Solid state disk performance evaluation method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of solid state disk performance evaluation. The solid state disk performance evaluation method comprises the steps of obtaining solid state disk operation information, conducting exception analysis on the solid state disk operation information through a deep learning model to obtain solid state disk exception information, conducting correction and interpolation on the solid state disk exception information, and obtaining the solid state disk performance evaluation result. Performing feature extraction on the corrected operation information to obtain a plurality of performance features of the solid state disk, screening the plurality of performance features based on a feature selection method to obtain a target feature set required by performance evaluation, and analyzing the target feature set required by performance evaluation through a time sequence prediction model. Obtaining performance change trend information under different loads, predicting the residual life and the performance degradation condition of the solid state disk according to the performance change trend information, and obtaining a performance evaluation result. The method has the effect of improving the performance evaluation accuracy of the solid state disk.

Description

Technical Field

[0001] The present invention relates to the technical field of solid state drive performance evaluation, and in particular to a solid state drive performance evaluation method, device, computer equipment and storage medium. Background Art

[0002] Currently, solid-state drives (SSDs) are widely used in storage devices, with high read and write speeds and long service life. However, over time, the performance of SSDs will gradually degrade, especially under high load or harsh working environments, where the performance degradation rate of the hard disk may accelerate. Existing SSD performance evaluation methods usually rely on health monitoring indicators and simple statistical analysis, which can only provide limited information and cannot comprehensively and accurately predict the remaining life and performance degradation trend of the hard disk.

[0003] There is a major flaw in current technology: the existing SSD performance evaluation methods lack accurate correction and interpolation mechanisms for operating information, especially in the processing of abnormal information, which can easily lead to inaccurate data, thus affecting the accuracy and reliability of performance evaluation results. Traditional methods fail to effectively combine deep learning models to correct abnormal information, and fail to use multiple performance characteristics for optimization and screening, thus failing to accurately reflect the performance change trend of hard disks under different loads. Summary of the invention

[0004] In order to improve the accuracy of solid-state drive performance evaluation, the present application provides a solid-state drive performance evaluation method, apparatus, computer equipment, and storage medium.

[0005] The above-mentioned invention objective of the present application is achieved through the following technical solutions: A solid state drive performance evaluation method, the solid state drive performance evaluation method comprising: Get the SSD operation information; Performing an abnormal analysis on the operation information of the solid-state hard disk through a deep learning model to obtain abnormal information of the solid-state hard disk, and correcting and interpolating the abnormal information of the solid-state hard disk to obtain corrected operation information; Performing feature extraction on the corrected operation information to obtain multiple performance features of the solid state drive; Screening the multiple performance features based on a feature selection method to obtain a target feature set required for performance evaluation; Analyzing the target feature set required for the performance evaluation through a time series prediction model to obtain performance change trend information under different loads; Based on the performance change trend information, the remaining life and performance degradation of the solid state drive are predicted to obtain a performance evaluation result.

[0006] By adopting the above technical solution, by inputting the operation information of the solid-state drive into the deep learning model for encoding and decoding, potential abnormal information can be effectively identified, and these abnormal information can be corrected and interpolated based on the generative adversarial network to ensure the integrity and accuracy of the data. By extracting features from the corrected operation information and combining feature selection methods to screen out the most representative performance features, redundant data can be greatly reduced and the computational efficiency of the model can be improved. Furthermore, by analyzing the target feature set through the time series prediction model, the performance change trend under different loads can be accurately captured, and the remaining life and performance degradation of the solid-state drive can be predicted in a timely manner. A detailed performance evaluation report is provided to users to help users perform preventive maintenance, extend the service life of the hard drive, and reduce the risk of data loss.

[0007] In a preferred example, the present application may be further configured as follows: performing abnormal analysis on the operation information of the solid-state hard disk by using a deep learning model to obtain abnormal information of the solid-state hard disk, and correcting the abnormal information of the solid-state hard disk to obtain corrected operation information, including: Encoding and decoding the solid state drive operation information by using the deep learning model, calculating a reconstruction error, and if the reconstruction error exceeds a predetermined error threshold, determining that the solid state drive operation information contains the solid state drive abnormal information; The abnormal information is corrected based on the generative adversarial network in the deep learning model, the correction information is generated based on the generator of the generative adversarial network, and the corrected information is compared with the standard information through the discriminator of the generative adversarial network. If the distribution difference between the corrected information and the standard information is less than a predetermined difference threshold, the solid-state hard disk abnormal information is replaced by the correction information to obtain the corrected operation information.

[0008] By adopting the above technical solutions, abnormal information of SSDs can be accurately identified, and these abnormal data can be automatically corrected through deep learning models to ensure the accuracy and completeness of the data. The generator and discriminator mechanism based on the generative adversarial network can effectively generate corrected data from existing normal information and compare it with standard information to ensure that the generated data conforms to the real distribution. This method not only improves the accuracy of data processing, but also can automatically correct abnormal information without human intervention, improves the quality and reliability of SSD operation data, and provides a solid data foundation for subsequent performance evaluation and life prediction.

[0009] In a preferred example, the present application may be further configured as follows: encoding and decoding the solid state drive operation information by the deep learning model, and calculating the reconstruction error, including: The reconstruction error is obtained based on the following formula: , where E is the reconstruction error, N is the total number of samples of the SSD operation information, and X i is the original input of the i-th sample, the Y i is the predicted output of the i-th sample, the w i is the sample weight, and p is the order of error calculation.

[0010] By adopting the above technical solution, the reconstruction error of the SSD operation information can be accurately calculated, and the degree of data abnormality can be accurately evaluated by comparing the predicted output of each sample with the original input. This solution enhances the flexibility and accuracy of error evaluation by introducing sample weights and error calculation orders, and can more accurately identify abnormal information. In addition, the calculation of the reconstruction error takes into account the importance of each sample, further improving the model's adaptability and correction capabilities for SSD operation information.

[0011] In a preferred example, the present application may be further configured as follows: the feature selection method is used to screen the multiple performance features to obtain a target feature set required for performance evaluation, including: Calculate the correlation of the multiple performance features, and select performance features whose correlation coefficients are lower than a discrimination threshold to obtain a first feature set; Performing variance calculation on the first feature set, selecting performance features whose variance is greater than a predetermined variance threshold, and obtaining a second feature set; The second feature set is input into the L1 regularization model, and the performance features that contribute greatly to the solid state drive performance evaluation are selected to obtain the target feature set required for the performance evaluation.

