 MB / s (to meet business requirements)";

[0102] 2. Long-term stability dimension: Set the test parameter range to "72 hours of continuous high load (CPU usage rate >80%, hard disk IO load >70%) running, no blue screen, restart or performance drop (single drop ≤3%)";

[0103] 3. Environmental adaptation dimension: Set the temperature fluctuation range (20-40℃) and voltage fluctuation range (±5%), and test the parameters "under this environment, the component parameters can still be stable, and the performance deviation ≤5%".

[0104] After that, the embodiment of the present application can establish a test execution process based on the determined test parameter range, as follows:

[0105] First, configure the target server parameters according to the adjustment data, and preheat for 30 minutes;

[0106] Second, perform the test in the order of performance compliance dimension-long-term stability dimension-environmental adaptation dimension, and record the initial performance benchmark value before each dimension test;

[0107] Third, record the performance data every 1 hour in the long-term stability test, and adjust the temperature every 5℃ in the environmental adaptation test and record the parameter stability; at the same time, formulate the test result collection rules: clear the collection tools (such as CPU-Z to monitor computing power, CrystalDiskMark to test hard disk speed), data recording frequency (performance data every 15 minutes, stability log real-time collection), and abnormality judgment standard (such as performance lower than the lower limit of the parameter range is judged as abnormal).

[0108] Finally, the embodiment of the present application can structure and integrate the test execution process (including steps, order, and duration) and the test result collection rules (including tools, frequency, and standards), supplement the server state check list before the test (such as component connection, system driver version) and the abnormality handling plan after the test (such as reverting the parameters and analyzing the reasons when the performance is not up to standard), and form a complete aging stability test scheme.

[0109] Thus, the embodiment of the present application constructs a targeted test scheme around parameter adjustment, clearly defines the test dimensions, parameters, and processes, thereby ensuring the effectiveness of comprehensive verification of parameter adjustment and guaranteeing the stable operation of the server under the aging state.

[0110] Optionally, in an embodiment of the present application, after performing the aging stability test on the target server according to the aging stability test scheme to obtain the aging stability result of the target server, the method further comprises: performing aging stability analysis on the aging stability result to obtain aging stability analysis data corresponding to the target server, and determining whether the target server meets a preset aging stability requirement according to the aging stability analysis data; if the target server meets the aging stability requirement, determining that the target server is aging stability performance qualified; if the target server does not meet the aging stability requirement, determining that the target server is aging stability performance unqualified, and generating a corresponding aging stability performance test report, and sending the aging stability performance test report to the user end, so that the target user optimizes and adjusts the target server according to the aging stability performance test report.

[0111] In the specific implementation process, after completing the aging stability test of the target server according to the aging stability test scheme, the embodiment of the present application can first perform a test data collection operation. For example, the embodiment of the present application can collect performance data (such as CPU computing power fluctuation value, memory read / write speed change, hard disk IOPS attenuation), running state data (component temperature, power consumption, error code record), and associated data of different aging stages (such as 24 hours, 48 hours, 72 hours corresponding to the simulation aging degree) in real time for CPU, memory, hard disk and other core components, to ensure that the data covers the test full cycle and multiple component dimensions.

[0112] Secondly, the embodiment of the present application can integrate the collected unstructured data (such as log text) and structured data (such as performance numerical value), and classify and archive them according to the dimension of “test time period-component type-data category”, to build a standardized test data set containing original data and data annotation (such as “24-hour aging stage-CPU-computing power data”).

[0113] Thirdly, the embodiment of the present application can perform aging stability analysis on the test data set. Specifically, the embodiment of the present application performs corresponding analysis from two dimensions, as follows:

[0114] 1. Component aging degree and stability correlation analysis:

[0115] Compare the performance attenuation amplitude and failure rate of each component under different aging stages (such as simulated 1-year, 3-year, 5-year aging), to determine whether there is an abnormal situation of “stability sudden drop after aging degree improvement” (such as a sharp increase in hard disk error codes after 3 years of aging);

[0116] 2. Time dimension stability trend analysis:

[0117] The overall running state of the server in a test period (such as 72 hours) is combed to check whether there are problems such as periodic performance fluctuations, sudden outages, or parameter drifts, and the stable performance of the server at different time periods is determined.

