Solid state disk read-write rate analysis method and system

Abstract

The invention provides a solid state disk read-write rate analysis method and system, and relates to the technical field of solid state disks, and the method comprises the steps: collecting the state information and read-write rate of a solid state disk, analyzing the influence index of the read-write rate of the solid state disk through an extension analytic hierarchy process, and calculating the read-write rate of the solid state disk through a fuzzy close degree calculation method; preliminarily predicting the read-write rate of the solid state disk; performing joint simulation on the solid state disk based on a digital twinning technology, and constructing a read-write rate compensation model of the solid state disk in combination with a recurrent neural network; and performing error correction on the preliminary prediction result of the read-write rate of the solid state disk by using the read-write rate compensation model to obtain the final read-write rate of the solid state disk. The prediction accuracy of the read-write rate of the solid state disk can be improved, and powerful support can be provided for performance evaluation, system optimization and hardware selection of the solid state disk.

Description

Technical Field

[0001] The present invention relates to the technical field of solid state hard disks, and in particular to a method and system for analyzing the read and write rates of solid state hard disks. Background Art

[0002] With the rapid development and widespread popularity of computer technologies such as the Internet, cloud computing, and the Internet of Things, unprecedented amounts of data have been generated in our daily work and life, which urgently require efficient processing and storage solutions. The explosive growth of data, the increasing attention paid by users to data security, and the continuous innovation of information technology have all put forward more stringent requirements on the performance of storage systems.

[0003] Solid-state drive (SSD), as a new type of hard disk based on solid-state electronic storage chip array, has become a leader in the storage field due to its structure consisting of a control unit and a solid-state storage unit (such as NAND flash memory chip), the stable nature of flash memory, and the widespread application of NAND flash memory particles. Read and write performance is a key indicator to measure the quality of solid-state drives, which is directly related to data processing efficiency and user experience.

[0004] However, the read and write rates of SSDs are not static, but are affected by a combination of factors. Factors such as temperature, load, total bytes written (TBW), and hard drive health all affect the performance of SSDs in their own unique ways. These factors are not only diverse in dimension, but also intricately interrelated, making the behavior of SSDs vary greatly under different working conditions. Traditional hard drive read and write rate prediction methods, such as linear regression or simple statistical models, often have difficulty capturing this complexity. They ignore the dynamic changes of hard drives in different environments, resulting in prediction results that are often not accurate enough in practical applications.

[0005] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0006] In view of this, the present invention provides a solid state drive read and write rate analysis method and system to solve the above-mentioned problems.

[0007] In order to solve the above problems, the specific technical solutions adopted by the present invention are as follows: According to one aspect of the present invention, a method for analyzing the read and write rates of a solid state drive is provided, comprising: S1. Collect the status information and read / write speed of the solid state drive, use the extension hierarchical analysis method to analyze the influencing indicators of the solid state drive read / write speed, and make a preliminary prediction of the read / write speed of the solid state drive through the fuzzy proximity calculation method; S2. Based on digital twin technology, conduct joint simulation on the solid-state drive, and construct a read / write rate compensation model for the solid-state drive in combination with a recurrent neural network; S3. Use the read / write rate compensation model to correct the error of the preliminary prediction result of the solid-state drive read / write rate, and obtain the final solid-state drive read / write rate.

[0008] Preferably, the collecting the state information and read / write rate of the solid-state drive, analyzing the influencing indicators of the solid-state drive read / write rate by using the extension analytic hierarchy process, and preliminarily predicting the read / write rate of the solid-state drive by using the fuzzy closeness calculation method includes: S11. Take the state information of the solid-state drive as different influencing indicators of the solid-state drive read / write rate, and construct an extension interval number judgment matrix by comparing the indicators pairwise; S12. Based on the extension interval number judgment matrix, use the extension analytic hierarchy process to calculate the initial weights of each influencing indicator, and correct the initial weights by using the order relation analysis method to obtain the final indicator weights; S13. Based on the read / write rate of the solid-state drive, use data fitting technology to determine the fitting relationship between each influencing indicator and the solid-state drive read / write rate, and in combination with the final indicator weights, use the fuzzy closeness calculation method to preliminarily predict the read / write rate of the solid-state drive.

[0009] Preferably, the based on the extension interval number judgment matrix, using the extension analytic hierarchy process to calculate the initial weights of each influencing indicator, and correcting the initial weights by using the order relation analysis method to obtain the final indicator weights includes: S121. Conduct a consistency index test on the extension interval number judgment matrix, and for the extension interval number judgment matrix that passes the consistency test, calculate the normalized eigenvector corresponding to its largest eigenvalue to obtain the normalized weight vector; S122. According to the normalized weight vector, form a weight interval vector that satisfies the consistency condition, and calculate the single sorting weight of each influencing indicator corresponding to its upper-level indicator according to the weight interval vector; S123. Calculate the group decision-making average index weight vector according to the single sorting weight to obtain the initial weight; S124. Use the order relation analysis method to determine the importance ranking of the subjective weights, calculate the relative importance degree ratio, and correct the initial weights according to the importance ranking of the subjective weights and the relative importance degree ratio to obtain the final indicator weights.

[0010] Preferably, the based on the read / write rate of the solid-state drive, using data fitting technology to determine the fitting relationship between each influencing indicator and the solid-state drive read / write rate, and in combination with the final indicator weights, using the fuzzy closeness calculation method to preliminarily predict the read / write rate of the solid-state drive includes: S131. Preprocess the read / write rate of the solid-state drive, where the preprocessing includes: removing noise, filling missing values, and normalizing; S132. Use non-linear fitting to fit each influencing index and the read / write rate to obtain the fitting relationship between each index and the read / write rate of the solid-state drive; S133. Establish a training sample based on the fitting relationship between each index and the read / write rate of the solid-state drive, calculate the closeness between the training sample and the prediction sample in combination with the final index weight, and make a preliminary prediction of the read / write rate by the weighted average method.

