Diagnosis method and diagnosis device for memory fault and electronic equipment

Abstract

The embodiment of the invention provides a memory fault diagnosis method and device and electronic equipment. The method comprises the steps of obtaining index data with health indexes of normal memory performance index data larger than a health threshold value in multiple memory performance index data of a server; the unsupervised learning model is trained according to the unsupervised learning model and the memory performance index data sample, a fault diagnosis model is obtained, and a time sequence analysis algorithm is integrated in the unsupervised learning model; according to the multiple pieces of memory performance index data and a fault diagnosis model, determining a first average absolute error of the multiple pieces of memory performance index data; and judging whether the first mean absolute error is greater than an error threshold value or not, and sending first early warning information of the server fault when the mean absolute error is greater than the error threshold value. The problems that the built memory fault prediction model is inaccurate and the robustness is poor due to scarcity of fault memory samples in related technologies are solved.

Description

Technical Field

[0001] Embodiments of the present application relate to the technical field of server fault diagnosis, and in particular, to a method for diagnosing a memory fault, a device for diagnosing a memory fault, a computer-readable storage medium, a computer program product, and an electronic device. Background Art

[0002] In the existing technology, machine learning and deep learning technologies are beginning to be applied to build memory fault prediction models, which will have the following problems, such as the scarcity of faulty memory samples. Since the number of faulty memories is far less than that of normal memories, when using supervised learning methods to classify faulty and normal memories, there will be serious class imbalance problems. Although there are some methods to deal with class imbalance, they may lead to high false alarm rates, thereby increasing operation and maintenance costs; the robustness of the fault warning model is poor, the health status of memory samples will decrease as the usage time increases, and data of different quality will have a significant impact on the performance of the model; the fault location capability of the memory part is poor. Summary of the invention

[0003] The embodiments of the present application provide a method for diagnosing a memory fault, a device for diagnosing a memory fault, a computer-readable storage medium, a computer program product, and an electronic device, so as to at least solve the problem in the related art that a memory fault prediction model constructed is inaccurate and has poor robustness due to the scarcity of faulty memory samples.

[0004] According to an embodiment of the present application, a method for diagnosing a memory fault is provided, including: obtaining memory performance indicator data samples of multiple memory performance indicator data of a server, the memory performance indicator data samples being indicator data of normal memory performance indicator data in the memory performance indicator data whose health index is greater than a health threshold; training the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data samples to obtain a fault diagnosis model, wherein a time series analysis algorithm is integrated in the unsupervised learning model; determining a first mean absolute error of the multiple memory performance indicator data according to the multiple memory performance indicator data and the fault diagnosis model; judging whether the first mean absolute error is greater than an error threshold, and issuing a first warning information of a server fault when the mean absolute error is greater than the error threshold.

[0005] In an exemplary embodiment, the acquiring of the memory performance indicator data sample of the server includes: determining the memory performance indicator data sample data in the normal memory performance indicator data according to the normal memory performance indicator data and a principal component analysis technique.

[0006] In an exemplary embodiment, determining the memory performance indicator data sample data in the normal memory performance indicator data according to the normal memory performance indicator data and the principal component analysis technology includes: determining the memory performance indicator data sample data according to a quality scoring function and the normal memory performance indicator data, wherein the quality scoring function is:

[0007] ,in, The mth characteristic data representing the normal memory performance indicator data after decentralization, characteristic data representing the j-th characteristic of the normal memory performance indicator, represents the characteristic value of the j-th characteristic of the normal memory performance indicator, represents the number of eigenvalues ​​and feature data, express The score, The score is inversely proportional to the memory health of the server.

[0008] In an exemplary embodiment, the decentralizing the normal memory performance indicator data includes: obtaining an average of multiple normal memory performance indicator data based on all the normal memory performance indicator data; and decentralizing each normal memory performance indicator data based on the average.

[0009] In an exemplary embodiment, the diagnostic method further includes: performing normalization processing on the memory performance indicator data sample according to a normalization formula, and the normalization formula is: ,in, is the original memory performance index data, is the normalized memory performance index data, is the maximum value of the corresponding original memory performance indicator data, It is the minimum value in the corresponding original memory performance indicator data.

[0010] In an exemplary embodiment, the determining whether the first mean absolute error is greater than an error threshold, and issuing a warning message of the server failure if the first mean absolute error is greater than the error threshold, includes: obtaining a set of memory performance indicator data; determining a set of predicted memory performance indicator data based on the set of memory performance indicator data, the set of memory performance indicator data and the fault diagnosis model; determining the first mean absolute error of a plurality of the memory performance indicator data in the set of predicted memory performance indicator data based on the set of predicted memory performance indicator data and a preset relationship; determining whether the first mean absolute error is greater than an error threshold, and issuing a warning message of the server failure if the first mean absolute error is greater than the error threshold.