[0118] Further, the embodiment of the application can determine according to the analysis result and the preset aging stability requirement (such as “performance attenuation rate of each component ≤8%”, “no outage in the test period”, “error code occurrence frequency ≤3 times”); if all indicators meet the requirements, it is directly determined that the aging stability performance of the target server is qualified; if there is an indicator that does not meet the requirements (such as CPU aging for 3 years and power attenuation of 10%), it is determined as unqualified. At this time, the aging stability performance test report is generated, and the unqualified indicators, the corresponding aging stage, and the possible impact (such as “power attenuation exceeding the threshold value may cause business lag”) are marked in detail in the report, and are pushed to the user end of the operation and maintenance personnel for targeted optimization and adjustment (such as re-adjusting the CPU parameters or replacing the components with high aging risk).

[0119] Therefore, the embodiment of the application can accurately determine the server aging stability through multi-dimensional collection and test data analysis operation, and effectively improve the maintenance efficiency and stability of the server.

[0120] Optionally, in an embodiment of the application, the target user optimizes and adjusts the target server according to the aging stability performance test report, including: analyzing the aging stability performance test report to obtain corresponding test report analysis data, and determining a to-be-optimized component in the target server according to the test report analysis data; extracting unqualified data detail data, abnormal scene description data, and preliminary optimization guide data in the test report analysis data, and obtaining actual running load characteristics of the target server to determine a to-be-improved performance dimension of the to-be-optimized component according to the actual running load characteristics; generating an optimization adjustment scheme of the to-be-optimized component based on the unqualified data detail data, the abnormal scene description data, the preliminary optimization guide data, and the to-be-improved performance dimension, so that the target user performs component maintenance on the target server according to the optimization adjustment scheme; re-testing the aging stability of the target server after the component maintenance to obtain corresponding review test data, and determining whether the review test data meets the preset aging stability requirement standard, wherein, in the case that the review test data meets the aging stability requirement standard, it is determined that the aging stability performance of the target server is qualified.

[0121] It should be noted that, in the embodiment of the application, the process of optimizing and adjusting the server by the user according to the aging stability performance test report is as follows:

[0122] 1. Analyzing the aging stability performance test report, extracting test report analysis data, and determining a to-be-optimized component (such as a hard disk that appears performance unqualified in the report multiple times);

[0123] 2. Extract the substandard data details (such as the minimum hard disk read-write speed of 90MB / s, lower than the standard of 120MB / s), abnormal scene description (hard disk response delay exceeds 50ms under high load), preliminary optimization guide (suggesting to expand the cache), and the performance dimensions to be improved (hard disk read-write speed and response time under high load) of the components to be optimized according to the actual running load characteristics of the server (such as the read-write peak at 10-12 o'clock every day);

[0124] 3. Based on the substandard data details, abnormal scene description data, preliminary optimization guide data, and performance dimensions to be improved, generate an optimization adjustment scheme, for example, expand the cache of the hard disk to 512MB, and adjust the read-write scheduling algorithm to prioritize tasks during peak hours; the target user completes the component maintenance according to the optimization adjustment scheme, and then re-performs the aging stability test to collect review test data; if the data shows that the hard disk read-write speed under high load reaches 130MB / s and the response delay is reduced to 20ms, meeting the preset standard, it is determined to be qualified.

[0125] Therefore, the embodiments of the present application can optimize and review the aging stability test to accurately solve the problem of server aging stability, thereby ensuring that the performance meets the standard and improving the operation reliability and maintenance efficiency.

[0126] The execution logic of the server aging stability test method, the server aging stability test process, and the server component aging prediction process of the present application are described below by combining the accompanying drawings.

[0127] Figure 2 The execution logic of the server aging stability test method is shown in FIG. 1. As shown in FIG. 1, the execution process of the server aging stability test method of the present application is described as follows: Figure 2

[0128] S201: Obtain the server model, and obtain the component model parameters of the server through the server model;

[0129] S202: Perform aging prediction on the components of the server through the aging prediction model;

[0130] S203: Based on the aging prediction data of the components, predict the performance of the components at the aging time of the server through the performance prediction model to obtain the aging component performance data;

[0131] S204: Generate component parameter adjustment data according to the aging component performance data;

[0132] S205: Construct a server aging stability test scheme according to the generated component parameter adjustment data;

[0133] ​S206: The server is tested for aging stability using the server aging stability test scheme.