[0011] Preferably, the step of establishing a training sample based on the fitting relationship between each index and the read / write rate of the solid-state drive, calculating the closeness between the training sample and the prediction sample in combination with the final index weight, and making a preliminary prediction of the read / write rate by the weighted average method includes: S1331. Use the preprocessed read / write rate as the output of the sample, and use the influencing indexes of the read / write rate as the input of the sample to form a training sample matrix; S1332. Obtain the prediction sample of the read / write rate of the solid-state drive to be predicted, and calculate the closeness between the training sample and the prediction sample using the closeness calculation formula according to the final index weight; S1333. Sort the obtained closeness, select a preset number of training samples according to the sorting result, and predict the read / write rate of the solid-state drive to be measured by the weighted average method to obtain a preliminary prediction result.

[0012] Preferably, the closeness calculation formula is: ; ; ; In the formula, H ( T , T ′) represents the closeness, T i represents the i th training sample, T ′ represents the sample to be predicted, T i ⊕ T ′ represents the union operation of fuzzy sets, represents the intersection operation of fuzzy sets, Y represents the set of read / write rates corresponding to all influencing indexes, y ki represents the i rd read / write rate corresponding to the k th influencing index in the y′ kIndicates the read / write rate corresponding to the k th influencing indicator in the sample to be predicted, w k Indicates the k final indicator weight of the th influencing indicator.

[0013] Preferably, the method for jointly simulating a solid-state drive based on digital twin technology and constructing a read / write rate compensation model for the solid-state drive by combining a recurrent neural network includes: S21. Using digital twin technology, construct a virtual model of the solid-state drive, and perform joint simulation under different conditions using the virtual model of the solid-state drive to obtain simulated read / write rate data; S22. Combine the theoretical read / write rate data and the simulated read / write rate data of the solid-state drive, and construct a data-driven read / write rate compensation model based on a convolutional neural network with a fusion attention mechanism.

[0014] Preferably, the method for combining the theoretical read / write rate data and the simulated read / write rate data of the solid-state drive and constructing a data-driven read / write rate compensation model based on a convolutional neural network with a fusion attention mechanism includes: S221. Based on the simulated read / write rate data of the solid-state drive, determine the theoretical read / write rate data of the solid-state drive under the same conditions, and calculate the difference between the simulated read / write rate data and the theoretical read / write rate data to obtain a read / write rate error; S222. Through a convolutional neural network, use the read / write rate error between the simulated read / write rate data and the theoretical read / write rate data to extract local features, and use a channel attention mechanism to weight the extracted local features to generate a weighted feature map; S223. According to the weighted feature map, construct a data-driven read / write rate compensation model through a fully connected layer.

[0015] Preferably, the method for using a convolutional neural network to extract local features using the read / write rate error between the simulated read / write rate data and the theoretical read / write rate data and using a channel attention mechanism to weight the extracted local features to generate a weighted feature map includes: S2221. Input the read / write rate error into the convolutional neural network, and extract local features from the input read / write rate error data through convolutional operations; S2222. Perform a non-linear mapping on the local features after convolutional operations through an activation function to obtain a feature map after non-linear mapping; S2223. Use a channel attention mechanism to weight the feature map after non-linear mapping to obtain a weighted feature map.

[0016] According to another aspect of the present invention, there is provided a solid-state drive read / write rate analysis system, including: A preliminary prediction module, which is used to collect the status information and read / write rate of the solid-state drive, analyze the influencing indicators of the solid-state drive's read / write rate using the extension analytic hierarchy process, and perform a preliminary prediction of the solid-state drive's read / write rate through the fuzzy closeness calculation method; A compensation model construction module, which is used to perform joint simulation on the solid-state drive based on digital twin technology and construct a read / write rate compensation model for the solid-state drive in combination with a recurrent neural network; A read / write rate correction module, which is used to correct the error of the preliminary prediction result of the solid-state drive's read / write rate using the read / write rate compensation model to obtain the final read / write rate of the solid-state drive.

[0017] The beneficial effects of the present invention are as follows: 1. The present invention combines a variety of advanced technologies and methods, and has strong adaptability and scalability. With the continuous development of solid-state drive technology and the change of the hardware environment, the accuracy and effectiveness of the method can be maintained by updating the digital twin model, adjusting the parameters of the recurrent neural network, etc. It can not only improve the prediction accuracy of the solid-state drive's read / write rate, but also provide strong support for the performance evaluation, system optimization and hardware selection of the solid-state drive.

[0018] 2. The present invention uses the extension analytic hierarchy process to calculate the initial weights of each influencing indicator, and corrects the initial weights through the order relation analysis method to obtain more reasonable final indicator weights. The data fitting technology is used to determine the fitting relationship between each influencing indicator and the solid-state drive's read / write rate, which can accurately reflect the internal relationship between the indicator and the read / write rate. The fuzzy closeness calculation method is used to perform a preliminary prediction of the solid-state drive's read / write rate, taking into account the fuzziness and similarity between samples. By calculating the closeness between the training samples and the prediction samples, and selecting a preset number of training samples for weighted average prediction, the prediction result is closer to the actual situation.