[0012] By adopting the above technical solution, the most critical features for SSD performance evaluation can be effectively screened out. First, through correlation calculation, redundant information is eliminated to ensure the independence and effectiveness of the feature set. Then, through variance calculation, invalid features with small variance are further filtered out to ensure that the selected features have sufficient discrimination and representativeness. Finally, through the introduction of the L1 regularization model, those features that have made significant contributions to the SSD performance evaluation are accurately selected, making the model more concise and efficient, while also improving the accuracy of the prediction.

[0013] In a preferred example, the present application may be further configured as follows: the second feature set is input into the L1 regularization model, performance features that contribute greatly to the solid-state drive performance evaluation are selected, and the target feature set required for the performance evaluation is obtained, including: Inputting the second feature set into the L1 regularization model, calculating the weight coefficient of each performance feature by introducing an L1 norm constraint on each feature, and obtaining the L1 regularization coefficient of each performance feature; Sorting the L1 regularization coefficient of each performance feature, selecting the performance feature whose absolute value of the regularization coefficient is greater than a predetermined coefficient threshold, and obtaining a preliminary target feature set; Redundant features are removed from the preliminary target feature set, and features with high correlation are removed through correlation analysis between features to obtain the target feature set required for the performance evaluation.

[0014] By adopting the above technical solutions, the features that have the greatest impact on the performance evaluation of solid-state drives can be accurately screened out. The L1 regularization model automatically compresses unimportant features by introducing L1 norm constraints on each feature, effectively removing redundant information, thereby improving the interpretability and predictive power of the model. By sorting the regularization coefficients, it ensures that features that contribute significantly to performance evaluation are given priority, while avoiding over-reliance on irrelevant features. Then, by removing redundant features and highly correlated features, the dimensionality of the feature set is further reduced, improving the efficiency and prediction accuracy of the model.

[0015] In a preferred example, the present application may be further configured as follows: the target feature set required for the performance evaluation is analyzed by the time series prediction model to obtain performance change trend information under different loads, including: Inputting the target feature set required for the performance evaluation into the time series prediction model, calculating the performance change trend within each time step based on the time series relationship in the time series prediction model, and obtaining performance change trend information; Perform difference analysis on the performance change trend information, identify the change rules of performance fluctuations under different loads, and obtain the performance change trend information under different loads.

[0016] By adopting the above technical solutions, the performance change trend of SSD under different load conditions can be accurately captured. Through the time series prediction model, the performance change trend within each time step is calculated based on the time series relationship, ensuring the dynamic and timeliness of performance evaluation. Further, through difference analysis, the performance fluctuation pattern under different loads is identified, providing more accurate information support for the health monitoring and life prediction of SSD in different usage scenarios.

[0017] In a preferred example, the present application may be further configured as follows: predicting the remaining life and performance degradation of the solid state drive according to the performance change trend information to obtain a performance evaluation result, including: The remaining life of the solid state drive under different loads is calculated based on the following formula using the performance change trend information: , wherein the L remaining is the remaining life, L0 is the initial life, and R current is the current performance degradation value, the Rmax is the performance degradation value, β is the degradation acceleration coefficient, γ is the load influence coefficient, and f load is the load factor; Based on the remaining life, the remaining life is corrected in combination with the working environment parameter information of the solid state drive to obtain a corrected remaining life; The performance degradation of the solid state drive is predicted based on the corrected remaining life to obtain the performance evaluation result.

[0018] By adopting the above technical solution, the remaining life of the solid-state drive can be accurately calculated based on the performance change trend information and load changes, and further corrected in combination with the working environment parameters to ensure the accuracy of the evaluation results. The corrected remaining life can more reasonably predict the performance degradation of the hard drive, thereby achieving more accurate health monitoring and life prediction. This solution provides a scientific basis for preventive maintenance and timely upgrades of solid-state drives, helping to reduce the risk of failure and extend the service life of the equipment.

[0019] The second object of the invention is achieved by the following technical solutions: A solid state hard disk performance evaluation device, the solid state hard disk performance evaluation device comprising: An operation information acquisition module is used to obtain the operation information of the solid state drive; An abnormal information correction module is used to perform abnormal analysis on the operation information of the solid-state hard disk through a deep learning model to obtain abnormal information of the solid-state hard disk, and to correct and interpolate the abnormal information of the solid-state hard disk to obtain corrected operation information; A performance feature extraction module, used to extract features from the corrected operation information to obtain multiple performance features of the solid state drive; A feature screening module, used to screen the multiple performance features based on a feature selection method to obtain a target feature set required for performance evaluation; A performance trend analysis module is used to analyze the target feature set required for the performance evaluation through a time series prediction model to obtain performance change trend information under different loads; The life prediction module is used to predict the remaining life and performance degradation of the solid state drive according to the performance change trend information to obtain a performance evaluation result.

[0020] By adopting the above technical solution, by inputting the operation information of the solid-state drive into the deep learning model for encoding and decoding, potential abnormal information can be effectively identified, and these abnormal information can be corrected and interpolated based on the generative adversarial network to ensure the integrity and accuracy of the data. By extracting features from the corrected operation information and combining feature selection methods to screen out the most representative performance features, redundant data can be greatly reduced and the computational efficiency of the model can be improved. Furthermore, by analyzing the target feature set through the time series prediction model, the performance change trend under different loads can be accurately captured, and the remaining life and performance degradation of the solid-state drive can be predicted in a timely manner. A detailed performance evaluation report is provided to users to help users perform preventive maintenance, extend the service life of the hard drive, and reduce the risk of data loss.

[0021] The third objective of the present application is achieved through the following technical solutions: A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned solid state drive performance evaluation method when executing the computer program.

[0022] The fourth objective of the present application is achieved through the following technical solutions: A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the solid-state hard disk performance evaluation method.

[0023] In summary, the present application includes at least one of the following beneficial technical effects: 1. By inputting the operation information of the SSD into the deep learning model for encoding and decoding, potential abnormal information can be effectively identified, and these abnormal information can be corrected and interpolated based on the generative adversarial network to ensure the integrity and accuracy of the data. By extracting features from the corrected operation information and combining feature selection methods to screen out the most representative performance features, redundant data can be greatly reduced and the computational efficiency of the model can be improved. Furthermore, by analyzing the target feature set through the time series prediction model, the performance change trend under different loads can be accurately captured, and the remaining life and performance degradation of the SSD can be predicted in a timely manner. A detailed performance evaluation report can be provided to users to help users perform preventive maintenance, extend the service life of the hard disk, and reduce the risk of data loss.