[0134] Figure 3 The execution logic diagram for testing the server for aging stability based on the aging stability test scheme is shown. As shown in Figure 3 the execution process of the server aging stability test based on the aging stability test scheme of the present application is as follows:

[0135] S301: The server is tested for aging stability using the constructed server aging stability test scheme;

[0136] S302: The data generated during the server test is collected, and after the test is completed, the corresponding test data set is generated;

[0137] S303: The test data set is analyzed, and the server aging stability analysis result is generated and output.

[0138] Figure 4 The execution logic diagram for server component aging prediction is shown. As shown in Figure 4 the execution process of the server component aging prediction of the present application is as follows:

[0139] S401: The components of the server are predicted for aging through the aging prediction model;

[0140] S402: Historical server component aging data is collected to construct an aging prediction training sample set;

[0141] S403: The aging prediction model is trained through the aging prediction training sample set.

[0142] It can be understood that before testing the server, the embodiments of the present application can first select the model of the server, and obtain the model parameters of each component of the server through the server model, and set the server running environment and the server purpose, and then predict the aging of each component of the server to obtain the aging degree of each component of the server in each time period; secondly, the embodiments of the present application can obtain the aging degree data of each component in each time period, and generate parameter adjustment data (such as the heat dissipation power of the heat sink, the read speed, the write speed, etc.) of the corresponding components at the corresponding time points according to the performance of the components in each time period after aging through the performance prediction model, so as to construct a server aging stability test scheme according to the generated parameter adjustment data and the running time period; then, the embodiments of the present application can use the constructed server aging stability test scheme to test the aging stability of the server, collect the data generated during the server test, and generate a test data set after the test is completed, so as to analyze the test data set, obtain the server aging stability analysis result, and output the analysis result, so that the staff can view the server aging stability test result.

[0143] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be realized by means of software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment.

[0144] The embodiments of the present application also provide a server aging stability test device.

[0145] As shown in Figure 5 The server aging stability test device 10 includes an aging prediction model training module 100, a component aging prediction module 200, a component parameter adjustment module 300, and a stability test module 400.

[0146] The aging prediction model training module 100 is configured to collect historical aging data of a plurality of components of a target server, and determine whether the historical aging data meets a preset quantity requirement, and in the case that the historical aging data does not meet the quantity requirement, perform an aging simulation experiment on the corresponding components to generate experimental aging data, and train a pre-constructed aging prediction model through the experimental aging data and the historical aging data.

[0147] The component aging prediction module 200 is configured to determine the component model parameters corresponding to the target server based on the server identifier of the target server, and input the component model parameters into the trained aging prediction model to output the component aging parameters corresponding to the target server.

[0148] The component parameter adjustment module 300 is configured to predict the performance parameters of the aging components of the target server in different time periods based on the component aging parameters and the pre-constructed performance prediction model, acquire the component parameter data of the plurality of constituent components, determine the parameter adjustment requirements of the corresponding constituent components according to the component parameter data and the aging component performance parameters, and adjust the component parameter data through the parameter adjustment requirements to generate corresponding component parameter adjustment data.

[0149] The stability test module 400 is configured to generate an aging stability test scheme corresponding to the target server through the component parameter adjustment data, and perform an aging stability test on the target server according to the aging stability test scheme to obtain an aging stability result of the target server.

[0150] Optionally, in an embodiment of the present application, the aging prediction model training module 100 comprises a judgment unit, a first sample set construction unit and a second sample set construction unit.

[0151] The judgment unit is configured to acquire historical aging data of the plurality of constituent components of the target server before inputting the model parameters of the constituent components into the trained aging prediction model to output the component aging parameters corresponding to the target server, and determine the data amount corresponding to the historical aging data, and judge whether the data amount of the historical aging data meets the preset quantity requirement.

[0152] The first sample set construction unit is configured to construct a corresponding aging prediction training sample set through the historical aging data when the data amount of the historical aging data meets the quantity requirement, so as to train the aging prediction model based on the aging prediction training sample set.

