Hard disk performance prediction method and device, equipment, storage medium and program product
By using performance image analysis model and abnormal feature analysis model in hard disk performance prediction, the historical performance data of hard disk is used to solve the problem of reduced performance prediction accuracy caused by ignoring historical data in the existing technology, and more accurate hard disk performance is achieved predict.
Patent Information
- Application Number
- CN202411627728.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-14
- Publication Date
- 2025-02-18
AI Technical Summary
The existing hard disk performance prediction methods are based only on the performance data within the current period, ignoring the historical performance data of the hard disk, resulting in a decrease in the accuracy of performance prediction.
The performance image analysis model and anomaly feature analysis model are used to train historical performance data to predict the future performance change trends and abnormal characteristics of the hard disk.
Improve the accuracy of hard disk performance prediction, and provide more accurate performance prediction results by considering the periodic changes of hard disk and historical performance data.
Smart Images
Figure CN119473821A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a hard disk performance prediction method, device, equipment, storage medium and program product. Background Art
[0002] With the continuous development of the computer field, in order to ensure the reliability of information storage, it is necessary to predict the performance of the hard disk. The existing hard disk performance prediction method generally predicts the performance change trend of the hard disk based on the performance data of the hard disk in the current period.
[0003] However, in the existing technology, the performance change trend of the hard disk is predicted only based on the performance data of the hard disk in the current period, which ignores the impact of the hard disk's historical performance on future performance, thereby reducing the accuracy of the hard disk performance prediction. Summary of the invention
[0004] Based on this, it is necessary to provide a hard disk performance prediction method, device, equipment, storage medium and program product that can improve the accuracy of hard disk performance prediction in response to the above technical problems.
[0005] In a first aspect, the present application provides a hard disk performance prediction method, comprising:
[0006] In response to the performance prediction request, obtaining a current performance data sequence of the target hard disk in a current sampling period;
[0007] A performance profile analysis model is used to process the current performance data sequence to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period;
[0008] The abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained by using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period;
[0009] Determine the performance prediction result of the target hard disk based on the predicted performance profile and predicted abnormal characteristics.
[0010] In one embodiment, the performance profile analysis model includes a feature extraction branch and a performance profile prediction branch; the performance profile analysis model is used to process the current performance data sequence to obtain a predicted performance profile of the target hard disk, including:
[0011] The current performance data sequence is input into the feature extraction branch to obtain the current performance profile output by the feature extraction branch after extracting features from the current performance data sequence; the current performance profile is input into the performance profile prediction branch to obtain the predicted performance profile of the target hard disk output by the performance profile prediction branch after processing the current performance profile.
[0012] In one embodiment, the abnormal feature analysis model includes an abnormal feature analysis branch and an abnormal feature prediction branch; the abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk, including:
[0013] The current performance data sequence is input into the anomaly analysis branch, and the current anomaly features are output after the anomaly analysis branch performs outlier mining on the current performance data sequence; the current anomaly features are input into the anomaly feature prediction branch, and the predicted anomaly features of the target hard disk are output after the anomaly feature prediction branch processes the current anomaly features.
[0014] In one embodiment, the anomaly analysis branch includes a first feature network and a second feature network; the current performance data sequence is input into the anomaly analysis branch, and the current anomaly features output by the anomaly analysis branch after performing outlier mining on the current performance data sequence include:
[0015] The current performance data sequence is input into the first feature network to obtain the current isolated tree corresponding to the current performance data sequence output by the first feature network; the current isolated tree is input into the second feature network to obtain the current abnormal feature output by the second feature network after performing outlier mining on the current isolated tree.
[0016] In one of the embodiments, the current abnormal feature is determined based on the difference between the current performance data corresponding to the abnormal node and the standard performance parameter; the abnormal node is determined based on the isolation of each node under the current isolation tree; the standard performance parameter is the average number of reads and writes per second IOPS of the target hard disk in each sample sampling period.
[0017] In one embodiment, the performance profile prediction branch is trained in the following manner:
[0018] The real sample performance portrait corresponding to the sample performance data sequence in each sample sampling period is obtained; the real sample performance portrait of each sampling period is input into the long short-term memory neural network to obtain the predicted performance portrait of each sample sampling period; according to the difference between the real performance portrait and the predicted performance portrait corresponding to each sample sampling period, the long short-term memory neural network is optimized to obtain the performance portrait prediction branch.
[0019] In a second aspect, the present application also provides a hard disk performance prediction device, comprising:
[0020] A sequence acquisition module, used to obtain a current performance data sequence of a target hard disk in a current sampling period in response to a performance prediction request;
[0021] The first prediction module is used to process the current performance data sequence using a performance profile analysis model to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period;
[0022] The second prediction module is used to process the current performance data sequence using an abnormal feature analysis model to obtain the predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period;
[0023] The result determination module is used to determine the performance prediction result of the target hard disk based on the predicted performance profile and predicted abnormal characteristics.
[0024] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0025] In response to the performance prediction request, obtaining a current performance data sequence of the target hard disk in a current sampling period;
[0026] A performance profile analysis model is used to process the current performance data sequence to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period;
[0027] The abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained by using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period;
[0028] Determine the performance prediction result of the target hard disk based on the predicted performance profile and predicted abnormal characteristics.