[0011] In an exemplary embodiment, an acquisition step: acquiring normal memory performance indicator data and abnormal memory performance indicator data in the memory performance indicator data; a first determination step: determining a second mean absolute error of the normal memory performance indicator data and a third mean absolute error of the abnormal memory performance indicator data according to the memory performance indicator data, the fault diagnosis model and the preset relationship; a first judgment step: judging whether the second mean absolute error is greater than a preset error threshold, and if the second mean absolute error is greater than the preset error threshold, classifying the normal memory performance indicator data as abnormal data, and if the second mean absolute error is less than or equal to the preset error threshold, classifying the normal memory performance indicator data as normal data; a second judgment step: judging whether the third mean absolute error is greater than the preset error threshold, and if the third mean absolute error is greater than the preset error threshold, classifying the normal memory performance indicator data as normal data. The abnormal memory performance indicator data is classified as abnormal data, and when the third mean absolute error is less than or equal to the preset error threshold, the abnormal memory performance indicator data is classified as normal data; a second determination step: determining the accuracy of judging the normal data according to the data volume of the normal data and the data volume of the normal memory performance indicator data in the normal data, or determining the accuracy of judging the abnormal data according to the data volume of the abnormal data and the data volume of the abnormal memory performance indicator data in the abnormal data; a loop step: determining whether the accuracy rate meets the preset accuracy rate, and when the accuracy rate does not meet the preset accuracy rate, re-executing the acquisition step, the first determination step, the first judgment step, the second judgment step and the second determination step at least once, until the accuracy rate meets the preset accuracy rate, and stopping, and determining the preset error threshold as the error threshold.

[0012] In an exemplary embodiment, the preset relationship is: ,in, is the mean absolute error, is the average absolute error between the input sequence and the output sequence of the mth feature in the unsupervised learning model at time t; W and M are natural numbers.

[0013] In an exemplary embodiment, an error set of the first mean absolute error of the memory performance indicator data is obtained, and adjacent memory performance indicator data differ by a window length; it is determined whether the first mean absolute error in the error set is greater than the error threshold, and when it is determined that more than half of the mean absolute errors in the error set are greater than the error threshold, a second warning information of the server failure is issued.

[0014] In an exemplary embodiment, the diagnostic method further includes: when the first warning information and the second warning information are issued, determining a maximum mean absolute error based on all the first mean absolute errors in the error set; and determining a faulty component of the server based on the maximum mean absolute error.

[0015] In an exemplary embodiment, the diagnostic method also includes: when the first warning information and the second warning information of the server failure are issued, determining whether the first mean absolute error of each memory performance indicator data meets the standard threshold of the memory performance indicator data; when it is determined that the first mean absolute error does not meet the standard threshold, determining the faulty component of the server based on the first mean absolute error.

[0016] According to another embodiment of the present application, a memory fault diagnosis device is provided, including: an acquisition module, used to acquire memory performance indicator data samples of multiple memory performance indicator data of a server, the memory performance indicator data samples being indicator data of normal memory performance indicator data in the memory performance indicator data whose health index is greater than a health threshold; a first determination module, used to train the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data samples to obtain a fault diagnosis model, wherein a time series analysis algorithm is integrated in the unsupervised learning model; a second determination module, used to determine a first mean absolute error of multiple memory performance indicator data according to multiple memory performance indicator data and the fault diagnosis model; a control module, used to determine whether the first mean absolute error is greater than an error threshold, and to issue a first warning information of a server failure when the mean absolute error is greater than the error threshold.

[0017] According to another embodiment of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the memory fault diagnosis method.

[0018] According to another embodiment of the present application, a computer program product is provided, including computer instructions, and when the computer instructions are executed by a processor, the memory fault diagnosis method is implemented.

[0019] According to another embodiment of the present application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for executing the memory failure diagnosis method.

[0020] The present application first obtains memory performance indicator data samples of various memory performance indicator data of the server, and then trains the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data samples to obtain a fault diagnosis model. According to the multiple memory performance indicator data and the fault diagnosis model, the first mean absolute error of the multiple memory performance indicator data is determined, and it is judged whether the first mean absolute error is greater than the error threshold. When the mean absolute error is greater than the error threshold, the first warning information of the server fault is issued. The present application obtains a fault diagnosis model by training the unsupervised learning model with normal memory performance indicator data samples. Since the probability of the server generating normal memory performance indicators is much higher than that of generating faulty memory performance indicators, when generating faulty memory performance indicators, the output of the fault diagnosis model will have a larger deviation from the normal memory performance indicators, so the diagnosis accuracy is higher, and the time series analysis algorithm is integrated into the unsupervised learning model to capture the dynamic changes of the memory performance indicators, further enhancing the robustness of the model. It solves the problem in the related art that the memory fault prediction model constructed is inaccurate and has poor robustness due to the scarcity of faulty memory samples. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a hardware structure block diagram of a mobile terminal according to a method for diagnosing a memory fault in an embodiment of the present application;

[0022] Figure 2 is a flow chart of a method for diagnosing a memory fault according to an embodiment of the present application;

[0023] Figure 3 is a training diagram of a fault diagnosis model according to an embodiment of the present application;

[0024] Figure 4It is a structural block diagram of a memory fault diagnosis method device according to an embodiment of the present application.