[0153] The second sample set construction unit is configured to determine the target constituent component that does not meet the quantity requirement when the data amount of the historical aging data does not meet the quantity requirement, perform an aging simulation experiment on the target constituent component, record the experimental equipment data corresponding to the target constituent component, generate experimental aging data of the target constituent component according to the experimental equipment data, construct a corresponding aging prediction training sample set based on the experimental aging data and the historical aging data, and train the aging prediction model based on the aging prediction training sample set.

[0154] Optionally, in an embodiment of the present application, the component parameter adjustment module 300 comprises a collection unit and a prediction unit.

[0155] The collection unit is configured to collect the component performance data of the plurality of constituent components, and construct a performance prediction training sample data set according to the historical aging data and the component performance data.

[0156] The prediction unit is configured to train a performance prediction model by using the performance prediction training sample data set, and input the component aging parameter into the trained performance prediction model to predict the aging component performance parameter of the target server in different time periods.

[0157] Optionally, in an embodiment of the present application, the component parameter adjustment module 300 further comprises a matching unit and an adjustment unit.

[0158] The matching unit is configured to collect component parameter data corresponding to a plurality of constituent components, and match the component parameter data of the plurality of constituent components with the aging component performance parameter to obtain a matching data pair corresponding to different constituent components.

[0159] The adjustment unit is configured to determine a parameter adjustment requirement of a corresponding constituent component according to the aging component performance parameter in the matching data pair, and adjust the component parameter data based on the parameter adjustment requirement to generate corresponding component parameter adjustment data.

[0160] Optionally, in an embodiment of the present application, the stability test module 400 comprises a determination unit and an integration unit.

[0161] The determination unit is configured to determine a test dimension of the aging stability test corresponding to the target server, and determine a test parameter range under different test dimensions according to the component parameter adjustment data and the test dimension.

[0162] The integration unit is configured to establish a test execution process and a test result collection rule of the aging stability test based on the test parameter range, and integrate the test execution process and the test result collection rule to generate an aging stability test scheme.

[0163] Optionally, in an embodiment of the present application, the server aging stability test device 10 further comprises an analysis module, a judgment module and an optimization module.

[0164] The analysis module is configured to perform aging stability analysis on the aging stability result after performing the aging stability test on the target server according to the aging stability test scheme to obtain the aging stability result of the target server, to obtain aging stability analysis data corresponding to the target server, and to judge whether the target server meets a preset aging stability requirement according to the aging stability analysis data.

[0165] The judgment module is configured to judge that the target server is aging stability qualified if the target server meets the aging stability requirement.

[0166] The optimization module is configured to determine that the target server fails to meet the aging stability requirement if the target server does not meet the aging stability requirement, and generate an aging stability performance test report. The aging stability performance test report is sent to the user terminal, so that the target user optimizes and adjusts the target server according to the aging stability performance test report.

[0167] Optionally, in an embodiment of the present application, the server aging stability test device 10 further comprises an extraction module, a screening module, a test module and a selection module.

[0168] The extraction module is configured to extract data features corresponding to the aging prediction training sample set before training the aging prediction model pre-constructed by the experimental aging data and the historical aging data, wherein the data features include data dimension type, data time sequence correlation and data feature complexity.

[0169] The screening module is configured to determine the convolutional neural network structure requirement according to the data features, and screen a plurality of candidate convolutional neural network structures meeting the convolutional neural network structure requirement from a pre-set convolutional neural network structure library.

[0170] The test module is configured to input the aging prediction training sample set into the plurality of candidate convolutional neural network structures respectively, so as to train and test the performance of the plurality of candidate convolutional neural network structures, and generate a corresponding preliminary performance test result.

[0171] The selection module is configured to select a target candidate convolutional neural network structure meeting a pre-set performance requirement based on the preliminary performance test result, so as to construct the aging prediction model by using the target candidate convolutional neural network structure.

[0172] Optionally, in an embodiment of the present application, the judgment unit comprises an aging data acquisition subunit, a preprocessing subunit and a verification subunit.

[0173] The aging data acquisition subunit is configured to acquire original historical aging data of a plurality of component parts in the target server, wherein the original historical aging data includes part runtime length data, performance attenuation data, environmental influence data and fault record data.