[0024] 2. It can accurately identify abnormal information of SSDs and automatically correct these abnormal data through deep learning models to ensure the accuracy and completeness of the data. The generator and discriminator mechanism based on the generative adversarial network can effectively generate corrected data from existing normal information and compare it with standard information to ensure that the generated data conforms to the real distribution. This method not only improves the accuracy of data processing, but also can automatically correct abnormal information without human intervention, improves the quality and reliability of SSD operation data, and provides a solid data foundation for subsequent performance evaluation and life prediction; 3. It can accurately calculate the remaining life of the SSD based on performance trend information and load changes, and further correct it in combination with working environment parameters to ensure the accuracy of the evaluation results. The corrected remaining life can more reasonably predict the performance degradation of the hard disk, thereby achieving more accurate health monitoring and life prediction. This solution provides a scientific basis for preventive maintenance and timely upgrades of SSDs, helping to reduce the risk of failure and extend the service life of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a flow chart of a solid state drive performance evaluation method in one embodiment of the present application; Figure 2 is a flowchart for implementing step S20 in the solid state drive performance evaluation method in one embodiment of the present application; Figure 3 is a flowchart for implementing step S201 in the solid state drive performance evaluation method in one embodiment of the present application; Figure 4 is a flowchart for implementing step S40 in the solid state drive performance evaluation method in one embodiment of the present application; Figure 5 is a flowchart for implementing step S403 in the solid state drive performance evaluation method in one embodiment of the present application; Figure 6 is a flowchart for implementing step S50 in the solid state drive performance evaluation method in one embodiment of the present application; Figure 7 is a flowchart for implementing step S60 in the solid state drive performance evaluation method in one embodiment of the present application; Figure 8 is a principle block diagram of a solid state drive performance evaluation device in one embodiment of the present application; Fig. 9 It is a schematic diagram of a device in an embodiment of the present application. DETAILED DESCRIPTION

[0026] The present application is further described in detail below in conjunction with the accompanying drawings.

[0027] In one embodiment, if Figure 1 As shown, the present application discloses a solid state drive performance evaluation method, which specifically includes the following steps: S10: Obtaining the operation information of the solid state drive.

[0028] Specifically, various operating parameters such as temperature, power consumption, read and write speed, number of erase and write times, life estimation value, etc. are collected in real time from the hard disk controller or storage management system. This information can be obtained through the SMART protocol or accessed through the NVMe or SATA interface of the solid-state drive. For example, when the hard disk starts working, it will regularly record the amount of data written each time, the number of data blocks read, the ECC correction error rate and other information, and store them in the log system. In addition, a timed sampling mechanism can be set to automatically collect operating data every once in a while and store it in the data cache area for subsequent analysis. The collected data is preprocessed, such as removing duplicate records, filling missing values, and normalization to ensure the integrity and consistency of the data, so as to obtain the operating information of the solid-state drive.

[0029] S20: Perform an abnormal analysis on the operation information of the solid-state hard disk through a deep learning model to obtain abnormal information of the solid-state hard disk, and correct and interpolate the abnormal information of the solid-state hard disk to obtain corrected operation information.

[0030] Specifically, first, the collected SSD operation information is input into the deep learning model. The deep learning model uses a neural network-based autoencoder structure to detect abnormal values ​​by encoding and decoding the operation information. For example, when an operation parameter such as the number of writes suddenly deviates significantly from the historical information, the model will identify this deviation and mark it as abnormal. The generative adversarial network (GAN) in the deep learning model can correct these abnormal information by training a generator and a discriminator. The generator generates corrected information based on the existing normal information, and the discriminator compares the generated information with the real information to ensure that the generated information is consistent with the normal information distribution. The generated information is only accepted when it is considered reasonable by the discriminator. The corrected information will replace the original abnormal information, and the interpolation process will also fill in the missing part of the information. The interpolation method (such as linear interpolation or polynomial interpolation) is used to ensure that the generated corrected information can represent the real operation status, thereby obtaining the corrected operation information.

[0031] S30: extracting features from the corrected operating information to obtain multiple performance features of the solid state drive.

[0032] Specifically, feature extraction is based on the corrected operating information, which includes timing features, frequency domain features, and statistical features. For example, by performing timing analysis on the corrected temperature, power consumption, read / write rate and other data, the trend of each feature changing over time can be extracted, and then the change rate of each feature (such as the temperature change rate), fluctuation amplitude (such as the degree of fluctuation of the read / write rate) and other information can be obtained. At the same time, in order to capture possible periodic changes, the corrected data will be Fourier transformed to convert the time domain signal into a frequency domain signal, thereby extracting periodic frequency features, such as the periodic fluctuation of read / write operations. In addition, the stability of performance is evaluated by calculating the statistical features such as the mean, variance, and standard deviation of the data, and then the regularity and abnormal pattern of each performance feature are extracted, and finally multiple performance features of the solid-state drive are formed.

[0033] S40: Screening multiple performance features based on a feature selection method to obtain a target feature set required for performance evaluation.

[0034] Specifically, the first step of feature selection is to calculate the correlation between various performance features. For example, the linear relationship between each feature is calculated through the Pearson correlation coefficient, and redundant features with too high correlation are eliminated to avoid information duplication. Then, variance analysis is performed on the remaining features to eliminate features with too low variance. For example, if the variance of a feature is close to zero, it means that it varies very little between different samples and has limited contribution to the evaluation results, so it can be removed. Next, the L1 regularization method is used to further screen the features. L1 regularization imposes sparsity constraints on the features, so that features with smaller weights are automatically eliminated. For example, after applying LASSO regression, the regression coefficients of some features that contribute less to the prediction will become zero. These features can be removed, and finally a set of target feature sets that are most representative of the SSD performance evaluation are retained as the target feature set required for performance evaluation.

[0035] S50: Analyze the target feature set required for performance evaluation through the time series prediction model to obtain performance change trend information under different loads.