[0019] 3. The present invention uses digital twin technology to construct a virtual model of the solid-state drive, which can realize the real-time simulation and emulation of the solid-state drive's status. By integrating the attention mechanism, the convolutional neural network can pay more attention to the features that have a greater impact on the read / write rate, improving the sensitivity and accuracy of the model. The introduction of the attention mechanism enables the model to dynamically adjust the attention to different features, enhancing the adaptability and robustness of the model. By calculating the difference between the simulated read / write rate data and the theoretical read / write rate data, the read / write rate error is obtained as the input of the compensation model. The data-driven compensation model can make full use of the existing data to improve the prediction accuracy and reliability. Description of the Drawings

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for use in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings. In the accompanying drawings: Figure 1 is a flowchart of a method for analyzing the read / write rate of a solid-state drive according to an embodiment of the present invention; Figure 2 is a schematic block diagram of a system for analyzing the read / write rate of a solid-state drive according to an embodiment of the present invention.

[0021] In the figure: 1. Preliminary prediction module; 2. Compensation model construction module; 3. Read / write rate correction module. Detailed implementation manners

[0022] In order to enable those skilled in the art of the present technology to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only some embodiments of this application, rather than all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts should fall within the protection scope of this application.

[0023] According to an embodiment of the present invention, a method and system for analyzing the read / write rate of a solid-state drive are provided.

[0024] Now, the present invention will be further described in conjunction with the accompanying drawings and specific implementation manners. As Figure 1 shown, according to an embodiment of an embodiment of the present invention, a method for analyzing the read / write rate of a solid-state drive is provided, including: S1. Collect the status information and read / write rate of the solid-state drive, analyze the influencing indicators of the read / write rate of the solid-state drive using the extension analytic hierarchy process, and perform a preliminary prediction of the read / write rate of the solid-state drive through the fuzzy closeness calculation method; It should be noted that the status information of the solid-state drive includes many factors that affect its read / write rate. Common influencing indicators include: Temperature: The temperature of the solid-state drive usually affects its performance. Excessive temperature may cause a speed reduction.

[0025] Remaining available space: The remaining space of the solid-state drive affects the read / write rate. The less the space, the lower the write rate may be.

[0026] Block Erase Count (Total Bytes Written): This metric reflects the wear level of the solid-state drive. The more severe the wear, the more likely the performance will degrade.

[0027] Drive Health Status: Solid-state drives with poor health may exhibit lower read and write speeds.

[0028] Input / Output Operations Per Second (IOPS): An important metric reflecting the hard drive's response performance, which affects the data transfer speed.

[0029] Power Status: Voltage fluctuations or unstable power supply may cause the hard drive's performance to degrade.

[0030] As a preferred implementation, collecting the status information and read / write speeds of the solid-state drive, analyzing the influencing metrics of the solid-state drive's read / write speeds using the extension analytic hierarchy process, and preliminarily predicting the read / write speeds of the solid-state drive through the fuzzy closeness calculation method includes: S11. Taking the status information of the solid-state drive as different influencing metrics for the solid-state drive's read / write speeds, and constructing an extension interval number judgment matrix by pairwise comparison of the metrics; It should be noted that through the experience of experts or existing data, pairwise comparison of these influencing metrics is carried out to construct an extension interval number judgment matrix.

[0031] Extension interval number is a method used to describe and handle uncertainty, especially suitable for dealing with the fuzziness and uncertainty in complex decision-making problems. It can describe the relationship between each pair of metrics during pairwise comparison, while taking into account the uncertainty of these relationships.

[0032] Experts or relevant data analysts give the relative importance scores of each pair of metrics according to the importance of the influencing metrics through pairwise comparison. The scoring can use the 1-9 ratio scale method (AHP scale method) or other similar methods. For each pair of influencing metrics, experts need to answer the following questions: Metric A Relative to metric B What is the degree of importance? For example, if A Relative to B is "extremely important", a very high score (e.g., 9) may be given; if it is "slightly important", a low score (e.g., 3 or 4) may be given.

[0033] Introduction of Extension Interval Numbers: In pairwise comparisons, uncertainty often occurs (for example, the priority of an indicator is higher in some situations and lower in others). To handle this uncertainty, extension interval numbers are adopted. For the comparison of each pair of indicators, the constructed judgment matrix is not a single value but an interval value, which can represent the range of changes in the importance of indicators. For example, the importance between indicators can be represented by an interval, such as [a, b], where a and b represent the minimum and maximum values of the comparison of this pair of indicators respectively, taking into account the fuzziness of subjective judgment.

[0034] Suppose there are three influencing indicators T 1 (temperature), T 2 (remaining space), and T 3 (total number of written bytes). After pairwise comparison, the obtained judgment matrix is as follows: ; Among them: The "1" in the first row and first column represents T the importance ratio of 1 and T 1 itself is equal.

[0035] The interval value "[2, 3]" in the first row and second column indicates that the expert believes T the importance degree of 1 compared to T 2 is between "2 and 3".

[0036] The interval value "[3, 4]" in the first row and third column indicates that the expert believes T the importance degree of 1 compared to T 3 is between "3 and 4".

[0037] S12. Based on the extension interval number judgment matrix, use the extension analytic hierarchy process to calculate the initial weights of each influencing indicator, and correct the initial weights through the order relation analysis method to obtain the final indicator weights; As a preferred implementation manner, the above-mentioned based on the extension interval number judgment matrix, using the extension analytic hierarchy process to calculate the initial weights of each influencing indicator, and correcting the initial weights through the order relation analysis method to obtain the final indicator weights includes: S121. Conduct a consistency index test on the extension interval number judgment matrix, and for the extension interval number judgment matrix that passes the consistency test, calculate the normalized eigenvector corresponding to its maximum eigenvalue to obtain the normalized weight vector; Specifically, the purpose of the consistency test is to ensure that the judgment results of the extension interval number judgment matrix are reasonable, that is, the relative importance ratios between the indicators in the matrix are consistent. In the traditional analytic hierarchy process, this test is carried out by calculating the consistency ratio ( CRIt is completed by... For the extension interval number judgment matrix, a similar method can also be used for consistency checking.