[0174] The preprocessing subunit is configured to perform a data preprocessing operation on the original historical aging data to obtain corresponding standard historical aging data.

[0175] The verification subunit is configured to perform reliability verification on the standard historical aging data based on a pre-set data reliability verification rule to obtain a corresponding reliability verification result, and screen historical aging data meeting a pre-set reliability requirement from the standard historical aging data according to the reliability verification result.

[0176] Optionally, in an embodiment of the present application, the prediction unit comprises a division subunit, an iteration subunit, a comparison subunit, a disassembly subunit and a generation subunit.

[0177] The division subunit is configured to divide the performance prediction training sample data set into a training subset and a verification subset of the performance prediction model based on a preset data set division ratio.

[0178] The iteration subunit is configured to input the training subset into the performance prediction model to perform iterative training of the performance prediction model, and input the verification subset into the performance prediction model after the iterative training to output predicted performance verification data corresponding to the verification subset.

[0179] The comparison subunit is configured to compare the predicted performance verification data and the component performance data in the verification subset to obtain a corresponding comparison result, and adjust a plurality of model parameters of the performance prediction model after the iterative training according to the comparison result to generate a final trained performance prediction model.

[0180] The disassembly subunit is configured to determine a prediction time period division rule of the target server, and perform a disassembly operation on the component aging parameters according to the prediction time period division rule to obtain component aging prediction sub-data corresponding to different time periods.

[0181] The generation subunit is configured to input the component aging prediction sub-data into the final trained performance prediction model respectively to calculate performance prediction data corresponding to a plurality of constituent components in the target server in different time periods through the final trained performance prediction model, and integrate the performance prediction data corresponding to the plurality of constituent components in different time periods to generate aging component performance parameters of the target server in different time periods.

[0182] Optionally, in an embodiment of the present application, the second sample set construction unit comprises an experimental parameter determination subunit, a simulation subunit and a data cleaning subunit.

[0183] The experimental parameter determination subunit is configured to determine actual operating environment parameters and aging influencing factors of the target constituent component, and determine a plurality of experimental parameters of an aging simulation simulation experiment according to the actual operating environment parameters and the aging influencing factors, wherein the plurality of experimental parameters comprise environmental stress parameters, operating load parameters and experimental duration parameters.

[0184] The simulation subunit is configured to build an aging simulation simulation experiment environment of the target constituent component based on the plurality of experimental parameters, and start the aging simulation simulation experiment in the aging simulation simulation experiment environment, and record current device data of the target constituent component in the aging simulation simulation experiment process, wherein the current device data comprises component operating state data, performance change data and stress response data.

[0185] The data cleaning subunit is configured to perform a data cleaning operation on the current device data to obtain corresponding current standard device data, and extract performance degradation trend features and aging state features in the current standard device data, so as to generate experimental aging data of the target component according to the performance degradation trend features and the aging state features.

[0186] Optionally, in an embodiment of the present application, the optimization module comprises an analysis unit, an acquisition unit, a maintenance unit and a review unit.

[0187] The analysis unit is configured to analyze the aging stability performance test report to obtain corresponding test report analysis data, and determine the to-be-optimized component in the target server according to the test report analysis data.

[0188] The acquisition unit is configured to extract non-compliance data details, abnormal scenario description data and preliminary optimization guidance data in the test report analysis data, and acquire actual running load features of the target server, so as to determine a to-be-improved performance dimension of the to-be-optimized component according to the actual running load features.

[0189] The maintenance unit is configured to generate an optimization adjustment scheme of the to-be-optimized component based on the non-compliance data details, the abnormal scenario description data, the preliminary optimization guidance data and the to-be-improved performance dimension, so that the target user performs component maintenance on the target server according to the optimization adjustment scheme.

[0190] The review unit is configured to re-perform the aging stability test on the target server after the component maintenance to obtain corresponding review test data, and determine whether the review test data meets a preset aging stability requirement standard, wherein, in a case where the review test data meets the aging stability requirement standard, it is determined that the target server is aging stable and performs well.

[0191] Optionally, in an embodiment of the present application, the component aging prediction module 200 comprises a reading unit configured to read a server model of the target server, and input the server model into a preset association database to extract component model parameters corresponding to a plurality of component parts in the target server.