[0036] Specifically, the time series prediction model uses LSTM to analyze the target feature set. The model can identify the time dependency of data and accurately predict the performance change trend. For example, the data of read and write rate, temperature, erase and write times in the previous 30 days are input into the model, and the model will predict the possible change trend of hard disk performance in the next 7 days. The performance trend information under different loads is obtained by adjusting the load parameters in the input data. For example, under low load, the read and write rate of the hard disk fluctuates less, while under high load, the temperature and power consumption may rise faster. The prediction results are presented in the form of a time series curve, which can clearly observe the performance fluctuation of the hard disk under different loads, thus providing a basis for the remaining life prediction.

[0037] S60: Predict the remaining life and performance degradation of the solid state drive according to the performance change trend information, and obtain a performance evaluation result.

[0038] Specifically, the degradation rate of the solid-state drive is calculated based on the predicted performance change trend information, and the lifespan is estimated in combination with the initial lifespan of the hard drive. For example, if the design lifespan of a hard drive is 10,000 erases and writes when it leaves the factory, and the current erase and write times have reached 6,000 times, and the model predicts that the average erase and write times in the next 30 days will be 500 times, then it can be inferred that the remaining lifespan of the hard drive is approximately 8 months. In addition, by analyzing the degradation patterns in historical data, the lifespan prediction can be further revised. For example, if a hard drive degrades slowly in the early stages of use and accelerates in the later stages, this trend can be taken into account in the calculation, and a nonlinear model can be used to estimate the lifespan. Finally, the system generates performance evaluation results.

[0039] In one embodiment, if Figure 2 As shown, in step S20, the SSD operation information is analyzed abnormally by a deep learning model to obtain abnormal information of the SSD, and the abnormal information of the SSD is corrected to obtain corrected operation information, including: S201: Encode and decode the solid-state hard disk operation information through a deep learning model, calculate the reconstruction error, and if the reconstruction error exceeds a predetermined error threshold, determine that the solid-state hard disk operation information contains solid-state hard disk abnormal information.

[0040] Specifically, the operation information of the solid-state drive is input into a deep learning model. Specifically, this model uses an autoencoder architecture. The autoencoder encodes the input data into a lower-dimensional representation, and then decodes it to restore the output that is as similar as possible to the original data. The goal of model training is to minimize the reconstruction error between the input and the output. For example, common error measurement methods include mean square error (MSE) or mean absolute error (MAE). After the model is trained, new operation information is input and encoded and decoded through the autoencoder to obtain the reconstruction error between the decoding result and the original input. When the reconstruction error is greater than the predetermined error threshold, it means that the model has a large deviation when reconstructing this part of the data, which may be because this part of the data contains abnormal information. At this time, it is determined that this part of the solid-state drive operation information is abnormal information. For example, if the temperature value or the number of writes of a hard disk deviates significantly from the normal range, the model may not be able to accurately reconstruct these values, resulting in a large reconstruction error, and thus being judged as abnormal.

[0041] S202: Correct the abnormal information based on the generative adversarial network in the deep learning model, generate correction information based on the generator of the generative adversarial network, and compare the corrected information with the standard information through the discriminator of the generative adversarial network. If the distribution difference between the corrected information and the standard information is less than a predetermined difference threshold, replace the abnormal information of the solid-state drive with the corrected information to obtain the corrected operation information.

[0042] Specifically, first, the generator in the generative adversarial network (GAN) generates correction information based on the existing normal operation information to repair the abnormal information. The generator generates false information similar to the real information by learning the distribution pattern of normal information. For example, if the number of writes to the hard disk is abnormal, the generator can generate a reasonable correction value based on historical information and other normal parameters. Then, the generated correction information is passed to the discriminator, whose task is to judge the difference between the generated information and the real information. By calculating the distribution difference between the generated information and the standard information (real information), common measurement methods include Kullback-Leibler divergence or Wasserstein distance. If the distribution difference between the generated information and the standard information is less than the predetermined difference threshold, the generated correction information is considered to be true enough and can be used to replace the abnormal information to complete the correction of the abnormal information. Through this process, the model can repair the information that is judged to be abnormal, and ensure that the corrected information is consistent with the real information in statistical characteristics and distribution. The corrected operation information will replace the original abnormal information.

[0043] In one embodiment, if Figure 3 As shown, in step S201, the operation information of the solid state drive is encoded and decoded by a deep learning model, and a reconstruction error is calculated, including: S2011: The reconstruction error is obtained based on the following formula: , where E is the reconstruction error, N is the total number of samples of SSD operation information, and X i is the original input of the i-th sample, Y i is the predicted output of the i-th sample, w i is the sample weight, and p is the order of error calculation.

[0044] Specifically, in this formula, the reconstruction error E is used to measure the model's prediction accuracy of the SSD operation information. First, the original input X of each sample i is calculated. i and the predicted output Y i The difference between the two is weighted, where the weight w iUsed to adjust the contribution of each sample to the final error. Then, the absolute error to the power of p is used to measure the size of the difference, where p is the order of error calculation. Usually p=2 represents the mean square error (MSE), or p=1 represents the mean absolute error (MAE). The weighted errors of all samples are summed and divided by the total number of samples N to calculate the final average reconstruction error, which reflects the deviation of the overall prediction. For example, suppose that in a certain period of time, the system collects N=5 SSD operating parameters, which represent the temperature changes of the SSD at different time points. The original input data (real temperature) is X={40, 42, 45, 47, 50}, the predicted output data (temperature predicted by the model) is Y={39, 43, 44, 48, 51}, the sample weight w={1, 1, 1, 1, 1}, and the error calculation order p=2 (that is, the mean square error is used). The final reconstruction error E=1.

[0045] In one embodiment, if Figure 4 As shown, in step S40, multiple performance features are screened based on the feature selection method to obtain a target feature set required for performance evaluation, including: S401: Calculate the correlation of multiple performance features, select performance features with correlation coefficients lower than a discrimination threshold, and obtain a first feature set.

[0046] Specifically, first, all the collected performance characteristics are subjected to pairwise correlation calculation. The correlation calculation is performed using the Pearson correlation coefficient, and the Pearson correlation coefficient formula is as follows: , where A i and B i Represents the data values ​​of two performance features, A and B are the means of features A and B respectively. The calculation result ranges from -1 to 1. If the value is close to 1 or -1, it means that the two features have a strong linear relationship and one of the features should be removed. If the correlation is low (for example, less than 0.5), it means that the two features are almost independent and can be retained. For example, suppose that in the calculated correlation matrix, the correlation between feature A and feature B is 0.9, which means that the two features are highly correlated, so we only retain one of them.