[0025] Among them, the above reference numerals include:

[0026] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0027] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0029] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0030] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 FIG. 1 is a hardware structure block diagram of a mobile terminal of a memory fault diagnosis method according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0031] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the memory fault diagnosis method in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the mobile terminal via a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0032] In this embodiment, a method for determining power supply recovery of a faulty distribution network running on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0033] In this embodiment, a method for diagnosing a memory fault is provided. Figure 2 is a flow chart of a method for diagnosing a memory failure according to an embodiment of the present application, such as Figure 2 As shown, the diagnostic method of memory failure includes:

[0034] Step S201, obtain memory performance indicator data samples of various memory performance indicator data of the server, where the memory performance indicator data samples are indicator data in which the health index of normal memory performance indicator data in the memory performance indicator data is greater than the health threshold; since there are too few abnormal memory performance indicator samples, the present application uses high-quality indicators in normal memory performance indicators for model training, and firstly, it is necessary to screen out high-quality memory performance indicator data samples from the various memory performance indicator data of the server.

[0035] Step S202, training the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data samples to obtain a fault diagnosis model, wherein the unsupervised learning model is integrated with a time series analysis algorithm; since there is a large amount of data on normal memory performance indicators of the server, the unsupervised learning model is trained with high-quality memory performance indicator data samples, and the fault diagnosis model obtained by training is very accurate in the output of normal memory performance indicators. Therefore, once an abnormal memory performance indicator is generated, the output of the fault diagnosis model will produce obvious abnormalities, and at this time, it can be known which performance indicators are abnormal indicators.

[0036] Step S203, determining the first mean absolute error of multiple memory performance indicator data based on multiple memory performance indicator data and a fault diagnosis model; inputting a performance indicator sequence mixed with normal and abnormal data into the fault diagnosis model, and calculating the first mean absolute error of the above indicators based on the output. The first mean absolute error can characterize the overall error level of the sequence indicators.

[0037] Step S204, determine whether the first mean absolute error is greater than the error threshold, and issue a first warning message of server failure if the mean absolute error is greater than the error threshold. When there is an abnormal indicator in the memory performance indicator, the first mean absolute error obtained will be larger than the error of the normal indicator. By comparing the first mean absolute error obtained with the error threshold, it can be known which indicators are most likely to be abnormal.

[0038] By using normal memory performance indicator data samples to train the unsupervised learning model, a fault diagnosis model is obtained. Since the probability of a server generating a normal memory performance indicator is much higher than that of generating a faulty memory performance indicator, when a faulty memory performance indicator is generated, the output of the fault diagnosis model will have a larger deviation from the normal memory performance indicator, so the diagnosis accuracy is higher. This solves the problem of inaccurate memory fault prediction models constructed due to the scarcity of faulty memory samples in related technologies. In addition, considering that the memory performance indicator is time-varying, the time series analysis algorithm (Long Short-Term Memory, LSTM) is integrated into the unsupervised learning model to capture the dynamic changes of the memory performance indicator and further enhance the robustness of the model. The above unsupervised learning model can be a variational auto-encoder (VAE) model. The fault diagnosis model obtained by training the unsupervised learning model integrated with the time series analysis algorithm is a VAE-LSTM model.

[0039] In an optional solution, obtaining the memory performance indicator data sample of the server in step S201 includes: determining the memory performance indicator data sample data in the normal memory performance indicator data according to the normal memory performance indicator data and the principal component analysis technology.

[0040] In the above embodiment, for the memory, the abnormality of the above performance indicators can reflect the health of the memory. The greater the abnormality of the performance indicators, the worse the health of the memory. To this end, a method based on principal component analysis technology (PCA) is proposed as an auxiliary strategy to find memory performance indicators with good health.

[0041] In an optional solution, the training process of step S202 includes the following steps: determining memory performance indicator data sample data in the normal memory performance indicator data according to the normal memory performance indicator data and the principal component analysis technology, including: determining the memory performance indicator data sample data according to the quality scoring function and the normal memory performance indicator data, the quality scoring function being:

[0042] ,

[0043] in, Represents the mth feature data of the normal memory performance index data after decentralization. The feature data representing the jth feature of the normal memory performance indicator, represents the eigenvalue of the jth feature of the normal memory performance indicator, represents the number of eigenvalues ​​and feature data, express The score, The score is inversely proportional to the memory health of the server.

[0044] In the above embodiment, a scoring mechanism is adopted based on PCA to score the normal data in the memory performance indicator data, and the high-quality data in the normal data is selected to train the unsupervised learning model (VAE), wherein the lower the score, the better the quality of the data, and the higher the score, the worse the quality of the data. The data with low scores and a certain threshold are screened out for the next step of model training.

[0045] In an optional solution, the normal memory performance indicator data is decentralized, including: obtaining an average of multiple normal memory performance indicator data based on all normal memory performance indicator data; and decentralizing each normal memory performance indicator data based on the average.

[0046] In the above embodiment, the normal memory performance index data needs to be decentralized during the scoring process. Decentralization can prevent the data itself from being too large, which in turn leads to the problem of excessive computation during the scoring process. The decentralized result is obtained by subtracting the internal performance index data from the mean. The structure obtained in this way not only retains the differences between the various indicators, but also reduces the computation load.

[0047] In an optional solution, the diagnostic method further includes: normalizing the memory performance indicator data sample according to a normalization formula, the normalization formula being: ,in, It is the raw memory performance index data. is the normalized memory performance index data, is the maximum value of the corresponding original memory performance indicator data, It is the minimum value of the corresponding original memory performance indicator data.