[0192] The features of the embodiment of the server aging stability test device can be referred to the related description of the embodiment of the server aging stability test method, which will not be repeated here.

[0193] Embodiments of the present application also provide an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above-mentioned server aging stability test method embodiments.

[0194] The embodiment of the present application further provides a non-volatile computer readable storage medium, which stores a computer program, wherein the computer program is arranged to execute the steps in any of the above-mentioned server aging stability test method embodiments when running.

[0195] In an example embodiment, the above-mentioned non-volatile computer readable storage medium can include, but is not limited to, a U disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store computer programs.

[0196] The embodiment of the present application further provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the steps in any of the above-mentioned server aging stability test method embodiments.

[0197] The embodiment of the present application further provides another computer program product, which comprises a non-volatile computer readable storage medium, and the non-volatile computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps in any of the above-mentioned server aging stability test method embodiments.

[0198] The skilled person can further realize that the units and algorithm steps of the examples described in connection with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in general terms in the above description. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0199] The above has introduced in detail a server aging stability test method, device, equipment and medium provided by the present application. The principles and implementation manners of the present application have been described by applying specific examples in this paper, and the above example description is only used to help understand the method and core idea of the present application. It should be pointed out that for ordinary skilled person in the technical field, some improvements and modifications can be made to the present application without departing from the principles of the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims

1. A method of server burn-in stability testing, the method comprising: The method comprises the following steps: Collecting historical aging data of a plurality of component parts in a target server, and determining whether the historical aging data meets a preset quantity requirement; and in the case that the historical aging data does not meet the quantity requirement, performing an aging simulation experiment on the corresponding component parts to generate experimental aging data, and training a pre-constructed aging prediction model by using the experimental aging data and the historical aging data; Determining a component part model parameter corresponding to the target server based on a server identifier of the target server, and inputting the component part model parameter into the trained aging prediction model to output a component part aging parameter corresponding to the target server; Based on the component part aging parameter and a pre-constructed performance prediction model, predicting aging component part performance parameters of the target server in different time periods, obtaining component part parameter data of the plurality of component parts, and determining parameter adjustment requirements of the corresponding component parts according to the component part parameter data and the aging component part performance parameters, and adjusting the component part parameter data according to the parameter adjustment requirements to generate corresponding component part parameter adjustment data; Generating an aging stability test scheme corresponding to the target server by using the component part parameter adjustment data, and performing an aging stability test on the target server according to the aging stability test scheme to obtain an aging stability result of the target server; Before training the pre-constructed aging prediction model by using the experimental aging data and the historical aging data, the method further comprises the following steps: Extracting data features corresponding to an aging prediction training sample set, wherein the data features include data dimension types, data time sequence correlation, and data feature complexity; Determining a convolutional neural network structure requirement corresponding to the data features, and screening a plurality of candidate convolutional neural network structures meeting the convolutional neural network structure requirement from a pre-set convolutional neural network structure library; Inputting the aging prediction training sample set into the plurality of candidate convolutional neural network structures respectively to train and test the performance of the plurality of candidate convolutional neural network structures, and generating corresponding preliminary performance test results; Based on the preliminary performance test results, selecting a target candidate convolutional neural network structure meeting a pre-set performance requirement to construct the aging prediction model by using the target candidate convolutional neural network structure; The method of determining a convolutional neural network structure requirement corresponding to the data features, and screening a plurality of candidate convolutional neural network structures meeting the convolutional neural network structure requirement from a pre-set convolutional neural network structure library comprises the following steps: Obtaining time sequence data of the plurality of component parts in the target server, and calculating mutual information entropy values between different component parts according to the time sequence data, so as to divide the plurality of component parts into corresponding strongly correlated groups, weakly correlated groups, and uncorrelated groups based on the mutual information entropy values, a first pre-set entropy value threshold, and a second pre-set entropy value threshold, wherein the first pre-set entropy value threshold is greater than the second pre-set entropy value threshold; The strong correlation group shares a preset feature extraction backbone and is connected in series with a preset timing attention module, the weak correlation group adopts independent branches and interacts through a preset jump connection strategy, and the irrelevant group is configured with completely independent branches, so as to construct a basic branch structure in the target candidate convolutional neural network structure based on the strong correlation group, the weak correlation group and the irrelevant group; A preset adaptive fusion unit is embedded in the basic branch structure, and a current component correlation degree between the plurality of component parts is calculated every preset period, and the current component correlation degree is converted into a corresponding normalized correlation weight value through the adaptive fusion unit, and the normalized correlation weight value is used to adjust the cross-fusion weight of the feature extraction backbone, the weak correlation group independent branch and the irrelevant group independent branch in the basic branch structure, so as to determine the plurality of candidate convolutional neural network structures.