[0029] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:
[0030] In response to the performance prediction request, obtaining a current performance data sequence of the target hard disk in a current sampling period;
[0031] A performance profile analysis model is used to process the current performance data sequence to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period;
[0032] The abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained by using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period;
[0033] Determine the performance prediction result of the target hard disk based on the predicted performance profile and predicted abnormal characteristics.
[0034] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:
[0035] In response to the performance prediction request, obtaining a current performance data sequence of the target hard disk in a current sampling period;
[0036] A performance profile analysis model is used to process the current performance data sequence to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period;
[0037] The abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained by using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period;
[0038] Determine the performance prediction result of the target hard disk based on the predicted performance profile and predicted abnormal characteristics.
[0039] The hard disk performance prediction method, device, equipment, storage medium and program product are trained to obtain a performance profile analysis model and an abnormal feature analysis model by pre-adopting the sample performance data sequence in each sample sampling period and the real performance portrait and real abnormal features corresponding to each sample sampling period; then, in response to the performance prediction request, the performance profile analysis model is used to process the current performance data sequence to obtain the predicted performance profile of the target hard disk, and the abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk; and then the performance prediction result of the target hard disk is determined based on the predicted performance portrait and the predicted abnormal features. Compared with the related art, which only predicts the performance change trend of the hard disk based on the performance data of the hard disk in the current period, the above method uses the performance profile analysis model and the abnormal feature analysis model to predict the predicted performance portrait and predicted abnormal features based on the periodic change law of the target hard disk, which can ensure the accuracy of the performance prediction result of the target hard disk. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0041] Figure 1 A flowchart of a hard disk performance prediction method in one embodiment;
[0042] Figure 2 A schematic diagram of a process for determining a predicted performance profile in one embodiment;
[0043] Figure 3 A schematic diagram of a process for determining predicted abnormal features in one embodiment;
[0044] Figure 4 A schematic diagram of a process for determining current abnormal features in one embodiment;
[0045] Figure 5 A schematic diagram of a process of training a performance profile prediction branch in one embodiment;
[0046] Figure 6 is a flow chart of a hard disk performance prediction method in another embodiment;
[0047] Figure 7 is a structural block diagram of a hard disk performance prediction device in one embodiment;
[0048] Figure 8FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0049] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0050] With the continuous development of the computer field, in order to ensure the reliability of information storage, it is necessary to predict the performance of the hard disk. The existing hard disk performance prediction method generally predicts the performance change trend of the hard disk based on the performance data of the hard disk in the current period.
[0051] However, in the existing technology, the performance change trend of the hard disk is predicted only based on the performance data of the hard disk in the current period, which ignores the impact of the hard disk's historical performance on future performance, thereby reducing the accuracy of the hard disk performance prediction.
[0052] Based on this, in an exemplary embodiment, Figure 1 As shown, a hard disk performance prediction method is provided, which is described by taking the method applied to a performance prediction device in a server as an example, and specifically includes the following steps:
[0053] S101, in response to a performance prediction request, obtaining a current performance data sequence of a target hard disk in a current sampling period.
[0054] Among them, the performance prediction request is used to represent the request for hard disk performance prediction; the target hard disk is used to represent the hard disk with performance prediction requirements; the current sampling period is the period for collecting the performance data sequence of the target hard disk; the current performance data sequence is the performance data with a time series relationship acquired during the current sampling period. The performance data can be any dimensional data that can reflect the performance of the hard disk, for example, it can include but is not limited to the average number of reads and writes per second of the hard disk IOPS, that is, the average number of reads per second and the average number of writes per second.
[0055] In one possible implementation, after a performance prediction request is detected, identification information of the target hard disk can be obtained from the performance prediction request; then, based on the identification information of the target hard disk and a preset sampling period, performance data at each sampling moment in the current sampling period is obtained, and the performance data at each sampling moment is processed according to the timing relationship between the sampling moments to obtain a current performance data sequence.
[0056] In the embodiment of the present application, the performance prediction request can be actively sent by the operation and maintenance personnel, or it can be automatically sent according to the preset prediction period, and this is not limited in this application. In addition, the sampling period can be a fixed period set by the operation and maintenance personnel based on experience, or it can be an initial period determined based on a large number of experiments, and then the initial period is adaptively adjusted according to the prediction effect in the specific prediction process, and this is not limited in this application.
[0057] S102, using a performance profile analysis model to process the current performance data sequence to obtain a predicted performance profile of the target hard disk.
[0058] Among them, the predicted performance profile can characterize the performance change trend of the target hard disk in the future period. The performance profile analysis model is trained using the sample performance data sequence in each sample sampling period and the real performance profile corresponding to each sample sampling period. In the embodiment of the present application, the sample sampling period can be the historical sampling period of the target hard disk, and accordingly, the sample performance data sequence can be the historical performance data sequence of the target hard disk.
[0059] It can be understood that the initial model can be trained using the sample performance data sequences within each sample sampling period and the actual performance portraits corresponding to each sample sampling period to obtain a performance portrait analysis model, so that the performance portrait analysis model can learn the periodic trend change rules of the target hard disk based on multiple historical performance data sequences of the target hard disk, so as to predict the performance portrait of the target hard disk in the future period.