[0048] In the above embodiment, after high-quality memory performance index data is screened out, data preprocessing (normalization) is performed on the data. Different performance indicators have different units and their value ranges vary greatly. This application considers memory performance indicators such as read and write delays, voltage drift, memory error rate, storage loss, power consumption changes, and memory refresh rates. If the screened data is used directly for model training, the model training results may be affected due to the large differences between the various indicators. Therefore, in order to eliminate the problem of poor comparability between performance indicators, the normalization method is used to process the data sets separately, so that each performance indicator data is scaled at (0, 1), reducing the difference between each performance indicator and improving the accuracy of model training.

[0049] The step of determining the first mean absolute error in step S203 includes: determining the first mean absolute error of multiple memory performance indicator data in the set of predicted memory performance indicator data according to the set of memory performance indicator data, the set of predicted memory performance indicator data, and a preset relationship; the preset relationship is:

[0050] ,

[0051] in, is the mean absolute error (MAE), is the mean absolute error between the input sequence and the output sequence of the mth feature in the unsupervised learning model at time t, ; W and M are natural numbers. The above preset relationship is not limited to the calculation formula of the first mean absolute error, and the second mean absolute error and the third mean absolute error in the following text of this application can also be calculated.

[0052] Furthermore, the mean absolute error can also be calculated using:

[0053] ,

[0054] in, is the mean absolute error of M memory performance indicators at time t; is the mean absolute error of the mth feature within the window length W.

[0055] In an optional solution, step S204, determining whether the first mean absolute error is greater than an error threshold, and issuing a warning message of a server failure when the first mean absolute error is greater than the error threshold, includes:

[0056] Get a collection of memory performance indicator data; VAE encoder input window length W time series , is the memory performance indicator vector at time t, expressed as , M is the number of performance indicators.

[0057] According to the set of memory performance indicator data and the fault diagnosis model, determine the set of predicted memory performance indicator data; after the fault diagnosis model is trained, the memory performance indicator data in the server is input into the fault diagnosis model as a set of multiple sequences, and the fault diagnosis model outputs a set of predicted memory performance indicator data. The set of predicted memory performance indicator data and the set of input memory performance indicator data are used for subsequent error calculation. The VAE-LSTM decoder gives a sequence of predicted memory performance indicator data corresponding to the output sequence. ,in ,for Output data.

[0058] It is determined whether the first mean absolute error is greater than the error threshold, and when the first mean absolute error is greater than the error threshold, a warning message of a server failure is issued. The above error threshold is obtained through specific experiments, and can make the fault diagnosis model judgment accuracy reach the highest level.

[0059] The process of determining the error threshold is described below.

[0060] Acquisition steps: obtain normal memory performance indicator data and abnormal memory performance indicator data in the memory performance indicator data; classify the memory performance indicator data in the server into known normal memory performance indicator data and abnormal memory performance indicator data, and mark the normal memory performance indicator data and the abnormal memory performance indicator data for subsequent judgment.

[0061] The first determination step: determine the second mean absolute error of normal memory performance index data and the third mean absolute error of abnormal memory performance index data according to the memory performance index data, the fault diagnosis model and the preset relationship; input the selected normal and abnormal index data at multiple moments with a fixed window length between adjacent moments as a sequence into the fault diagnosis model to obtain an output memory performance index data sequence corresponding to the input data sequence, calculate the absolute error of the fault diagnosis model according to the input and output data of each index data, and multiple absolute errors constitute the average absolute error of the sequence. The second mean absolute error is calculated for the normal memory performance index data and the third mean absolute error is calculated for the abnormal memory performance index data.

[0062] The first judgment step: judge whether the second mean absolute error is greater than the preset error threshold. When the second mean absolute error is greater than the preset error threshold, classify the normal memory performance indicator data as abnormal data; when the second mean absolute error is less than or equal to the preset error threshold, classify the normal memory performance indicator data as normal data; the second judgment step: judge whether the third mean absolute error is greater than the preset error threshold. When the third mean absolute error is greater than the preset error threshold, classify the abnormal memory performance indicator data as abnormal data; when the third mean absolute error is less than or equal to the preset error threshold, classify the abnormal memory performance indicator data as normal data; first pre-select a preset error threshold, then compare the obtained second mean absolute error and third mean absolute error with the preset error threshold, classify those greater than the preset error threshold as abnormal, and those less than the preset error threshold as normal.

[0063] The second determination step: determine the accuracy of judging normal data based on the data volume of normal data and the data volume of normal memory performance indicator data in the normal data, or determine the accuracy of judging abnormal data based on the data volume of abnormal data and the data volume of abnormal memory performance indicator data located in the abnormal data; count the memory performance indicator data marked as normal in the data classified as normal, calculate the proportion of memory performance indicator data in the total data volume, and obtain the diagnostic accuracy of the fault diagnosis model; you can also calculate the accuracy of the data classified as abnormal, and choose one of the two calculation methods.