[0047] S402: Calculate the variance of the first feature set, select performance features whose variance is greater than a predetermined variance threshold, and obtain a second feature set.

[0048] Specifically, the variance is calculated using the following formula: , where D irepresents the i-th sample value of the feature, D is the mean of the feature, and N is the total number of samples. If the variance of a feature is very small, it means that its variation in all samples is very small and cannot effectively reflect the performance difference of the hard disk. For example, if the variance of feature A is 0.02 and the variance of feature B is 10, then the variation of feature A is very limited and may not provide effective information for performance evaluation. Therefore, we will remove the features with smaller variance and only retain the features with larger variance that can effectively distinguish the performance of different hard disks to form the second feature set.

[0049] S403: Input the second feature set into the L1 regularization model, select performance features that contribute greatly to the solid state drive performance evaluation, and obtain a target feature set required for performance evaluation.

[0050] Specifically, the second feature set (i.e., the features obtained after correlation and variance screening) is input into the L1 regularization model. Assuming that the second feature set contains multiple performance features, such as F1 (continuous read and write speed), F2 (random read and write speed), F3 (number of erases), F4 (operating temperature fluctuations), and F5 (current consumption of the solid-state drive), these features are used as input variables to predict the remaining life or performance status of the solid-state drive. The L1 regularization model is based on the following formula: , where Z k The performance evaluation target value (such as health or remaining life) of the kth sample, F k，m is the mth feature of the kth sample, β m is the regression coefficient corresponding to the feature, and α is the regularization strength parameter, which controls the sparsity of the feature. The core idea of ​​L1 regularization is to impose sparsity constraints to automatically eliminate features with small contributions. When the α value is large, the model screens features more strictly and more features will be eliminated; when the α value is small, the model retains more features, but may increase redundant information. The appropriate α value is selected through cross-validation so that the model only retains the features that contribute the most to the hard disk performance prediction. If the regression coefficients of β4 (temperature fluctuation) and β5 (current consumption) are compressed to zero during the optimization process, it means that these two features contribute little to the hard disk performance evaluation and should be eliminated. On the contrary, if the regression coefficients of β1 (continuous read and write speed) and β3 (number of erases) and β2 (random read and write speed) are large, these features are retained in the final model. After L1 regularization, the feature set that is finally retained is the target feature set required for performance evaluation. These features have high predictive power for the performance evaluation of solid-state drives, while eliminating redundant features and improving the computational efficiency of the model. For example, suppose there are 5 original input features. After L1 regularization screening, 3 important features are finally retained: {F1, F2, F3}.

[0051] In one embodiment, if Figure 5As shown, in step S403, the second feature set is input into the L1 regularization model, and the performance features that contribute greatly to the performance evaluation of the solid state drive are selected to obtain the target feature set required for the performance evaluation, including: S4031: Input the second feature set into the L1 regularization model, introduce an L1 norm constraint on each feature, calculate the weight coefficient of each performance feature, and obtain the L1 regularization coefficient of each performance feature.

[0052] Specifically, the second feature set {P1, P2, ..., P M} as input, where P m represents the mth performance feature. For each sample n, the model’s goal is to predict the remaining life Y of the hard disk. n Or health score. Use L1 regularized regression model for feature selection, and optimize the objective function as follows: , where Y n is the target variable of the nth sample, P n，m is the mth feature of the nth sample, β m is the regression coefficient of the mth feature, i.e., the L1 regularization coefficient, and λ is the regularization strength parameter, which determines the strictness of L1 regularization on feature selection. The L1 regularization coefficient β of each feature is calculated by optimizing through gradient descent or coordinate descent. m If λ is set properly, the regression coefficients of some features will be compressed to zero, indicating that these features contribute less to the model, and other features will be assigned different weights. Finally, the calculated L1 regularization coefficient set is: {β1, β2, ..., β M}.

[0053] S4032: Sort the L1 regularization coefficient of each performance feature, select the performance features whose absolute value of the regularization coefficient is greater than a predetermined coefficient threshold, and obtain a preliminary target feature set.

[0054] Specifically, the calculated L1 regularization coefficient set {β1, β2, ..., β M} are sorted in descending order to ensure that the most important features are at the front. For example, suppose the calculated L1 regularization coefficients are as follows: β1=0.65, β2=0.02, β3=0.48, β4=0.00, β5=0.31, after sorting: β1=0.65, β3=0.48, β5=0.31, β2=0.02, β4=0.00, set a predetermined coefficient threshold T β , only retain β m ≥T β The features of the target are obtained by 初 ={P1, P3, P5}.

[0055] S4033: Remove redundant features from the preliminary target feature set, and remove features with high correlation through correlation analysis between features to obtain the target feature set required for performance evaluation.

[0056] Specifically, all features in the preliminary target feature set are compared to analyze the similarity between the numerical changes of two features. If the change trends of two features are close, it is considered that there is a high similarity between the two features. The features with higher similarity are screened to determine the features that are more important for performance evaluation, and the features with relatively smaller impact are eliminated. In the screening process, based on the performance of historical data, it is determined which features can more accurately reflect the status of the solid-state drive under different load environments. For features that change at the same time and have similar trends, the features with more stable data distribution and greater evaluation contribution are retained first, and the features with less impact are deleted. After the screening is completed, the target feature set required for performance evaluation is finally obtained.

[0057] In one embodiment, if Figure 6 As shown, in step S50, the target feature set required for performance evaluation is analyzed by the time series prediction model to obtain performance change trend information under different loads, including: S501: Input the target feature set required for performance evaluation into the time series prediction model, calculate the performance change trend within each time step based on the time series relationship in the time series prediction model, and obtain performance change trend information.

[0058] Specifically, first, the target feature set (such as read and write speed, storage capacity, temperature, IOPS, etc.) extracted from the SSD operation information is input into the time series prediction model. The time series prediction model adopts a deep learning architecture based on LSTM (Long Short-Term Memory Network) or GRU (Gated Recurrent Unit), which can effectively capture long-term dependencies in time series data. By inputting the target feature set into the model in chronological order, the model learns from past performance data and predicts future performance changes based on time series relationships. For example, the model can predict the trend of these performance indicators in the next few time steps by comparing the changes in indicators such as write frequency and read and write latency of the SSD over the past period of time. The model calculates the performance change trend in each time step and finally outputs a trend information about the performance change in each time step. This information is used to evaluate the performance and degradation trend of the SSD in different time periods.