[0038] Calculate the consistency index CI , and through the maximum eigenvalue λ max Calculate the consistency ratio CR .

[0039] For the consistency check of the extension interval number matrix, a fuzzy consistency check method can be adopted, considering the fuzziness and uncertainty of the interval numbers. If the check passes, the matrix can be used for subsequent eigenvector calculations.

[0040] By solving the characteristic equation of the extension interval number judgment matrix, its maximum eigenvalue λ max and the corresponding eigenvector are obtained.

[0041] The steps to normalize the eigenvector are to divide each component of the eigenvector by the sum of the vector to ensure that the sum of all weights is 1. The finally obtained normalized eigenvector represents the preliminary weights of each influencing index.

[0042] In addition, the maximum eigenvalue λ max and the consistency index CI Calculation formulas: ; Among them, n represents the order of the judgment matrix (the number of influencing indicators).

[0043] S122. According to the normalized weight vector, form a weight interval vector that satisfies the consistency condition, and calculate the single sorting weight of each influencing index corresponding to its upper-level index according to the weight interval vector; It should be noted that the weight interval vector is formed by expanding the normalized weight vector to form the weight interval of each influencing index to represent uncertainty and fuzziness. Specifically, the weight interval vector expresses the range of the weight of each index by assigning an upper and lower bound (minimum and maximum value) to the weight of each index.

[0044] The single sorting weight is the relative importance of each influencing index calculated according to the weight interval of the index relative to its upper-level index. In the analytic hierarchy process, the weight of each lower-level index is usually determined according to the weight of its upper-level index. Specifically, it includes the following steps: Determine the upper-level indicators: Assuming there are multiple upper-level indicators, the weight interval vector will help determine the relative importance of each lower-level indicator with respect to the upper-level indicator. For example, the lower-level indicators affecting the read / write speed of a solid-state drive may be "temperature", "remaining space", and "total written bytes", and the sorting weights of these lower-level indicators with respect to the upper-level indicator "drive performance" need to be calculated.

[0045] Weighting of interval weights: Based on the weight interval vector of each indicator, the single-sorting weight of the lower-level indicators can be calculated through interval weighting. For example, assume the weights of the upper-level indicators of "temperature" and "remaining space" are w 1 = [0.45, 0.55] and w 2 = [0.25, 0.35], then the single-sorting weights of the lower-level indicators can be calculated according to these intervals.

[0046] Calculate the single-sorting weight: The calculation of the single-sorting weight is usually achieved by normalizing and weighting the weight interval vector. In the process of pairwise comparison, the obtained weight interval vector will affect the importance ranking of the lower-level indicators relative to the upper-level indicator. For a given weight interval, the single-sorting weight of the lower-level indicators can be calculated by the interval number weighting method.

[0047] S123. Calculate the group decision average index weight vector according to the single-sorting weight to obtain the initial weight; It should be noted that the group decision-making method obtains the final decision result by weighted averaging the judgment results of multiple experts. In this step, the group decision average weight vector represents the comprehensive result of all experts' opinions.

[0048] Calculate the group decision average weight of each influencing indicator, usually by weighted averaging the judgments of each expert on the indicator (such as the single-sorting weight).

[0049] The group decision average weight vector is the initial weight of all influencing indicators. These weights reflect the relative importance of each influencing indicator in the overall decision-making.

[0050] S124. Use the order relation analysis method to determine the importance ranking of the subjective weights, calculate the relative importance ratio, and correct the initial weight according to the importance ranking of the subjective weights and the relative importance ratio to obtain the final index weight.

[0051] It should be noted that the Ranking Analysis Method is a technique that helps decision-makers rank indicators by comparing the relative importance of each indicator. Through the Ranking Analysis Method, the importance ranking of each indicator can be effectively determined, and the ratio of the relative importance between indicators can be calculated. This is very useful for correcting the initial weights obtained from multiple experts or different data sources. Especially in the case of inconsistent weights, it can integrate different views and optimize the final indicator weights.

[0052] In the Ranking Analysis Method, the first step is to determine the importance ranking of each indicator relative to other indicators based on subjective judgment or expert opinion. Usually, experts will rank the indicators according to experience or historical data and give the relative importance of each indicator. For example, assume there are three indicators T 1, T 2, and T 3. The expert gives a ranking based on their influence on the read / write speed of the solid-state drive. For example: T 1 (temperature) > T 2 (remaining space) > T 3 (total written bytes); This means that the expert believes that temperature ( T 1) has the greatest impact on the read / write speed of the solid-state drive, followed by the remaining space ( T 2), and finally the total written bytes ( T 3).

[0053] In addition, after determining the ranking of the subjective weights, the next step is to calculate the ratio of the relative importance between each indicator. The ratio of relative importance describes the importance of one indicator relative to another. In the Ranking Analysis Method, the following method is usually used to calculate: If T 1 is more important than T 2, then set the ratio r 12 > 1, and the specific value can be determined according to the expert's opinion.

[0054] If T 1 and T 2 are equally important, then set the ratio r 12 = 1.

[0055] If T 1 is less important than T 2, then set the ratio r 12 < 1.

[0056] Then the example of the ratio of relative importance is: Assumed temperature T 1 relative to the remaining space T 2 has a ratio of 2:1 ( r 12 = 2), indicating that T the importance of 1 is T twice that of 2.

[0057] Assume the remaining space T 2 relative to the total number of written bytes T 3 has a ratio of 3:1 (i.e., r 23 = 3).