2. The server burn-in stability test method of claim 1, wherein, The history aging data of the plurality of component parts in the target server is collected, and it is judged whether the history aging data meets a preset quantity requirement, and in the case that the history aging data does not meet the quantity requirement, an aging simulation experiment is performed on the corresponding component part to generate experimental aging data, and the aging prediction model is trained through the experimental aging data and the history aging data, including: The history aging data of the plurality of component parts in the target server is collected, and the data amount corresponding to the history aging data is determined, and it is judged whether the data amount of the history aging data meets a preset quantity requirement; When the data amount of the history aging data meets the quantity requirement, a corresponding aging prediction training sample set is constructed through the history aging data, and the aging prediction model is trained based on the aging prediction training sample set; When the data amount of the history aging data does not meet the quantity requirement, the target component part that does not meet the quantity requirement is determined, and an aging simulation experiment is performed on the target component part, and experimental equipment data corresponding to the target component part is recorded, and experimental aging data of the target component part is generated according to the experimental equipment data, and a corresponding aging prediction training sample set is constructed based on the experimental aging data and the history aging data, and the aging prediction model is trained based on the aging prediction training sample set.

3. The server aging stability test method of claim 2, wherein, The aging component performance parameters of the target server in different time periods are predicted based on the component aging parameters and a pre-constructed performance prediction model, including: The component performance data of the plurality of component parts is collected, and a performance prediction training sample data set is constructed according to the history aging data and the component performance data; The performance prediction model is trained through the performance prediction training sample data set, and the component aging parameters are input into the trained performance prediction model to predict the aging component performance parameters of the target server in different time periods.

4. The server burn-in stability test method of claim 1, wherein, The component parameter data of the plurality of constituent components is acquired, and parameter adjustment requirements of the corresponding constituent components are determined according to the component parameter data and the aging component performance parameter, and the component parameter data is adjusted through the parameter adjustment requirements to generate corresponding component parameter adjustment data, including: The component parameter data corresponding to the plurality of constituent components is collected, and the component parameter data of the plurality of constituent components and the aging component performance parameter are matched to obtain a matching data pair corresponding to different constituent components; According to the aging component performance parameter in the matching data pair, the parameter adjustment requirements of the corresponding constituent components are determined, and the component parameter data is adjusted based on the parameter adjustment requirements to generate corresponding component parameter adjustment data.

5. The server burn-in stability test method of claim 1, wherein, The aging stability test scheme corresponding to the target server is generated through the component parameter adjustment data, including: The test dimensions of the aging stability test corresponding to the target server are determined, and the test parameter range under different test dimensions is determined according to the component parameter adjustment data and the test dimensions; Based on the test parameter range, the test execution process and the test result collection rule of the aging stability test are established, and the test execution process and the test result collection rule are integrated to generate the aging stability test scheme.

6. The server burn-in stability test method of claim 1, wherein, After performing the aging stability test on the target server according to the aging stability test scheme to obtain the aging stability result of the target server, it further includes: The aging stability analysis is performed on the aging stability result to obtain the aging stability analysis data corresponding to the target server, and it is judged whether the target server meets the preset aging stability requirement according to the aging stability analysis data; If the target server meets the aging stability requirement, it is determined that the aging stability performance of the target server is qualified; If the target server does not meet the aging stability requirement, it is determined that the aging stability performance of the target server is unqualified, and a corresponding aging stability performance test report is generated, and the aging stability performance test report is sent to the user end to enable the target user to optimize and adjust the target server according to the aging stability performance test report.