[0060] In one possible implementation, the current performance data sequence may be input into a performance profile analysis model, and the performance profile analysis model outputs a predicted performance profile of the target hard disk based on the current performance data sequence and model parameters.
[0061] S103, using an abnormal feature analysis model to process the current performance data sequence to obtain predicted abnormal features of the target hard disk.
[0062] The predicted abnormal features can characterize the abnormal features of the target hard disk in the future period. The abnormal feature analysis model is trained by using the sample performance data sequence in each sample sampling period and the real abnormal features corresponding to each sample sampling period.
[0063] It can be understood that the initial model can be trained using the sample performance data sequences within each sample sampling period and the actual abnormal features corresponding to each sample sampling period to obtain an abnormal feature analysis model, so that the abnormal feature analysis model can learn the periodic abnormal occurrence rules of the target hard disk based on multiple historical performance data sequences of the target hard disk, so as to predict the abnormal features of the target hard disk in the future period.
[0064] In one possible implementation, the current performance data sequence may be input into an abnormal feature analysis model, and the abnormal feature analysis model outputs predicted abnormal features of the target hard disk based on the current performance data sequence and model parameters.
[0065] S104, determining the performance prediction result of the target hard disk according to the predicted performance profile and the predicted abnormal characteristics.
[0066] The performance prediction result is used to represent the predicted performance information of the target hard disk in the future period.
[0067] In one possible implementation, the predicted performance profile and predicted abnormal features can be input into a trained result generation model, and the result generation model outputs the performance prediction result of the target hard disk based on the predicted performance profile, predicted abnormal features and model parameters. Alternatively, the predicted performance profile and predicted abnormal features are directly used as the performance prediction result of the target hard disk.
[0068] In another possible implementation, the predicted performance profile and predicted abnormal features can be mapped to a pre-configured result template to obtain the performance prediction result of the target hard disk. It is understandable that in order to remind the operation and maintenance personnel to handle the abnormality, the predicted abnormal features need to be highlighted in the performance prediction result, for example, by highlighting or bolding the font.
[0069] After the performance prediction result is determined, the result display method can be determined according to the abnormality degree of the predicted abnormal feature, and then the performance prediction result of the target hard disk is displayed to the operation and maintenance personnel using the result display method.
[0070] For example, when the degree of abnormality of the predicted abnormal feature is low, the performance prediction result of the target hard disk can be sent to the operation and maintenance personnel in the form of a data report; when the degree of abnormality of the predicted abnormal feature is high, the performance prediction result of the target hard disk can be sent to the operation and maintenance personnel in the form of a data report, and an abnormal prompt can be sent to the terminal held by the operation and maintenance personnel to prompt the operation and maintenance personnel to check in time.
[0071] In the hard disk performance prediction method, the performance profile analysis model and the abnormal feature analysis model are trained by pre-adopting the sample performance data sequence in each sample sampling period and the real performance portrait and real abnormal features corresponding to each sample sampling period; then, in response to the performance prediction request, the performance profile analysis model is used to process the current performance data sequence to obtain the predicted performance profile of the target hard disk, and the abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk; and then the performance prediction result of the target hard disk is determined based on the predicted performance portrait and the predicted abnormal features. Compared with the related art, which only predicts the performance change trend of the hard disk based on the performance data of the hard disk in the current period, the above method uses the performance profile analysis model and the abnormal feature analysis model to predict the predicted performance portrait and predicted abnormal features based on the periodic change law of the target hard disk, which can ensure the accuracy of the performance prediction result of the target hard disk.
[0072] In order to ensure the accuracy of the predicted performance portrait, based on the above embodiment, the performance portrait analysis model includes a feature extraction branch and a performance portrait prediction branch; further, in the embodiment of the present application, an optional method for determining the predicted performance portrait is provided, such as Figure 2 As shown, the specific steps include:
[0073] S201, input the current performance data sequence into the feature extraction branch, and obtain the current performance portrait outputted by the feature extraction branch after extracting features from the current performance data sequence.
[0074] Among them, the feature extraction branch is used to characterize the network structure for extracting features from the current performance data sequence in the performance profile analysis model; the current performance profile is the performance profile of the target hard disk during the current sampling period.
[0075] In one possible implementation, the current performance data sequence may be input into a feature extraction branch in a performance profile analysis model, and the feature extraction branch extracts features from the current performance data sequence based on model parameters to obtain a current performance profile and output it.
[0076] S202, inputting the current performance profile into the performance profile prediction branch, and obtaining a predicted performance profile of the target hard disk output by the performance profile prediction branch after processing the current performance profile.
[0077] Among them, the performance profile prediction branch is used to characterize the network structure in the performance profile analysis model for prediction based on the current performance profile.
[0078] In one possible implementation, the current performance profile output by the feature extraction branch can be input into the performance profile prediction branch, which processes the current performance profile based on model parameters to obtain a predicted performance profile of the target hard disk and outputs it.
[0079] In the embodiment of the present application, by adopting the feature extraction branch and the performance portrait prediction branch, the current performance data sequence is processed in sequence to obtain a predicted performance portrait, which can ensure the accuracy of the predicted performance portrait.