[0058] In addition, once the importance ranking and relative importance degree ratio of the subjective weights are obtained, these ratios can be used to correct the initial weights. The key to the correction process is to reallocate the weights of each index according to the ranking and ratio. The following is the correction method: Adjust the weight ratio: By calculating the relative importance degree ratio, the initial weights can be weighted and corrected. For example, if the initial weight vector is W init = w 1, w 2, w 3], and the ratio matrix is R , then the weights can be adjusted according to the ratio matrix.

[0059] Weighted average method: Use the weighted average method to correct the initial weights. The weighted average method is based on the relative importance degree ratio r ij to determine the final weights of each index. For example, assume the initial weights are W init =[0.5, 0.3, 0.2], and the relative importance ratio matrix is R , then the final weights of each index can be calculated by weighted average.

[0060] Through the weighted correction of the order relation analysis method, the final weights of each influencing index can be obtained, and these weights can more accurately reflect the actual importance of each index relative to the read / write speed of the solid-state drive.

[0061] For example: Assume the weights corrected by the order relation analysis method are: W final =[0.55, 0.35, 0.10]; Then it means that: the final weight of temperature ( T 1) is 0.55; the final weight of the remaining space ( T 2) is 0.35; the total number of written bytes ( TThe final weight of (3) is 0.10.

[0062] S13. Based on the read / write rate of the solid-state drive, use data fitting technology to determine the fitting relationship between each influencing index and the read / write rate of the solid-state drive, and combine the final index weights to preliminarily predict the read / write rate of the solid-state drive using the fuzzy closeness calculation method.

[0063] As a preferred embodiment, the step of using data fitting technology to determine the fitting relationship between each influencing index and the read / write rate of the solid-state drive, and combining the final index weights to preliminarily predict the read / write rate of the solid-state drive using the fuzzy closeness calculation method includes: S131. Preprocess the read / write rate of the solid-state drive, and the preprocessing includes: removing noise, filling missing values, and normalizing. It should be noted that the purpose of preprocessing is to ensure the quality and usability of the data. The read / write rate data is often affected by noise, missing values, or outliers, so it needs to be cleaned and processed.

[0064] Common preprocessing steps include: Removing noise: Remove outliers or extreme values from the read / write rate data.

[0065] Filling missing values: For missing read / write rate data, it can be filled by interpolation, mean filling method, or other methods.

[0066] Normalizing: Normalize the read / write rate data to make its range unified, usually standardizing the data between 0 and 1.

[0067] S132. Use non-linear fitting to fit each influencing index and the read / write rate to obtain the fitting relationship between each index and the read / write rate of the solid-state drive. Specifically, common non-linear fitting methods include: Quadratic polynomial fitting: Use a polynomial function to fit the relationship between the influencing index and the read / write rate.

[0068] Support vector regression (SVR): Perform non-linear fitting through the regression algorithm of the support vector machine.

[0069] Neural network fitting: Use a neural network model in deep learning to learn complex non-linear relationships.

[0070] For example, assume that a quadratic polynomial model is selected to fit the relationship between the read / write rate and the influencing index. After fitting, a model as follows is obtained: R =1.2 + 0.5 T 1 + 0.3 T2 + 0.1 T 3 + 0.02 T 1 2 + 0.05 T 2 2 + 0.01 T 3 2 ; Among them, a 0 = 1.2, a 1 = 0.5, a 2 = 0.3, a 3 = 0.1, … are coefficients obtained by fitting.

[0071] S133. Establish training samples based on the fitting relationship between each index and the read / write speed of the solid-state drive, calculate the closeness between the training samples and the prediction samples in combination with the final index weights, and perform a preliminary prediction on the read / write speed through the weighted average method.

[0072] As a preferred implementation manner, the establishment of training samples based on the fitting relationship between each index and the read / write speed of the solid-state drive, calculating the closeness between the training samples and the prediction samples in combination with the final index weights, and performing a preliminary prediction on the read / write speed through the weighted average method includes: S1331. Use the preprocessed read / write speed as the output of the sample, and use the influencing indicators of the read / write speed as the input of the sample to form a training sample matrix; In addition, combine each index (such as temperature, remaining space, total written bytes, etc.) that affects the read / write speed of the solid-state drive with the preprocessed read / write speed data to form a training sample matrix. Each row represents a sample, listing the corresponding influencing indicators and their corresponding read / write speeds.

[0073] Suppose there are 3 influencing indicators: temperature T 1, remaining space T 2, and total written bytes T 3, and the corresponding read / write speed R . The sample matrix X can be expressed as: ; Among them, T 1 (i) , T 2 (i) , T 3 (i) represents the influencing indicator value of the i th sample, R (i) represents the corresponding read / write speed.

[0074] S1332. Obtain the prediction samples of the read / write speed of the solid-state drive to be predicted, and calculate the closeness between the training samples and the prediction samples according to the final index weights using the closeness calculation formula; As a preferred implementation, the closeness calculation formula is: ; ; ; In the formula, H ( T , T ′) represents the closeness, T i represents the i th training sample, T ′ represents the sample to be predicted, T i ⊕ T ′ represents the union operation of fuzzy sets, represents the intersection operation of fuzzy sets, Y represents the set of read / write speeds corresponding to all influencing indicators, y ki represents the read / write speed corresponding to the i th influencing indicator in the k th training sample, y′ k represents the read / write speed corresponding to the k th influencing indicator in the sample to be predicted, w k represents the final index weight of the k th influencing indicator.

[0075] S1333. Sort the obtained closeness, and select a preset number of training samples according to the sorting result. Predict the read / write speed of the solid-state drive to be tested by the weighted average method to obtain a preliminary prediction result.