[0064] Loop step: determine whether the accuracy meets the preset accuracy. If the accuracy does not meet the preset accuracy, re-execute at least one acquisition step, first determination step, first judgment step, second judgment step and second determination step until the accuracy meets the preset accuracy, and determine the preset error threshold as the error threshold. The above-obtained diagnostic accuracy can determine a preset accuracy according to the actual server type. If the accuracy does not meet the preset accuracy, re-execute the above acquisition step, first determination step, first judgment step, second judgment step and second determination step, and re-select the preset error threshold until the accuracy meets the preset accuracy, and determine the preset error threshold as the error threshold.

[0065] like Figure 3 As shown in the figure, the VAE model architecture used in this application has high robustness in fault diagnosis. The goal of VAE is to find the true distribution of training data and use this distribution to generate data. The basic unit is the most basic unit in the network. It is the core component of the LSTM network and is responsible for storing and updating information. It contains input gates, forget gates, unit states, output gates, and output values. m is the characteristic length of LSTM. VAE can be divided into two parts: encoder (AE encoder) and decoder (AE decoder) (corresponding to Figure 3 The left and right parts), the AE encoder maps the input data f to the latent variable, and the AE decoder maps the latent variable back to f. Specifically, in this solution, when constructing the VAE-LSTM model, the normal memory performance indicator data is modeled. The normal memory performance indicator data will be implied to the high probability area in the data distribution. On the contrary, if there is an abnormality in the input data, the output result will be implied to the low probability area in the normal data distribution. That is, VAE believes that the probability of normal memory occurrence is high and the probability of abnormal memory occurrence is low. Based on this assumption, abnormal memory detection can be performed.

[0066] In addition, due to the time-varying properties of memory, incorporating LSTM results into VAE can better capture this time-varying property. When modeling VAE, the input is the sequence data within a time window. Specifically, for each sample in the time series, in addition to the LSTM structure of the normal memory performance indicator data of the sample itself as input, the LSTM structure of the previous sample will also be used as hidden traversal input.

[0067] In an optional scheme, an error set of the first mean absolute error of the memory performance indicator data is obtained, and the difference between adjacent memory performance indicator data is a window length; it is determined whether the first mean absolute error in the error set is greater than the error threshold, and when it is determined that more than half of the mean absolute errors in the error set are greater than the error threshold, a second warning information of the server failure is issued.

[0068] In the above embodiment, when making the final warning decision, all samples in the time window will be used to vote to determine whether the memory will be abnormal. That is, for the error set of the first average error with a window length of W ,in is the absolute error between the input sequence and the output sequence in the tth sequence in the time window sequence. If there are more than W / 2 errors greater than the error threshold, it is considered that the server's memory will be abnormal in the future. This early warning decision-making method that integrates voting can further improve the robustness of the model.

[0069] In an optional solution, the diagnostic method further includes: when the first warning information and the second warning information are issued, determining the maximum mean absolute error based on all first mean absolute errors in the error set; and determining the faulty component of the server based on the maximum mean absolute error.

[0070] In the above embodiment, when it is determined that the memory will fail in the future, the greater the error of the memory performance indicator data, the greater the possibility that the component or software corresponding to the memory performance indicator data will fail. According to the first mean absolute error of the VAE input sequence and the output sequence, the first mean absolute error of the memory performance indicator data is selected from the set of the first mean absolute error of the memory performance indicator data. Filter out the memory performance indicators with the largest error values ​​(among which, as defined above) as the attribution of memory failures.

[0071] In an optional scheme, the diagnostic method also includes: when issuing a first warning message and a second warning message of a server failure, determining whether a first mean absolute error of each memory performance indicator data meets a standard threshold value of the memory performance indicator data; when determining that the first mean absolute error does not meet the standard threshold value, determining the faulty component of the server based on the first mean absolute error.

[0072] Furthermore, the collection The error of each memory performance indicator data is compared with the error threshold corresponding to the memory performance indicator data, and then each memory performance indicator data is judged, so that each part of the server can be diagnosed more comprehensively.

[0073] The embodiment of the present application also provides a memory failure diagnosis device. It should be noted that the memory failure diagnosis device of the embodiment of the present application can be used to execute the memory failure diagnosis method provided by the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0074] The following is an introduction to the memory fault diagnosis device provided in the embodiment of the present application.

[0075] Figure 4 Schematic diagram of a memory fault diagnosis device according to an embodiment of the present application. Figure 4 As shown, the device includes: an acquisition module 100, which is used to obtain memory performance indicator data samples of multiple memory performance indicator data of the server, and the memory performance indicator data samples are indicator data in which the health index of normal memory performance indicator data in the memory performance indicator data is greater than the health threshold; a first determination module 200, which is used to train the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data samples to obtain a fault diagnosis model, and a time series analysis algorithm is integrated in the unsupervised learning model; a second determination module 300, which is used to determine the first mean absolute error of multiple memory performance indicator data according to multiple memory performance indicator data and the fault diagnosis model; a control module 400, which is used to determine whether the first mean absolute error is greater than the error threshold, and when the mean absolute error is greater than the error threshold, issue a first warning information of the server failure.

[0076] The memory fault diagnosis device of the present application comprises an acquisition module, a first determination module, a second determination module and a control module, wherein the acquisition module is used for, the first determination module is used for, the second determination module is used for, and the control module is used for.