7. The server burn-in stability test method of claim 2, wherein, The historical aging data of the plurality of constituent components in the target server is acquired, including: The original historical aging data of the plurality of constituent components in the target server is collected, wherein the original historical aging data includes component running time data, performance degradation data, environmental influence data and fault record data; The original historical aging data is subjected to a data preprocessing operation to obtain corresponding standard historical aging data; Based on a preset data reliability verification rule, the standard historical aging data is subjected to reliability verification to obtain a corresponding reliability verification result, and historical aging data meeting a preset reliability requirement is selected from the standard historical aging data according to the reliability verification result.

8. The server burn-in stability test method of claim 3, wherein, The performance prediction model is trained through the performance prediction training sample data set, and the component aging parameter is input into the trained performance prediction model to predict the aging component performance parameter of the target server in different time periods, including: The performance prediction training sample data set is divided into a training subset and a verification subset of the performance prediction model based on a preset data set division ratio; The training subset is input into the performance prediction model for iterative training, and the verification subset is input into the performance prediction model after iterative training to output predicted performance verification data corresponding to the verification subset; The predicted performance verification data and the component performance data in the verification subset are compared to obtain a corresponding comparison result, and the model parameters of the performance prediction model after iterative training are adjusted according to the comparison result to generate a final trained performance prediction model; The prediction time period division rule of the target server is determined, and the component aging parameter is decomposed according to the prediction time period division rule to obtain component aging prediction sub-data corresponding to different time periods; The component aging prediction sub-data is input into the final trained performance prediction model respectively to calculate performance prediction data corresponding to a plurality of component parts in the target server in different time periods through the final trained performance prediction model, and the performance prediction data corresponding to the plurality of component parts in different time periods is integrated to generate aging component performance parameters of the target server in different time periods.

9. The server burn-in stability test method of claim 2, wherein, When the data amount of the historical aging data does not meet the quantity requirement, a target component part that does not meet the quantity requirement is determined, and an aging simulation experiment is performed on the target component part, and experimental equipment data corresponding to the target component part is recorded, including: The actual operating environment parameters and aging influencing factors of the target component part are determined to determine a plurality of experimental parameters of the aging simulation experiment according to the actual operating environment parameters and the aging influencing factors, wherein the plurality of experimental parameters include environmental stress parameters, operating load parameters and experimental duration parameters; Based on the plurality of experimental parameters, an aging simulation experiment environment of the target component part is built, and the aging simulation experiment is started in the aging simulation experiment environment, and current equipment data of the target component part in the aging simulation experiment process is recorded, wherein the current equipment data includes component operating state data, performance change data and stress response data; The current equipment data is subjected to data cleaning operation to obtain corresponding current standard equipment data, and the performance attenuation trend feature and the aging state feature in the current standard equipment data are extracted to generate experimental aging data of the target component part according to the performance attenuation trend feature and the aging state feature.

10. The server aging stability test method of claim 6, wherein, The target user optimizes and adjusts the target server according to the aging stable performance test report, including: parsing the aging stability performance test report to obtain corresponding test report parsing data, and determining a to-be-optimized component in the target server according to the test report parsing data; extracting non-compliance data details, abnormal scenario description data, and preliminary optimization guidance data in the test report parsing data, and obtaining actual running load characteristics of the target server to determine a to-be-improved performance dimension of the to-be-optimized component according to the actual running load characteristics; generating an optimization adjustment scheme of the to-be-optimized component based on the non-compliance data details, the abnormal scenario description data, the preliminary optimization guidance data, and the to-be-improved performance dimension, so that the target user performs component maintenance on the target server according to the optimization adjustment scheme; re-performing aging stability testing on the target server after component maintenance to obtain corresponding review test data, and determining whether the review test data meets a preset aging stability requirement standard, wherein, in a case where the review test data meets the aging stability requirement standard, it is determined that the target server is aging stability performance qualified.

11. The server burn-in stability testing method of claim 1, wherein, The server identification of the target server is used to determine the corresponding component model parameters of the target server, and the component model parameters include: reading the server model of the target server, and inputting the server model into a preset association database to extract the component model parameters corresponding to a plurality of components in the target server.

12. An electronic device, comprising: including: a memory for storing a computer program; a processor for executing the computer program to implement the steps of the server aging stability test method according to any one of claims 1 to 11.

Images