[0080] In order to ensure the accuracy of predicting abnormal features, based on the above embodiment, the abnormal feature analysis model includes an abnormal analysis branch and an abnormal feature prediction branch; further, in the embodiment of the present application, an optional method for determining the predicted abnormal features is provided, such as Figure 3 As shown, the specific steps include:
[0081] S301, inputting the current performance data sequence into the abnormal analysis branch, and obtaining the current abnormal features outputted by the abnormal analysis branch after performing outlier mining processing on the current performance data sequence.
[0082] Among them, the abnormal analysis branch is used to characterize the network structure of the abnormal feature analysis model for extracting abnormal features of the current performance data sequence; the current abnormal feature is the abnormal feature of the target hard disk in the current sampling period.
[0083] In one possible implementation, the current performance data sequence may be input into an abnormality analysis branch in an abnormality feature analysis model, and the abnormality analysis branch extracts abnormal features from the current performance data sequence based on model parameters to obtain and output the current abnormal features.
[0084] S302, inputting the current abnormal feature into the abnormal feature prediction branch, and obtaining the predicted abnormal feature of the target hard disk output by the abnormal feature prediction branch after processing the current abnormal feature.
[0085] Among them, the abnormal feature prediction branch is used to characterize the network structure of the abnormal feature analysis model for prediction based on the current abnormal features.
[0086] In one possible implementation, the current abnormal feature output by the abnormal analysis branch can be input into the abnormal feature prediction branch, and the abnormal feature prediction branch processes the current abnormal feature based on the model parameters to obtain the predicted abnormal feature of the target hard disk and output it.
[0087] In the embodiment of the present application, by adopting the abnormal analysis branch and the abnormal feature prediction branch, the current performance data sequence is processed in sequence to obtain the predicted abnormal features, which can ensure the accuracy of the predicted abnormal features.
[0088] In order to ensure the accuracy of the current abnormal features, based on the above embodiment, the abnormal analysis branch includes a first feature network and a second feature network; further, in the embodiment of the present application, an optional method for determining the current abnormal features is provided, such as Figure 4 As shown, the specific steps include:
[0089] S401, input the current performance data sequence into the first feature network to obtain the current isolated tree corresponding to the current performance data sequence output by the first feature network.
[0090] Among them, the first feature network is used to characterize the network structure of constructing an isolated tree in the abnormal analysis branch; the current isolated tree is the isolated tree corresponding to the current performance data sequence.
[0091] In one possible implementation, the current performance data sequence can be input into the first feature network in the abnormal analysis branch, and the first feature network constructs a current isolation tree that can characterize the abnormal situation of each performance data according to the current performance data sequence. For example, IOPS can be used as a feature dimension, and a current isolation tree corresponding to IOPS can be constructed according to the IOPS in each current performance data in the current performance data sequence.
[0092] In the process of the first feature network constructing the current isolated tree according to the current performance data sequence, it is necessary to limit the height of the current isolated tree. For example, when the height of the constructed current isolated tree reaches the preset limit height, the current isolated tree under the limit height can be directly output. Among them, the preset limit height can be determined by relevant technical personnel based on historical experience, or it can be determined through a large number of experiments, and this is not limited in this application.
[0093] S402, inputting the current isolated tree into the second feature network, and obtaining the current abnormal feature outputted by the second feature network after performing outlier mining processing on the current isolated tree.
[0094] Among them, the second feature network is used to characterize the network structure of mining abnormal features in the abnormal analysis branch. In the embodiment of the present application, the current abnormal feature is determined based on the difference between the current performance data corresponding to the abnormal node and the standard performance parameter; the abnormal node is determined based on the isolation of each node under the current isolation tree; the standard performance parameter is the average number of reads and writes per second IOPS of the target hard disk in each sample sampling period.
[0095] In one possible implementation, the current isolated tree can be input into the second feature network, and the second feature network determines the abnormal nodes from each node based on the isolation status of each node under the current isolated tree, and determines the current abnormal characteristics based on the difference between the current performance data contained in the abnormal node and the standard performance parameters, and outputs them.
[0096] Exemplarily, the abnormal node can be determined from each node according to the node height corresponding to each node, and the current performance data contained in the abnormal node is used as the abnormal performance data. In the case where there is only one abnormal performance data, the current abnormal feature can be determined directly according to the difference between the abnormal performance data and the standard performance parameters; in the case where there are multiple abnormal performance data, the abnormal performance data with the largest abnormal degree can be determined from each abnormal performance data, and the current abnormal feature can be determined according to the difference between the abnormal performance data with the largest abnormal degree and the standard performance parameters.
[0097] For example, when the standard performance parameter is average IOPS, if there is only one abnormal performance data, the current abnormal feature is determined based on the difference between the IOPS of the abnormal performance data and the average IOPS; if there are multiple abnormal performance data, the variance value between the IOPS of each abnormal performance data and the average IOPS can be determined first, and the current abnormal feature is determined based on the difference between the abnormal performance data with the largest variance value and the average IOPS.
[0098] In the embodiment of the present application, the current abnormal feature is determined by performing outlier mining processing on the current isolated tree constructed based on the current performance data sequence, so that the accuracy of the current abnormal feature can be guaranteed.