[0059] S502: Perform difference analysis on the performance change trend information, identify the change rules of performance fluctuations under different loads, and obtain the performance change trend information under different loads.

[0060] Specifically, the performance change trend information output by the time series prediction model is first processed through the difference analysis method. Specifically, the difference analysis can distinguish and compare the performance change trend information based on different load conditions (for example, low load, medium load, and high load). Statistical analysis methods, such as analysis of variance (ANOVA) or clustering-based methods (such as K-means clustering), are used to divide the performance fluctuations under different load conditions. The analysis can focus on the fluctuation range, change rate, and abnormal points of the performance indicators to identify different performance change patterns of the solid-state drive under different loads. For example, under high load conditions, the read and write delays of the hard disk may increase significantly, while under low loads, the delay changes are small and stable. Through such difference analysis, the performance change law of the solid-state drive under different load conditions can be described more accurately, thereby providing a basis for further performance evaluation. Finally, the performance change trend information under different loads is obtained.

[0061] In one embodiment, if Figure 7 As shown, in step S60, the remaining life and performance degradation of the solid state drive are predicted based on the performance change trend information to obtain a performance evaluation result, including: S601: Calculate the remaining life of the solid state drive under different loads based on the following formula according to the performance change trend information: , where L remaining is the remaining life, L0 is the initial life, R current is the current performance degradation value, R max is the performance degradation value, β is the degradation acceleration coefficient, γ is the load influence coefficient, and f load is the load factor.

[0062] Specifically, based on the performance change trend information obtained from the time series prediction model, the performance degradation value R at the current time point is extracted. current。 Performance degradation value R current It is usually calculated by comparing the current performance indicators of the SSD with the initial performance indicators. For example, the current degradation degree is estimated by using the deviation of performance indicators such as write speed and read / write latency from the initial state. Then, determine R max, which represents the maximum performance degradation value of the SSD, that is, the performance level of the hard disk when it reaches the ultimate failure (usually zero or a very low value). Next, by defining the degradation acceleration coefficient β, consider how the performance degradation of the SSD accelerates over time. β can be obtained by fitting historical performance degradation data, for example, the regression coefficient obtained by regression analysis method. The introduction of the degradation acceleration coefficient β takes into account the nonlinear trend of performance degradation of the SSD during use, and it can be adjusted according to different types of hard disks and usage environments. Then, the load influence coefficient γ is calculated, which reflects the degree of aggravation of the performance degradation of the SSD under different load conditions. For example, under high load conditions, frequent read and write operations will accelerate the performance degradation of the hard disk, so the value of γ may increase. Load factor f load It is calculated based on the current load situation. It is related to the current workload of the SSD and can be determined by measuring the IO operation frequency, data transfer volume and other indicators of the hard disk. Finally, the remaining life L is calculated using the above formula remaining The formula combines the initial life L0, the current degradation value R current , maximum degradation value R max , degradation acceleration coefficient β, load influence coefficient γ and load factor f load , by weighing these factors, we can predict the remaining life of the hard drive under the current workload.

[0063] S602: Based on the remaining life, the remaining life is corrected in combination with the working environment parameter information of the solid state drive to obtain a corrected remaining life.

[0064] Specifically, the current working environment parameter information of the solid-state drive is obtained. The environmental parameters include but are not limited to temperature, humidity, power supply stability, operating load, and data writing frequency, among which temperature and load are key factors affecting the life of the solid-state drive. For example, in a high-temperature environment, the storage unit of the solid-state drive is prone to charge leakage, resulting in a decrease in data retention capability, thereby accelerating performance degradation; under high-load conditions, continuous high-intensity read and write operations will accelerate the wear of the storage unit, so it is necessary to correct the calculated remaining life based on these environmental factors. Secondly, set the impact weights of the environmental parameters, quantify the degree of impact of different environmental factors on the life, and calculate the correction factor C env , C env It is obtained by weighted calculation of multiple environmental parameters. For example, setting the temperature influence coefficient α T 、Power supply fluctuation influence coefficient α V 、Operation load influence coefficient α L Etc., calculate the normalized values ​​of the corresponding environmental parameters respectively (for example, use the maximum and minimum normalization method to make data of different dimensions comparable), and then use the formula: , where T is the temperature deviation, V is the power supply fluctuation amplitude, and L is the load change rate. Each influencing factor α is set according to experimental or empirical data to ensure that it is within a reasonable range. env The value is usually between 0 and 1. If the environmental factors are bad, C env The smaller the value of , the greater the correction to the remaining life. Then, based on the calculated environmental correction factor C env The remaining life L remaining Correction is made, and the correction formula is as follows: L adjusted =L remaining ×C env This formula indicates that if the environmental impact is small (C env Close to 1), then the corrected life span L adjusted With the original life L remaining Close, if the environmental impact is large (C env For example, if the remaining life is 1000 hours, but the current ambient temperature is high and the power supply fluctuates frequently, the C env After calculation, it is 0.75, and the corrected life becomes 750 hours to reflect the additional impact of the environment on the hard disk life. Finally, the corrected remaining life L adjusted It will serve as the final evaluation result.

[0065] S603: Predicting the performance degradation of the solid state drive according to the corrected remaining lifespan, and obtaining a performance evaluation result.