[0077] In some optional schemes, the acquisition module includes a first sub-determination module, and the first sub-determination module is used to determine the memory performance indicator data sample data in the normal memory performance indicator data based on the normal memory performance indicator data and the principal component analysis technology.

[0078] In some optional solutions, the first sub-determination module includes a first sub-determination unit, and the first sub-determination unit is used to determine the memory performance indicator data sample data according to the quality scoring function and the normal memory performance indicator data, and the quality scoring function is: ,in, The mth feature data representing the normal memory performance index after decentralization, The feature data representing the jth feature of the normal memory performance indicator, represents the eigenvalue of the jth feature of the normal memory performance indicator, represents the number of eigenvalues ​​and feature data, express The score, The score is inversely proportional to the memory health of the server.

[0079] In some optional schemes, the normal memory performance indicator data is decentralized in the first sub-determination unit, including: obtaining the average of multiple normal memory performance indicator data based on all normal memory performance indicator data; and decentralizing each normal memory performance indicator data based on the average.

[0080] In some optional solutions, the device further includes a data processing module, which is used to normalize the memory performance indicator data samples according to a normalization formula, and the normalization formula is: ,in, It is the raw memory performance index data. is the normalized memory performance index data, is the maximum value of the corresponding original memory performance indicator data, It is the minimum value of the corresponding original memory performance indicator data.

[0081] In some optional schemes, the control module includes a first sub-acquisition module, a second sub-determination module, a third sub-determination module and a first sub-control module, wherein the first sub-acquisition module is used to acquire a set of memory performance indicator data; the second sub-determination module is used to determine a set of predicted memory performance indicator data based on the set of memory performance indicator data and an unsupervised learning model; the third sub-determination module is used to determine a first mean absolute error of multiple memory performance indicator data in the set of predicted memory performance indicator data based on the set of predicted memory performance indicator data and a preset relationship; the first sub-control module is used to determine whether the first mean absolute error is greater than an error threshold, and when the first mean absolute error is greater than the error threshold, issue a warning message of a server failure.

[0082] In some optional schemes, the device also includes a second acquisition module, a third determination module, a first judgment module, a second judgment module, a fourth determination module and a loop module for obtaining an error threshold in the first sub-control module, wherein the second acquisition module is used to obtain normal memory performance indicator data and abnormal memory performance indicator data in the memory performance indicator data; the third determination module is used to determine the second mean absolute error of the normal memory performance indicator data and the third mean absolute error of the abnormal memory performance indicator data according to the memory performance indicator data, the unsupervised learning model and the preset relationship; the first judgment module is used to determine whether the second mean absolute error is greater than the preset error threshold, and when the second mean absolute error is greater than the preset error threshold, the normal memory performance indicator data is classified as abnormal data, and when the second mean absolute error is less than or equal to the preset error threshold, the normal memory performance indicator data is classified as normal data; the second judgment module is used to judge Whether the third mean absolute error is greater than a preset error threshold. If the third mean absolute error is greater than the preset error threshold, the abnormal memory performance indicator data is classified as abnormal data. If the third mean absolute error is less than or equal to the preset error threshold, the abnormal memory performance indicator data is classified as normal data. The fourth determination module is used to determine the accuracy of judging normal data according to the data volume of normal data and the data volume of normal memory performance indicator data in the normal data, or to determine the accuracy of judging abnormal data according to the data volume of abnormal data and the data volume of abnormal memory performance indicator data located in the abnormal data. The loop module is used to determine whether the accuracy meets the preset accuracy. If the accuracy does not meet the preset accuracy, re-execute at least one acquisition step, the first determination step, the first judgment step, the second judgment step, and the second determination step until the accuracy meets the preset accuracy and stops, and determines the preset error threshold as the error threshold.

[0083] In some optional solutions, the preset relationship in the third sub-determination module is: ,in, is the mean absolute error, is the mean absolute error between the input sequence and the output sequence of the mth feature in the unsupervised learning model at time t; W and M are natural numbers.

[0084] In some optional schemes, the device also includes a third acquisition module and a second control module, wherein the third acquisition module is used to obtain an error set of the first mean absolute error of the memory performance indicator data, and adjacent memory performance indicator data differ by a window length; the second control module is used to determine whether the first mean absolute error in the error set is greater than an error threshold, and when it is determined that more than half of the mean absolute errors in the error set are greater than the error threshold, a second warning information of server failure is issued.

[0085] In some optional schemes, the device also includes a fifth determination module and a sixth determination module, wherein, when the first warning information and the second warning information are issued, the fifth determination module is used to determine the maximum mean absolute error based on all the first mean absolute errors in the error set; the sixth determination module is used to determine the faulty component of the server based on the maximum mean absolute error.

[0086] In some optional schemes, the device also includes a third judgment module and a seventh determination module, wherein, when the first warning information and the second warning information of the server failure are issued, the third judgment module is used to judge whether the first mean absolute error of each memory performance indicator data meets the standard threshold of the memory performance indicator data; the seventh determination module is used to determine the faulty component of the server based on the first mean absolute error when it is judged that the first mean absolute error does not meet the standard threshold.

[0087] The above memory fault diagnosis device includes a processor and a memory, and the above acquisition modules and the like are stored in the memory as program units, and the processor executes the above program units stored in the memory to implement corresponding functions. The above modules are all located in the same processor; or, the above modules are located in different processors in any combination.