[0099] In order to ensure the prediction reliability of the performance profile prediction branch, based on the above embodiment, in the embodiment of the present application, an optional method for training the performance profile prediction branch is provided, such as Figure 5 As shown, the specific steps include:
[0100] S501, obtaining the real sample performance portrait corresponding to the sample performance data sequence within each sample sampling period.
[0101] Among them, the real sample performance portrait is used to characterize the real performance change trend of the target hard disk during the sample sampling period.
[0102] In one possible implementation, the actual sample performance portrait corresponding to each sample sampling period may be obtained based on the performance analysis record corresponding to the sample performance data sequence within each sample sampling period.
[0103] It is understandable that for each sample sampling period, the sample performance data sequence in the sample sampling period can be input into the initial feature extraction branch, and the initial feature extraction branch performs feature extraction based on the sample performance data sequence to obtain the performance portrait to be verified in the sample sampling period. The performance portrait to be verified in the sample sampling period can be compared with the real sample performance portrait, and the initial feature extraction branch can be optimized according to the difference to obtain a trained feature extraction branch.
[0104] S502, input the actual sample performance portrait of each sampling period into the long short-term memory neural network to obtain the predicted performance portrait of each sample sampling period.
[0105] Among them, the long short-term memory neural network LSTM is a special recursive neural network. LSTM can use time series to analyze input data, and then establish a periodic prediction model, and can effectively avoid the drawbacks of gradient disappearance / explosion of traditional neural networks.
[0106] In an embodiment of the present application, a performance portrait prediction branch can be composed of multiple layers of LSTM. At this time, the actual sample performance portrait of each sampling period can be used as the input of the first LSTM layer, and the output of the first LSTM layer can be used as the input of the next LSTM layer until the output of the last LSTM layer to obtain a predicted performance portrait.
[0107] Each LSTM layer may include a neural network function layer, a forget gate, a memory gate, and an output gate. The neural network function layer may use the hyperbolic tangent tanh function as the activation function F(X). Correspondingly, the operation of the neural network function layer is Output i =F(W i ·A i +b i ). Among them, Output i W is the output of the neural network function layer that represents the performance of the i-th real sample; i A is the weight vector of the performance portrait of the ith real sample. In the performance portrait, it tends to have a period with a similar portrait, and in the abnormal features, it tends to have data with the same time regularity; i is the performance portrait of the i-th real sample; b i The bias vector corresponding to the performance portrait of the i-th real sample will be adjusted according to the feedback during the training process.
[0108] The forget gate is mainly used to eliminate performance data that does not have periodicity. When there are fewer real sample performance profiles, it is difficult to determine whether there is periodicity, and the use of the forget gate will be reduced. The real sample performance profile with periodicity tendency will be recorded by the memory gate and used as input to participate in the rating of effective information. Obviously, the more obvious the periodicity tendency, the higher the memory rating will be. Then the rating information is integrated through the output layer to obtain the predicted performance profile.
[0109] S503, according to the difference between the real performance portrait and the predicted performance portrait corresponding to each sample sampling period, the long short-term memory neural network is optimized to obtain a performance portrait prediction branch.
[0110] In one possible implementation, the network parameters of the long short-term memory neural network can be optimized based on the difference between the actual performance profile and the predicted performance profile corresponding to each sample sampling period, thereby obtaining a trained performance profile prediction branch.
[0111] By combining the trained feature extraction branch with the trained performance profile prediction branch, a performance profile analysis model can be obtained.
[0112] It is understandable that the abnormal feature analysis model can also be trained using the above steps, which will not be described in detail in the embodiments of the present application.
[0113] In an embodiment of the present application, the long short-term memory neural network is optimized according to the difference between the actual performance portrait and the predicted performance portrait corresponding to each sample sampling period to obtain a performance portrait prediction branch, which can ensure the prediction reliability of the performance portrait prediction branch.
[0114] Figure 6 FIG. 2 is a flow chart of a hard disk performance prediction method in another embodiment. Based on the above embodiment, this embodiment provides an optional example of a hard disk performance prediction method. Figure 6 The specific implementation process is as follows:
[0115] S601, in response to a performance prediction request, obtaining a current performance data sequence of a target hard disk in a current sampling period.
[0116] S602, input the current performance data sequence into the feature extraction branch, and obtain the current performance portrait outputted by the feature extraction branch after extracting features from the current performance data sequence.
[0117] S603, inputting the current performance profile into the performance profile prediction branch in the performance profile analysis model, and obtaining a predicted performance profile of the target hard disk output by the performance profile prediction branch after processing the current performance profile.
[0118] Among them, the performance portrait analysis model is obtained by training using the sample performance data sequence in each sample sampling period and the real performance portrait corresponding to each sample sampling period. Specifically, the real sample performance portrait corresponding to the sample performance data sequence in each sample sampling period is obtained; the real sample performance portrait of each sampling period is input into the long short-term memory neural network to obtain the predicted performance portrait of each sample sampling period; according to the difference between the real performance portrait and the predicted performance portrait corresponding to each sample sampling period, the long short-term memory neural network is optimized to obtain the performance portrait prediction branch in the performance portrait analysis model.
[0119] S604, input the current performance data sequence into the first feature network in the abnormal feature analysis model, and obtain the current isolated tree corresponding to the current performance data sequence output by the first feature network.