[0076] Specifically, sort according to the closeness between each calculated training sample and the sample to be predicted. The training samples with larger closeness are considered more representative and should have a greater weight in the final prediction. According to the sorting result, select the first m training samples with larger closeness as the basis for prediction. The more training samples are selected, the more accurate the prediction result usually is.

[0077] For the selected training samples, predict the read / write speed by the weighted average method. The weighted average method combines the closeness and read / write speed of each training sample to calculate the read / write speed of the solid-state drive to be predicted.

[0078] Among them, the weighted average method formula is: ; In the formula, R predict represents the preliminary predicted read / write rate, R i represents the i th training sample's read / write rate.

[0079] S2. Based on digital twin technology, conduct co-simulation on the solid-state drive, and construct a read / write rate compensation model for the solid-state drive in combination with a recurrent neural network; As a preferred implementation, the step of conducting co-simulation on the solid-state drive based on digital twin technology and constructing a read / write rate compensation model for the solid-state drive in combination with a recurrent neural network includes: S21. Use digital twin technology to construct a virtual model of the solid-state drive, and use the virtual model of the solid-state drive to conduct co-simulation under different conditions to obtain simulated read / write rate data; It should be noted that digital twin technology simulates the physical and behavioral characteristics of the hard drive by creating a virtual model of the solid-state drive (SSD). The virtual model can not only replicate the structure and working state of the hard drive, but also simulate the working conditions under different environments, including the impact of variables such as temperature, humidity, TBW (Total Bytes Written), and remaining space on the read / write rate.

[0080] Construct a digital twin model of the solid-state drive through advanced simulation tools (such as ANSYS, COMSOL, etc.). This model is based on the physical characteristics, working mechanism of the hard drive, and external environmental factors.

[0081] Conduct simulations under various conditions based on the digital twin model to simulate the performance of the solid-state drive under different working environments and operating conditions. The simulation can obtain different read / write rate data by adjusting environmental factors (such as temperature, load, hard drive health status, etc.). Through co-simulation, the read / write rate data of the solid-state drive under different environments and operating conditions can be obtained. These data will be used for subsequent comparison of theoretical and simulation data to calculate errors and provide training data for the compensation model.

[0082] S22. Combine the theoretical read / write rate data and simulated read / write rate data of the solid-state drive, and construct a data-driven read / write rate compensation model based on a convolutional neural network with a fusion attention mechanism.

[0083] As a preferred implementation, the step of combining the theoretical read / write rate data and simulated read / write rate data of the solid-state drive and constructing a data-driven read / write rate compensation model based on a convolutional neural network with a fusion attention mechanism includes: S221. Determine the theoretical read / write rate data of the solid-state drive under the same conditions based on the simulated read / write rate data of the solid-state drive, and calculate the difference between the simulated read / write rate data and the theoretical read / write rate data to obtain the read / write rate error. It should be noted that the theoretical read / write rate data is based on the design parameters, technical specifications of the solid-state drive, and test results under ideal conditions. Usually, these data can be obtained from the technical specification table provided by the solid-state drive manufacturer or a standard test environment.

[0084] Technical specification table: The maximum theoretical read / write rate of the solid-state drive is usually listed in the technical documentation provided by the hard drive manufacturer. For example, assume that the theoretical maximum read rate of a certain solid-state drive is 550 MB / s and the write rate is 500 MB / s.

[0085] Standard test conditions: The theoretical read / write rate is generally measured under specific standard conditions, such as when the temperature is 25 °C, the humidity is 40%, and the hard drive health condition is optimal.

[0086] For example, under ideal conditions, the theoretical read rate is 550 MB / s and the theoretical write rate is 500 MB / s.

[0087] In addition, in order to measure the difference between the simulated data and the theoretical data, the read / write rate error can be calculated, usually using the absolute error or relative error.

[0088] S222. Through the convolutional neural network, use the read / write rate error between the simulated read / write rate data and the theoretical read / write rate data to extract local features, and use the channel attention mechanism to weight the extracted local features to generate a weighted feature map. As a preferred embodiment, the step of using the convolutional neural network to use the read / write rate error between the simulated read / write rate data and the theoretical read / write rate data to extract local features, and using the channel attention mechanism to weight the extracted local features to generate a weighted feature map includes: S2221. Input the read / write rate error into the convolutional neural network, and extract local features from the input read / write rate error data through convolutional operations. It should be noted that the convolutional neural network will perform convolutional operations on these error data. The convolutional operation will scan the input data through different convolutional kernels (filters) to extract local features in the input data. Each convolutional kernel will detect specific local patterns or change trends.

[0089] Convolutional operation formula: Assume the input data is P , the convolutional kernel is W , and the output feature map is Q , the convolutional operation can be expressed as: ； Among them, represents the convolution operation, W is the convolution kernel, P is the input data.

[0090] Through the convolution operation, the model can extract local features from the error data, which may be the patterns of read / write rate errors under different environmental conditions. For example, the convolution kernel may identify the impact of temperature changes on read / write rate errors.

[0091] S2222. Perform a non-linear mapping on the local features after the convolution operation through an activation function to obtain a feature map after non-linear mapping; It should be noted that the role of the activation function is to introduce non-linear transformation, enabling the network to learn more complex patterns rather than simple linear relationships.

[0092] Common activation functions include ReLU (Rectified Linear Unit), Sigmoid, Tanh, etc.

[0093] ReLU is the most commonly used activation function, which changes all negative values to zero and keeps positive values unchanged. The formula is: ReLU( x ) = max(0, x ); Among them, x represents the value input to the activation function, usually the output of a node (neuron) in the neural network.

[0094] Through the ReLU activation function, perform a non-linear mapping on the local features after the convolution operation, thereby increasing the expression ability of the network. The feature map after non-linear mapping will better reflect the complex relationships in the error data.