[0066] Specifically, based on the corrected remaining life, the performance change trend of the solid-state drive is analyzed to predict possible performance degradation in the future. Performance degradation is mainly manifested as a decrease in multiple key indicators, including but not limited to a decrease in sequential read and write speed, a decrease in random read and write speed, an increase in write amplification factor, an increase in IO latency, an increase in the proportion of bad blocks, etc. By analyzing the change trends of these performance parameters, the performance degradation of the solid-state drive in the future can be predicted, and a degradation model can be established in combination with historical data to identify the performance degradation characteristics at different stages. Then, based on the corrected remaining life, the degradation of the solid-state drive is divided into different stages to more intuitively evaluate the health status of the hard drive. For example, in the early stage of life, the performance may remain stable and the degradation rate is slow, while in the middle and late stages of life, the performance degradation rate may accelerate and eventually enter an unusable state. In the prediction process, the performance changes that may occur in each stage can be inferred in combination with the use intensity and operating environment of the solid-state drive. For example, in a high-load environment, random read and write performance may decline earlier, while in a low-load environment, the solid-state drive may maintain stable performance for a longer time. Then, based on the performance degradation characteristics of different life stages, possible abnormal situations are further analyzed. For example, if the revised remaining life prediction result is shorter, but the current performance indicators remain at a high level, it may mean that the SSD is about to enter the accelerated degradation stage, and the user needs to make data backup and replacement plans in advance; on the contrary, if the revised remaining life is longer, but some performance parameters have already fluctuated abnormally, it is necessary to check whether there are short-term abnormalities caused by external environmental factors, such as excessive temperature, unstable power supply, etc. Finally, the performance evaluation result is generated by combining the revised remaining life and the predicted performance degradation.

[0067] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0068] In one embodiment, a solid state drive performance evaluation device is provided, and the solid state drive performance evaluation device corresponds one-to-one to the solid state drive performance evaluation method in the above embodiment. Figure 8 As shown, the solid state drive performance evaluation device includes an operation information acquisition module, an abnormal information correction module, a performance feature extraction module, a feature screening module, a performance trend analysis module and a life prediction module. The detailed description of each functional module is as follows: An operation information acquisition module is used to obtain the operation information of the solid state drive; The abnormal information correction module is used to perform abnormal analysis on the operation information of the solid-state drive through a deep learning model to obtain abnormal information of the solid-state drive, and to correct and interpolate the abnormal information of the solid-state drive to obtain corrected operation information; A performance feature extraction module, used to extract features from the corrected operation information to obtain multiple performance features of the solid state drive; A feature screening module is used to screen multiple performance features based on a feature selection method to obtain a target feature set required for performance evaluation; The performance trend analysis module is used to analyze the target feature set required for performance evaluation through the time series prediction model to obtain the performance change trend information under different loads; The life prediction module is used to predict the remaining life and performance degradation of the solid-state drive based on the performance change trend information to obtain the performance evaluation result.

[0069] Optionally, the abnormal information correction module includes: The anomaly detection and identification submodule is used to encode and decode the SSD operation information through a deep learning model, calculate the reconstruction error, and if the reconstruction error exceeds a predetermined error threshold, determine that the SSD operation information contains SSD abnormal information; The abnormal correction submodule is used to correct the abnormal information based on the generative adversarial network in the deep learning model, generate correction information based on the generator of the generative adversarial network, and compare the correction information with the standard information through the discriminator of the generative adversarial network. If the distribution difference between the correction information and the standard information is less than a predetermined difference threshold, the correction information is used to replace the abnormal information of the solid-state drive to obtain the corrected operation information.

[0070] Optionally, the anomaly detection and identification submodule includes: The reconstruction error calculation unit is used to obtain the reconstruction error based on the following formula: , where E is the reconstruction error, N is the total number of samples of SSD operation information, and X i is the original input of the i-th sample, Y i is the predicted output of the i-th sample, w i is the sample weight, and p is the order of error calculation.

[0071] Optional, feature screening modules include: A feature correlation screening submodule is used to calculate the correlation of multiple performance features, select performance features with correlation coefficients lower than a discrimination threshold, and obtain a first feature set; A feature variance screening submodule, used to perform variance calculation on the first feature set, select performance features whose variance is greater than a predetermined variance threshold, and obtain a second feature set; The L1 regularization feature optimization submodule is used to input the second feature set into the L1 regularization model, select the performance features that contribute greatly to the performance evaluation of the solid-state drive, and obtain the target feature set required for the performance evaluation.

[0072] Optionally, the L1 regularized feature optimization submodule includes: An L1 regularized feature weight calculation unit, used for inputting the second feature set into the L1 regularized model, calculating the weight coefficient of each performance feature by introducing an L1 norm constraint on each feature, and obtaining the L1 regularized coefficient of each performance feature; A feature weight screening unit is used to sort the L1 regularization coefficient of each performance feature, select the performance feature whose absolute value of the regularization coefficient is greater than a predetermined coefficient threshold, and obtain a preliminary target feature set; The redundant feature removal unit is used to remove redundant features from the preliminary target feature set, and remove features with high correlation through correlation analysis between features to obtain the target feature set required for performance evaluation.

[0073] Optional performance trend analysis module includes: The time series performance trend calculation submodule is used to input the target feature set required for performance evaluation into the time series prediction model, calculate the performance change trend within each time step based on the time series relationship in the time series prediction model, and obtain the performance change trend information; The load difference analysis submodule is used to perform difference analysis on the performance change trend information, identify the changing rules of performance fluctuations under different loads, and obtain the performance change trend information under different loads.

[0074] Optional, life prediction module includes: The remaining life calculation unit is used to calculate the remaining life of the solid-state drive under different loads based on the following formula according to the performance change trend information: , where L remaining is the remaining life, L0 is the initial life, R current is the current performance degradation value, R max is the performance degradation value, β is the degradation acceleration coefficient, γ is the load influence coefficient, and f load is the load factor; An environmental impact correction unit, used to correct the remaining life based on the remaining life in combination with the working environment parameter information of the solid state drive to obtain a corrected remaining life; The performance degradation prediction unit is used to predict the performance degradation of the solid state drive according to the corrected remaining life and obtain a performance evaluation result.

[0075] For the specific definition of the solid-state hard disk performance evaluation device, please refer to the definition of the solid-state hard disk performance evaluation method above, which will not be repeated here. Each module in the above-mentioned solid-state hard disk performance evaluation device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0076] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, 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 network 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 solid-state hard disk performance evaluation method is implemented.

[0077] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program: Get the SSD operation information; Perform abnormal analysis on the operation information of the solid-state drive through a deep learning model to obtain abnormal information of the solid-state drive, and correct and interpolate the abnormal information of the solid-state drive to obtain corrected operation information; Perform feature extraction on the corrected operation information to obtain multiple performance features of the solid state drive; Screen multiple performance features based on feature selection methods to obtain the target feature set required for performance evaluation; The target feature set required for performance evaluation is analyzed through the time series prediction model to obtain performance change trend information under different loads; Based on the performance change trend information, the remaining life and performance degradation of the solid-state drive are predicted to obtain the performance evaluation result.