[0088] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0089] An embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the memory fault diagnosis method.

[0090] Specifically, the memory failure diagnosis method includes:

[0091] Step S201, obtaining memory performance indicator data samples of multiple memory performance indicator data of the server, where the memory performance indicator data samples are indicator data of normal memory performance indicator data in the memory performance indicator data whose health index is greater than a health threshold;

[0092] Step S202, training the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data sample to obtain a fault diagnosis model, wherein the unsupervised learning model is integrated with a time series analysis algorithm;

[0093] Step S203, determining a first mean absolute error of the plurality of memory performance indicator data according to the plurality of memory performance indicator data and the fault diagnosis model;

[0094] Step S204, determining whether the first mean absolute error is greater than an error threshold, and issuing a first warning message of a server failure when the mean absolute error is greater than the error threshold.

[0095] An embodiment of the present invention provides a processor, and the processor is used to run a program, wherein the memory fault diagnosis method is executed when the program is running.

[0096] Specifically, the memory failure diagnosis method includes:

[0097] Step S201, obtaining memory performance indicator data samples of multiple memory performance indicator data of the server, where the memory performance indicator data samples are indicator data of normal memory performance indicator data in the memory performance indicator data whose health index is greater than a health threshold;

[0098] Step S202, training the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data sample to obtain a fault diagnosis model, wherein the unsupervised learning model is integrated with a time series analysis algorithm;

[0099] Step S203, determining a first mean absolute error of the plurality of memory performance indicator data according to the plurality of memory performance indicator data and the fault diagnosis model;

[0100] Step S204, determining whether the first mean absolute error is greater than an error threshold, and issuing a first warning message of a server failure when the mean absolute error is greater than the error threshold.

[0101] An embodiment of the present invention provides an electronic device, the device comprising a processor, a memory, and a program stored in the memory and executable on the processor, and when the processor executes the program, at least the following steps are implemented:

[0102] Step S201, obtaining memory performance indicator data samples of multiple memory performance indicator data of the server, where the memory performance indicator data samples are indicator data of normal memory performance indicator data in the memory performance indicator data whose health index is greater than a health threshold;

[0103] Step S202, training the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data sample to obtain a fault diagnosis model, wherein the unsupervised learning model is integrated with a time series analysis algorithm;

[0104] Step S203, determining a first mean absolute error of the plurality of memory performance indicator data according to the plurality of memory performance indicator data and the fault diagnosis model;

[0105] Step S204, determining whether the first mean absolute error is greater than an error threshold, and issuing a first warning message of a server failure when the mean absolute error is greater than the error threshold.

[0106] The devices in this article can be servers, PCs, PADs, mobile phones, etc.

[0107] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing at least the following method steps:

[0108] Step S201, obtaining memory performance indicator data samples of multiple memory performance indicator data of the server, where the memory performance indicator data samples are indicator data of normal memory performance indicator data in the memory performance indicator data whose health index is greater than a health threshold;

[0109] Step S202, training the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data sample to obtain a fault diagnosis model, wherein the unsupervised learning model is integrated with a time series analysis algorithm;

[0110] Step S203, determining a first mean absolute error of the plurality of memory performance indicator data according to the plurality of memory performance indicator data and the fault diagnosis model;

[0111] Step S204, determining whether the first mean absolute error is greater than an error threshold, and issuing a first warning message of a server failure when the mean absolute error is greater than the error threshold.

[0112] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0113] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0114] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0115] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0116] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0117] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0118] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0119] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0120] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0121] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0122] The above are only preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for diagnosing memory failure, characterized in that: include: Acquire memory performance indicator data samples of multiple memory performance indicator data of the server, wherein the memory performance indicator data samples are indicator data of normal memory performance indicator data in the memory performance indicator data whose health index is greater than a health threshold; The unsupervised learning model is trained according to the unsupervised learning model and the memory performance indicator data sample to obtain a fault diagnosis model, wherein the unsupervised learning model is integrated with a time series analysis algorithm; Determine a first mean absolute error of the plurality of memory performance indicator data according to the plurality of memory performance indicator data and the fault diagnosis model; It is determined whether the first mean absolute error is greater than an error threshold, and if the mean absolute error is greater than the error threshold, a first warning message of a server failure is issued.

2. The diagnostic method according to claim 1, characterized in that The obtaining of the memory performance indicator data sample of the server includes: According to the normal memory performance indicator data and the principal component analysis technology, memory performance indicator data sample data in the normal memory performance indicator data is determined.

3. The diagnostic method according to claim 2, characterized in that The determining, based on the normal memory performance indicator data and the principal component analysis technology, memory performance indicator data sample data in the normal memory performance indicator data includes: The memory performance indicator data sample data is determined according to a quality scoring function and the normal memory performance indicator data, wherein the quality scoring function is: ， in, The mth characteristic data representing the normal memory performance indicator data after decentralization, characteristic data representing the j-th characteristic of the normal memory performance indicator, represents the characteristic value of the j-th characteristic of the normal memory performance indicator, represents the number of eigenvalues ​​and feature data, express The score, The score is inversely proportional to the memory health of the server.