[0120] Among them, the abnormal feature analysis model is trained by using the sample performance data series within each sample sampling period and the real abnormal features corresponding to each sample sampling period.
[0121] S605, inputting the current isolated tree into the second feature network in the abnormal feature analysis model, and obtaining the current abnormal feature outputted by the second feature network after performing outlier mining processing on the current isolated tree.
[0122] Optionally, the current abnormal feature is determined based on the difference between the current performance data corresponding to the abnormal node and the standard performance parameters; the abnormal node is determined based on the isolation of each node under the current isolation tree; the standard performance parameter is the average IOPS of the target hard disk in each sample sampling period.
[0123] S606, inputting the current abnormal feature into the abnormal feature prediction branch in the abnormal feature analysis model, and obtaining the predicted abnormal feature of the target hard disk output by the abnormal feature prediction branch after processing the current abnormal feature.
[0124] S607, determining a performance prediction result of the target hard disk according to the predicted performance profile and the predicted abnormal characteristics.
[0125] The specific process of the above S601-S607 can refer to the description of the above method embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.
[0126] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.
[0127] Based on the same inventive concept, the embodiment of the present application also provides a hard disk performance prediction device for implementing the hard disk performance prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more hard disk performance prediction device embodiments provided below can refer to the limitations of the hard disk performance prediction method above, and will not be repeated here.
[0128] In an exemplary embodiment, Figure 7 As shown, a hard disk performance prediction device 1 is provided, comprising: a sequence acquisition module 10, a first prediction module 20, a second prediction module 30 and a result determination module 40, wherein:
[0129] The sequence acquisition module 10 is used to obtain the current performance data sequence of the target hard disk in the current sampling period in response to the performance prediction request;
[0130] The first prediction module 20 is used to process the current performance data sequence using a performance profile analysis model to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period;
[0131] The second prediction module 30 is used to process the current performance data sequence using an abnormal feature analysis model to obtain predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period;
[0132] The result determination module 40 is used to determine the performance prediction result of the target hard disk according to the predicted performance profile and the predicted abnormal characteristics.
[0133] In an exemplary embodiment, the performance profile analysis model includes a feature extraction branch and a performance profile prediction branch; the first prediction module 20 is specifically used to:
[0134] The current performance data sequence is input into the feature extraction branch to obtain the current performance profile output by the feature extraction branch after extracting features from the current performance data sequence; the current performance profile is input into the performance profile prediction branch to obtain the predicted performance profile of the target hard disk output by the performance profile prediction branch after processing the current performance profile.
[0135] In an exemplary embodiment, the abnormal feature analysis model includes an abnormal analysis branch and an abnormal feature prediction branch; the second prediction module 30 includes:
[0136] The first unit is used to input the current performance data sequence into the abnormal analysis branch, and obtain the current abnormal features output by the abnormal analysis branch after performing outlier mining processing on the current performance data sequence;
[0137] The second unit is used to input the current abnormal feature into the abnormal feature prediction branch, and obtain the predicted abnormal feature of the target hard disk output by the abnormal feature prediction branch after processing the current abnormal feature.
[0138] In an exemplary embodiment, the abnormal analysis branch includes a first feature network and a second feature network; the first unit is specifically used to:
[0139] The current performance data sequence is input into the first feature network to obtain the current isolated tree corresponding to the current performance data sequence output by the first feature network; the current isolated tree is input into the second feature network to obtain the current abnormal feature output by the second feature network after performing outlier mining on the current isolated tree.
[0140] In an exemplary embodiment, the current abnormal feature is determined based on the difference between the current performance data corresponding to the abnormal node and the standard performance parameter; the abnormal node is determined based on the isolation of each node under the current isolation tree; the standard performance parameter is the average number of read and write operations per second IOPS of the target hard disk in each sample sampling period.
[0141] In an exemplary embodiment, the hard disk performance prediction device 1 further includes a model training module, wherein the model training module is specifically used for:
[0142] The real sample performance portrait corresponding to the sample performance data sequence in each sample sampling period is obtained; the real sample performance portrait of each sampling period is input into the long short-term memory neural network to obtain the predicted performance portrait of each sample sampling period; according to the difference between the real performance portrait and the predicted performance portrait corresponding to each sample sampling period, the long short-term memory neural network is optimized to obtain the performance portrait prediction branch.
[0143] Each module in the hard disk performance prediction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.
[0144] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 8 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store hard disk performance data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a hard disk performance prediction method is implemented.
[0145] Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0146] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:
[0147] In response to the performance prediction request, obtaining a current performance data sequence of the target hard disk in a current sampling period;
[0148] A performance profile analysis model is used to process the current performance data sequence to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period;
[0149] The abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained by using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period;
[0150] Determine the performance prediction result of the target hard disk based on the predicted performance profile and predicted abnormal characteristics.
[0151] In one embodiment, the performance profile analysis model includes a feature extraction branch and a performance profile prediction branch; when the processor executes the computer program, the processor also implements the following steps:
[0152] The current performance data sequence is input into the feature extraction branch to obtain the current performance profile output by the feature extraction branch after extracting features from the current performance data sequence; the current performance profile is input into the performance profile prediction branch to obtain the predicted performance profile of the target hard disk output by the performance profile prediction branch after processing the current performance profile.