[0078] 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: Get the SSD operation information; Perform abnormal analysis on the operation information of the solid-state drive through a deep learning model to obtain abnormal information of the solid-state drive, and correct and interpolate the abnormal information of the solid-state drive to obtain corrected operation information; Perform feature extraction on the corrected operation information to obtain multiple performance features of the solid state drive; Screen multiple performance features based on feature selection methods to obtain the target feature set required for performance evaluation; The target feature set required for performance evaluation is analyzed through the time series prediction model to obtain performance change trend information under different loads; Based on the performance change trend information, the remaining life and performance degradation of the solid-state drive are predicted to obtain the performance evaluation result.

[0079] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods 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 memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0080] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0081] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A solid state drive performance evaluation method, characterized in that: The solid state drive performance evaluation method comprises: Get the SSD operation information; Performing an abnormal analysis on the operation information of the solid-state hard disk through a deep learning model to obtain abnormal information of the solid-state hard disk, and correcting and interpolating the abnormal information of the solid-state hard disk to obtain corrected operation information; Performing feature extraction on the corrected operation information to obtain multiple performance features of the solid state drive; Screening the multiple performance features based on a feature selection method to obtain a target feature set required for performance evaluation; Analyzing the target feature set required for the performance evaluation through a time series prediction model to obtain performance change trend information under different loads; Based on the performance change trend information, the remaining life and performance degradation of the solid state drive are predicted to obtain a performance evaluation result.

2. The solid state drive performance evaluation method according to claim 1, characterized in that: The performing abnormal analysis on the operation information of the solid state drive by the deep learning model to obtain abnormal information of the solid state drive, and correcting the abnormal information of the solid state drive to obtain corrected operation information includes: Encoding and decoding the solid state drive operation information by using the deep learning model, calculating a reconstruction error, and if the reconstruction error exceeds a predetermined error threshold, determining that the solid state drive operation information contains the solid state drive abnormal information; The abnormal information is corrected based on the generative adversarial network in the deep learning model, the correction information is generated based on the generator of the generative adversarial network, and the corrected information is compared with the standard information through the discriminator of the generative adversarial network. If the distribution difference between the corrected information and the standard information is less than a predetermined difference threshold, the solid-state hard disk abnormal information is replaced by the correction information to obtain the corrected operation information.

3. The solid state drive performance evaluation method according to claim 2, characterized in that: The encoding and decoding of the solid state drive operation information by the deep learning model and calculating the reconstruction error include: The reconstruction error is obtained based on the following formula: , where E is the reconstruction error, N is the total number of samples of the SSD operation information, and X i is the original input of the i-th sample, the Y i is the predicted output of the i-th sample, the w i is the sample weight, and p is the order of error calculation.

4. The solid state drive performance evaluation method according to claim 1, characterized in that: The method for selecting the multiple performance features based on the feature selection method is used to screen the multiple performance features to obtain a target feature set required for performance evaluation, including: Calculate the correlation of the multiple performance features, and select performance features whose correlation coefficients are lower than a discrimination threshold to obtain a first feature set; Performing variance calculation on the first feature set, selecting performance features whose variance is greater than a predetermined variance threshold, and obtaining a second feature set; The second feature set is input into the L1 regularization model, and the performance features that contribute greatly to the solid state drive performance evaluation are selected to obtain the target feature set required for the performance evaluation.

5. The solid state drive performance evaluation method according to claim 4, characterized in that: The step of inputting the second feature set into the L1 regularization model, selecting performance features that contribute greatly to the solid state drive performance evaluation, and obtaining the target feature set required for the performance evaluation includes: Inputting the second feature set into the L1 regularization model, calculating the weight coefficient of each performance feature by introducing an L1 norm constraint on each feature, and obtaining the L1 regularization coefficient of each performance feature; Sorting the L1 regularization coefficient of each performance feature, selecting the performance feature whose absolute value of the regularization coefficient is greater than a predetermined coefficient threshold, and obtaining a preliminary target feature set; Redundant features are removed from the preliminary target feature set, and features with high correlation are removed through correlation analysis between features to obtain the target feature set required for the performance evaluation.

6. The solid state drive performance evaluation method according to claim 1, characterized in that: The target feature set required for the performance evaluation is analyzed by the time series prediction model to obtain performance change trend information under different loads, including: Inputting the target feature set required for the performance evaluation into the time series prediction model, calculating the performance change trend within each time step based on the time series relationship in the time series prediction model, and obtaining performance change trend information; Perform difference analysis on the performance change trend information, identify the change rules of performance fluctuations under different loads, and obtain the performance change trend information under different loads.

7. The solid state drive performance evaluation method according to claim 1, characterized in that: The predicting the remaining life and performance degradation of the solid state drive according to the performance change trend information to obtain a performance evaluation result includes: The remaining life of the solid state drive under different loads is calculated based on the following formula using the performance change trend information: , wherein the L remaining is the remaining life, L0 is the initial life, and R current is the current performance degradation value, the R max is the performance degradation value, β is the degradation acceleration coefficient, γ is the load influence coefficient, and f load is the load factor; Based on the remaining life, the remaining life is corrected in combination with the working environment parameter information of the solid state drive to obtain a corrected remaining life; The performance degradation of the solid state drive is predicted based on the corrected remaining life to obtain the performance evaluation result.

8. A solid state drive performance evaluation device, characterized in that: The solid state drive performance evaluation device comprises: An operation information acquisition module is used to obtain the operation information of the solid state drive; An abnormal information correction module is used to perform abnormal analysis on the operation information of the solid-state hard disk through a deep learning model to obtain abnormal information of the solid-state hard disk, and to correct and interpolate the abnormal information of the solid-state hard disk to obtain corrected operation information; A performance feature extraction module, used to extract features from the corrected operation information to obtain multiple performance features of the solid state drive; A feature screening module, used to screen the multiple performance features based on a feature selection method to obtain a target feature set required for performance evaluation; A performance trend analysis module is used to analyze the target feature set required for the performance evaluation through a time series prediction model to obtain performance change trend information under different loads; The life prediction module is used to predict the remaining life and performance degradation of the solid state drive according to the performance change trend information to obtain a performance evaluation result.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the solid state drive performance evaluation method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the solid state drive performance evaluation method according to any one of claims 1 to 7 are implemented.

Images