4. The diagnostic method according to claim 3, characterized in that: The decentralization of the normal memory performance indicator data includes: According to all the normal memory performance indicator data, an average value of a plurality of the normal memory performance indicator data is obtained; According to the mean, each of the normal memory performance indicator data is decentralized.

5. The diagnostic method according to claim 3, characterized in that: The diagnostic method further comprises: The memory performance indicator data sample is normalized according to a normalization formula, and the normalization formula is: ， in, is the original memory performance index data, is the normalized memory performance index data, is the maximum value of the corresponding original memory performance indicator data, It is the minimum value in the corresponding original memory performance indicator data.

6. The diagnostic method according to claim 5, characterized in that: The determining whether the first mean absolute error is greater than an error threshold, and issuing warning information of the server failure when the first mean absolute error is greater than the error threshold, includes: Obtaining a collection of memory performance indicator data; Determine a set of predicted memory performance indicator data according to the set of memory performance indicator data and the fault diagnosis model; Determine the first mean absolute error of a plurality of the memory performance indicator data in the set of predicted memory performance indicator data according to the set of memory performance indicator data, the set of predicted memory performance indicator data and a preset relationship; It is determined whether the first mean absolute error is greater than an error threshold, and if the first mean absolute error is greater than the error threshold, a warning message of the server failure is issued.

7. The diagnostic method according to claim 6, characterized in that: Also includes: Acquisition step: acquiring normal memory performance indicator data and abnormal memory performance indicator data from the memory performance indicator data; A first determination step: determining a second mean absolute error of the normal memory performance indicator data and a third mean absolute error of the abnormal memory performance indicator data according to the memory performance indicator data, the fault diagnosis model and the preset relationship; The first judgment step: judging whether the second mean absolute error is greater than a preset error threshold, and classifying the normal memory performance indicator data as abnormal data when the second mean absolute error is greater than the preset error threshold, and classifying the normal memory performance indicator data as normal data when the second mean absolute error is less than or equal to the preset error threshold; A second judgment step: judging whether the third mean absolute error is greater than a preset error threshold, and classifying the abnormal memory performance indicator data as abnormal data when the third mean absolute error is greater than the preset error threshold, and classifying the abnormal memory performance indicator data as normal data when the third mean absolute error is less than or equal to the preset error threshold; The second determination step: determining the accuracy of judging the normal data according to the data volume of the normal data and the data volume of the normal memory performance indicator data in the normal data, or Determining the accuracy of judging the abnormal data according to the data volume of the abnormal data and the data volume of the abnormal memory performance indicator data located in the abnormal data; Loop step: determine whether the accuracy meets the preset accuracy. If the accuracy does not meet the preset accuracy, re-execute the acquisition step, the first determination step, the first judgment step, the second judgment step and the second determination step at least once until the accuracy meets the preset accuracy, and stop, and determine the preset error threshold as the error threshold.

8. The diagnostic method according to claim 6, characterized in that: The preset relationship is: ， in, is the mean absolute error, is the average absolute error between the input sequence and the output sequence of the mth feature in the unsupervised learning model at time t; W and M are natural numbers.

9. The diagnostic method according to claim 6, characterized in that: Also includes: Obtaining an error set of the first mean absolute errors of the memory performance indicator data, where adjacent memory performance indicator data differ by a window length; Determine whether the first mean absolute error in the error set is greater than the error threshold, and issue a second warning information of the server failure when it is determined that more than half of the mean absolute errors in the error set are greater than the error threshold.

10. The diagnostic method according to claim 9, characterized in that The diagnostic method further comprises: In the case of issuing the first warning information and the second warning information, determining a maximum mean absolute error according to all the first mean absolute errors in the error set; A faulty component of the server is determined according to the maximum mean absolute error.

11. The diagnostic method according to claim 9, characterized in that: The diagnostic method further comprises: In the case where the first warning information and the second warning information of the server failure are issued, determining whether the first mean absolute error of each memory performance indicator data meets a standard threshold value of the memory performance indicator data; When it is determined that the first mean absolute error does not meet the standard threshold, a faulty component of the server is determined according to the first mean absolute error.

12. A memory fault diagnosis device, characterized in that: include: An acquisition module, used to acquire memory performance indicator data samples of multiple memory performance indicator data of the server, wherein the memory performance indicator data samples are indicator data of normal memory performance indicator data in the memory performance indicator data whose health index is greater than a health threshold; A first determination module is used to train the unsupervised learning model according to the unsupervised learning model and the memory performance indicator data sample to obtain a fault diagnosis model, wherein the unsupervised learning model is integrated with a time series analysis algorithm; A second determination module, configured to determine a first mean absolute error of a plurality of memory performance indicator data according to the plurality of memory performance indicator data and the fault diagnosis model; The control module is used to determine whether the first mean absolute error is greater than an error threshold, and issue a first warning information of a server failure when the mean absolute error is greater than the error threshold.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the memory fault diagnosis method according to any one of claims 1 to 11.

14. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the memory fault diagnosis method according to any one of claims 1 to 11 is implemented.

15. An electronic device, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include a method for performing a memory failure diagnosis method as described in any one of claims 1 to 11.

Images