[0153] In one embodiment, the abnormal feature analysis model includes an abnormal feature analysis branch and an abnormal feature prediction branch; when the processor executes the computer program, the following steps are also implemented:
[0154] The current performance data sequence is input into the anomaly analysis branch, and the current anomaly features are output after the anomaly analysis branch performs outlier mining on the current performance data sequence; the current anomaly features are input into the anomaly feature prediction branch, and the predicted anomaly features of the target hard disk are output after the anomaly feature prediction branch processes the current anomaly features.
[0155] In one embodiment, the abnormal analysis branch includes a first feature network and a second feature network; when the processor executes the computer program, the following steps are also implemented:
[0156] The current performance data sequence is input into the first feature network to obtain the current isolated tree corresponding to the current performance data sequence output by the first feature network; the current isolated tree is input into the second feature network to obtain the current abnormal feature output by the second feature network after performing outlier mining on the current isolated tree.
[0157] In one embodiment, the current abnormal feature is determined based on the difference between the current performance data corresponding to the abnormal node and the standard performance parameter; the abnormal node is determined based on the isolation status of each node under the current isolation tree; the standard performance parameter is the average number of reads and writes per second IOPS of the target hard disk in each sample sampling period.
[0158] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0159] The real sample performance portrait corresponding to the sample performance data sequence in each sample sampling period is obtained; the real sample performance portrait of each sampling period is input into the long short-term memory neural network to obtain the predicted performance portrait of each sample sampling period; according to the difference between the real performance portrait and the predicted performance portrait corresponding to each sample sampling period, the long short-term memory neural network is optimized to obtain the performance portrait prediction branch.
[0160] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
[0161] In response to the performance prediction request, obtaining a current performance data sequence of the target hard disk in a current sampling period;
[0162] A performance profile analysis model is used to process the current performance data sequence to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period;
[0163] The abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained by using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period;
[0164] Determine the performance prediction result of the target hard disk based on the predicted performance profile and predicted abnormal characteristics.
[0165] In one embodiment, the performance profile analysis model includes a feature extraction branch and a performance profile prediction branch; when the computer program is executed by a processor, the following steps are also implemented:
[0166] The current performance data sequence is input into the feature extraction branch to obtain the current performance profile output by the feature extraction branch after extracting features from the current performance data sequence; the current performance profile is input into the performance profile prediction branch to obtain the predicted performance profile of the target hard disk output by the performance profile prediction branch after processing the current performance profile.
[0167] In one embodiment, the abnormal feature analysis model includes an abnormal feature analysis branch and an abnormal feature prediction branch; when the computer program is executed by the processor, the following steps are also implemented:
[0168] The current performance data sequence is input into the anomaly analysis branch, and the current anomaly features are output after the anomaly analysis branch performs outlier mining on the current performance data sequence; the current anomaly features are input into the anomaly feature prediction branch, and the predicted anomaly features of the target hard disk are output after the anomaly feature prediction branch processes the current anomaly features.
[0169] In one embodiment, the abnormal analysis branch includes a first feature network and a second feature network; when the computer program is executed by a processor, the following steps are also implemented:
[0170] The current performance data sequence is input into the first feature network to obtain the current isolated tree corresponding to the current performance data sequence output by the first feature network; the current isolated tree is input into the second feature network to obtain the current abnormal feature output by the second feature network after performing outlier mining on the current isolated tree.
[0171] In one embodiment, the current abnormal feature is determined based on the difference between the current performance data corresponding to the abnormal node and the standard performance parameter; the abnormal node is determined based on the isolation status of each node under the current isolation tree; the standard performance parameter is the average number of reads and writes per second IOPS of the target hard disk in each sample sampling period.
[0172] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0173] The real sample performance portrait corresponding to the sample performance data sequence in each sample sampling period is obtained; the real sample performance portrait of each sampling period is input into the long short-term memory neural network to obtain the predicted performance portrait of each sample sampling period; according to the difference between the real performance portrait and the predicted performance portrait corresponding to each sample sampling period, the long short-term memory neural network is optimized to obtain the performance portrait prediction branch.
[0174] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:
[0175] In response to the performance prediction request, obtaining a current performance data sequence of the target hard disk in a current sampling period;
[0176] A performance profile analysis model is used to process the current performance data sequence to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period;
[0177] The abnormal feature analysis model is used to process the current performance data sequence to obtain the predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained by using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period;
[0178] Determine the performance prediction result of the target hard disk based on the predicted performance profile and predicted abnormal characteristics.
[0179] In one embodiment, the performance profile analysis model includes a feature extraction branch and a performance profile prediction branch; when the computer program is executed by a processor, the following steps are also implemented:
[0180] The current performance data sequence is input into the feature extraction branch to obtain the current performance profile output by the feature extraction branch after extracting features from the current performance data sequence; the current performance profile is input into the performance profile prediction branch to obtain the predicted performance profile of the target hard disk output by the performance profile prediction branch after processing the current performance profile.