[0095] S2223. Use the channel attention mechanism to weight the feature map after non-linear mapping to obtain a weighted feature map.

[0096] Specifically, the channel attention mechanism determines the importance of each channel by assigning different weights to different feature channels. It allows the network to focus on the most important feature channels in the prediction task.

[0097] After the convolution operation and the processing of the activation function, multiple feature maps are obtained. Each feature map represents certain specific local features. Through the channel attention mechanism, a weight is assigned to each channel (i.e., each feature map). These weights represent the importance of the feature in the compensation task.

[0098] After calculating the attention weights for each channel, multiply each channel in the original feature map to obtain the weighted feature map.

[0099] S223. Construct a data-driven read-write rate compensation model through a fully connected layer based on the weighted feature map.

[0100] Specifically, the weighted feature map is a tensor with multiple channels, representing the weighted results of different features. Before inputting into the fully connected layer, it is usually necessary to flatten the weighted feature map into a one-dimensional vector because the fully connected layer generally accepts one-dimensional input vectors.

[0101] For example, if the shape of the weighted feature map is ( H , W , C ), after flattening, it will become a one-dimensional vector with a size of H × W × C .

[0102] The fully connected layer transforms the input vector into an output result through a set of weights and biases. Each neuron in the fully connected layer is connected to all neurons in the previous layer, so each output is a weighted sum of all inputs.

[0103] Assume the size of the input vector is H × W × C , then the size of the weight matrix of the fully connected layer is ( H × W × C , G ), where G is the output dimension of the fully connected layer (i.e., the target dimension predicted by the model, usually a single read-write rate value).

[0104] Assume the input vector is C, the output of the fully connected layer is V, the weight is O, and the bias is b. Then the calculation formula of the fully connected layer is: ; where C is the flattened weighted feature map, O is the weight matrix, b is the bias vector, and V is the output of the fully connected layer, i.e., the compensated read-write rate.

[0105] S3. Use the read-write rate compensation model to correct the error of the preliminary prediction result of the solid-state drive read-write rate to obtain the final solid-state drive read-write rate.

[0106] Specifically, using the read-write rate compensation model to correct the error of the preliminary prediction result of the solid-state drive read-write rate to obtain the final solid-state drive read-write rate includes the following steps: Take the preliminary prediction result as the input and input it into the trained compensation model.

[0107] The model outputs the corrected predicted read / write rate. After error correction, a more accurate predicted result of the SSD read / write rate can be obtained.

[0108] Among them, after obtaining the corrected read / write rate, these values can be used for multiple purposes, such as: Evaluating the performance of SSD (Solid State Drive): Through the read / write rate after error correction, the actual performance of the solid state drive can be evaluated more accurately, avoiding deviations caused by factors such as environmental changes and the health status of the hard disk.

[0109] Optimizing system configuration: According to the predicted result of the corrected read / write rate, the system configuration can be optimized, such as selecting a suitable SSD model in the data center, adjusting the storage hierarchy, etc.

[0110] Selecting a suitable SSD model: The corrected performance prediction can help users make a more scientific choice among different SSD models, so as to meet the requirements of specific application scenarios.

[0111] As Figure 2 shown, according to an embodiment of the present invention, a solid state drive read / write rate analysis system is provided, including: A preliminary prediction module 1, configured to collect the status information and read / write rate of the solid state drive, analyze the influencing indicators of the solid state drive read / write rate by using the extension analytic hierarchy process, and perform a preliminary prediction on the read / write rate of the solid state drive through a fuzzy closeness calculation method; A compensation model construction module 2, configured to perform co-simulation on the solid state drive based on digital twin technology, and construct a read / write rate compensation model of the solid state drive in combination with a recurrent neural network; A read / write rate correction module 3, configured to correct the error of the preliminary prediction result of the solid state drive read / write rate by using the read / write rate compensation model to obtain the final solid state drive read / write rate.

[0112] In summary, by means of the above technical solutions of the present invention, the present invention combines a variety of advanced technologies and methods, and has strong adaptability and scalability. With the continuous development of solid-state drive technology and the change of the hardware environment, the accuracy and effectiveness of the method can be maintained by updating the digital twin model, adjusting the parameters of the recurrent neural network, etc. It can not only improve the prediction accuracy of the read / write rate of the solid-state drive, but also provide strong support for the performance evaluation, system optimization and hardware selection of the solid-state drive. The present invention uses the extension analytic hierarchy process to calculate the initial weights of each influencing index, and modifies the initial weights through the order relation analysis method to obtain more reasonable final index weights. The data fitting technology is used to determine the fitting relationship between each influencing index and the read / write rate of the solid-state drive, which can accurately reflect the internal relationship between the index and the read / write rate. The fuzzy closeness calculation method is used to preliminarily predict the read / write rate of the solid-state drive, taking into account the fuzziness and similarity between samples. By calculating the closeness between the training samples and the prediction samples, and selecting a preset number of training samples for weighted average prediction, the prediction result is closer to the actual situation. The present invention uses digital twin technology to construct a virtual model of the solid-state drive, which can realize the real-time simulation and emulation of the solid-state drive state. By integrating the attention mechanism, the convolutional neural network can pay more attention to the features that have a greater impact on the read / write rate, improving the sensitivity and accuracy of the model. The introduction of the attention mechanism enables the model to dynamically adjust the attention to different features, enhancing the adaptability and robustness of the model. By calculating the difference between the simulated read / write rate data and the theoretical read / write rate data, the read / write rate error is obtained as the input of the compensation model. The data-driven compensation model can make full use of the existing data to improve the prediction accuracy and reliability.

[0113] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program code.