[0181] In one embodiment, the abnormal feature analysis model includes an abnormal feature analysis branch and an abnormal feature prediction branch; when the computer program is executed by the processor, the following steps are also implemented:
[0182] The current performance data sequence is input into the anomaly analysis branch, and the current anomaly features are output after the anomaly analysis branch performs outlier mining on the current performance data sequence; the current anomaly features are input into the anomaly feature prediction branch, and the predicted anomaly features of the target hard disk are output after the anomaly feature prediction branch processes the current anomaly features.
[0183] In one embodiment, the abnormal analysis branch includes a first feature network and a second feature network; when the computer program is executed by a processor, the following steps are also implemented:
[0184] The current performance data sequence is input into the first feature network to obtain the current isolated tree corresponding to the current performance data sequence output by the first feature network; the current isolated tree is input into the second feature network to obtain the current abnormal feature output by the second feature network after performing outlier mining on the current isolated tree.
[0185] In one embodiment, the current abnormal feature is determined based on the difference between the current performance data corresponding to the abnormal node and the standard performance parameter; the abnormal node is determined based on the isolation status of each node under the current isolation tree; the standard performance parameter is the average number of reads and writes per second IOPS of the target hard disk in each sample sampling period.
[0186] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0187] The real sample performance portrait corresponding to the sample performance data sequence in each sample sampling period is obtained; the real sample performance portrait of each sampling period is input into the long short-term memory neural network to obtain the predicted performance portrait of each sample sampling period; according to the difference between the real performance portrait and the predicted performance portrait corresponding to each sample sampling period, the long short-term memory neural network is optimized to obtain the performance portrait prediction branch.
[0188] It should be noted that the data involved in this application (including but not limited to hard disk performance data, etc.) are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0189] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.
[0190] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0191] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A hard disk performance prediction method, characterized in that: The method comprises: In response to the performance prediction request, obtaining a current performance data sequence of the target hard disk in a current sampling period; Using a performance profile analysis model, the current performance data sequence is processed to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period; Using an abnormal feature analysis model, the current performance data sequence is processed to obtain predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period; A performance prediction result of the target hard disk is determined based on the predicted performance portrait and the predicted abnormal characteristics.
2. The method according to claim 1, characterized in that The performance profile analysis model includes a feature extraction branch and a performance profile prediction branch; The adopting the performance profile analysis model to process the current performance data sequence to obtain the predicted performance profile of the target hard disk includes: Inputting the current performance data sequence into the feature extraction branch to obtain a current performance portrait outputted by the feature extraction branch after extracting features from the current performance data sequence; The current performance profile is input into the performance profile prediction branch to obtain a predicted performance profile of the target hard disk output by the performance profile prediction branch after processing the current performance profile.
3. The method according to claim 1, characterized in that The abnormal feature analysis model includes an abnormal analysis branch and an abnormal feature prediction branch; The adopting the abnormal feature analysis model to process the current performance data sequence to obtain the predicted abnormal feature of the target hard disk includes: Inputting the current performance data sequence into the abnormality analysis branch to obtain the current abnormality features outputted by the abnormality analysis branch after performing outlier mining processing on the current performance data sequence; The current abnormal feature is input into the abnormal feature prediction branch, and the abnormal feature prediction branch processes the current abnormal feature and outputs the predicted abnormal feature of the target hard disk.
4. The method according to claim 3, characterized in that The abnormal analysis branch includes a first feature network and a second feature network; The step of inputting the current performance data sequence into the abnormality analysis branch to obtain the current abnormality features outputted by the abnormality analysis branch after performing outlier mining processing on the current performance data sequence includes: Inputting the current performance data sequence into the first feature network to obtain a current isolated tree corresponding to the current performance data sequence output by the first feature network; The current isolated tree is input into the second feature network to obtain the current abnormal feature outputted by the second feature network after performing outlier mining processing on the current isolated tree.
5. The method according to claim 4, characterized in that The current abnormal feature is determined based on the difference between the current performance data corresponding to the abnormal node and the standard performance parameters; the abnormal node is determined based on the isolation status of each node under the current isolated tree; the standard performance parameter is the average number of reads and writes per second IOPS of the target hard disk in each sample sampling period.
6. The method according to claim 1, characterized in that The performance profile prediction branch is trained in the following way: Obtain the real sample performance portrait corresponding to the sample performance data sequence within each sample sampling period; The actual sample performance portrait of each sampling period is input into the long short-term memory neural network to obtain the predicted performance portrait of each sampling period; According to the difference between the real performance portrait and the predicted performance portrait corresponding to each sample sampling period, the long short-term memory neural network is optimized to obtain a performance portrait prediction branch.
7. A hard disk performance prediction device, characterized in that: The device comprises: A sequence acquisition module, used to obtain a current performance data sequence of a target hard disk in a current sampling period in response to a performance prediction request; A first prediction module is used to process the current performance data sequence using a performance profile analysis model to obtain a predicted performance profile of the target hard disk; wherein the performance profile analysis model is trained using the sample performance data sequence within each sample sampling period and the real performance profile corresponding to each sample sampling period; A second prediction module is used to process the current performance data sequence using an abnormal feature analysis model to obtain predicted abnormal features of the target hard disk; wherein the abnormal feature analysis model is trained using the sample performance data sequence within each sample sampling period and the real abnormal features corresponding to each sample sampling period; A result determination module is used to determine the performance prediction result of the target hard disk according to the predicted performance portrait and the predicted abnormal characteristics.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.