[0114] The above specific embodiments have further elaborated on the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for analyzing the read and write speeds of a solid-state drive, characterized in that, Including: S1. Collect the status information and read / write rate of the solid-state drive, analyze the influencing indicators of the solid-state drive's read / write rate using the extension analytic hierarchy process, and conduct a preliminary prediction of the solid-state drive's read / write rate through the fuzzy closeness calculation method; S2. Based on digital twin technology, conduct a co-simulation of the solid-state drive, and construct a read / write rate compensation model for the solid-state drive in combination with a recurrent neural network; S3. Use the read / write rate compensation model to correct the error in the preliminary prediction result of the solid-state drive's read / write rate to obtain the final read / write rate of the solid-state drive; The S1 includes: Based on the read / write rate of the solid-state drive, use data fitting technology to determine the fitting relationship between each influencing indicator and the solid-state drive's read / write rate, and in combination with the final index weights, conduct a preliminary prediction of the solid-state drive's read / write rate through the fuzzy closeness calculation method; Among them, the conducting a preliminary prediction of the solid-state drive's read / write rate through the fuzzy closeness calculation method in combination with the final index weights includes: establishing a training sample based on the fitting relationship between each indicator and the solid-state drive's read / write rate, calculating the closeness between the training sample and the prediction sample in combination with the final index weights, and conducting a preliminary prediction of the read / write rate through the weighted average method.

2. The method for analyzing the read and write speeds of a solid-state drive according to claim 1, characterized in that, Before the conducting a preliminary prediction of the solid-state drive's read / write rate through using data fitting technology to determine the fitting relationship between each influencing indicator and the solid-state drive's read / write rate and in combination with the final index weights, it includes: Taking the status information of the solid-state drive as different influencing indicators of the solid-state drive's read / write rate, and constructing an extension interval number judgment matrix by comparing the indicators pairwise; Based on the extension interval number judgment matrix, use the extension analytic hierarchy process to calculate the initial weights of each influencing indicator, and correct the initial weights through the order relation analysis method to obtain the final index weights.

3. The method for analyzing the read and write speeds of a solid-state drive according to claim 2, characterized in that, The using the extension analytic hierarchy process to calculate the initial weights of each influencing indicator based on the extension interval number judgment matrix and correcting the initial weights through the order relation analysis method to obtain the final index weights includes: S121. Conduct a consistency index test on the extension interval number judgment matrix, and for the extension interval number judgment matrix that passes the consistency test, calculate the normalized eigenvector corresponding to its largest eigenvalue to obtain the normalized weight vector; S122. According to the normalized weight vector, form a weight interval vector that satisfies the consistency condition, and calculate the single-rank weight of each influencing indicator corresponding to its upper-level indicator according to the weight interval vector; S123. Calculate the group decision-making average index weight vector according to the single-rank weight to obtain the initial weights; S124. Use the order relation analysis method to determine the importance ranking of the subjective weights, calculate the relative importance degree ratio, and correct the initial weights according to the importance ranking of the subjective weights and the relative importance degree ratio to obtain the final index weights.

4. The method for analyzing the read and write speeds of a solid-state drive according to claim 2, characterized in that, The using data fitting technology to determine the fitting relationship between each influencing indicator and the solid-state drive's read / write rate based on the read / write rate of the solid-state drive includes: S131. Preprocess the read / write rate of the solid-state drive, where the preprocessing includes: removing noise, filling in missing values, and normalizing; S132. Use non-linear fitting to fit each influencing index and the read / write rate to obtain the fitting relationship between each index and the read / write rate of the solid-state drive.

5. The method for analyzing the read and write speeds of a solid-state drive according to claim 4, characterized in that, The establishment of training samples based on the fitting relationship between each index and the read / write rate of the solid-state drive, the calculation of the closeness between the training samples and the prediction samples in combination with the final index weights, and the preliminary prediction of the read / write rate by the weighted average method include: S1331. Use the preprocessed read / write rate as the output of the sample and the influencing indexes of the read / write rate as the input of the sample to form a training sample matrix; S1332. Obtain the prediction samples of the read / write rate of the solid-state drive to be predicted, and calculate the closeness between the training samples and the prediction samples using the closeness calculation formula according to the final index weights; S1333. Sort the obtained closeness and select a preset number of training samples according to the sorting result, and predict the read / write rate of the solid-state drive to be measured by the weighted average method to obtain a preliminary prediction result.

6. The method for analyzing the read and write speeds of a solid-state drive according to claim 5, characterized in that, The closeness calculation formula is: ； ； ； In the formula, H ( T , T ) represents the closeness degree, T i represents the i th training sample, T ′ represents the sample to be predicted, T i ⊕ T ′ represents the union operation of fuzzy sets, represents the intersection operation of fuzzy sets, Y represents the set of read-write rates corresponding to all influencing indicators, y ki represents the read-write rate corresponding to the i th influencing indicator in the k th training sample, y′ k represents the read-write rate corresponding to the k th influencing indicator in the sample to be predicted, w k represents the final index weight of the k th influencing indicator.

7. The method for analyzing the read and write speeds of a solid-state drive according to claim 1, characterized in that, Based on the digital twin technology, conduct joint simulation on the solid-state drive, and construct a read / write rate compensation model for the solid-state drive in combination with a recurrent neural network, including: S21. Use the digital twin technology to construct a virtual model of the solid-state drive, and conduct joint simulation under different conditions using the virtual model of the solid-state drive to obtain simulated read / write rate data; S22. Combine the theoretical read / write rate data and the simulated read / write rate data of the solid-state drive, and construct a data-driven read / write rate compensation model based on a convolutional neural network with a fused attention mechanism.

8. A method for analyzing the read and write speeds of a solid-state drive according to claim 7, characterized in that, The combination of the theoretical read / write rate data and the simulated read / write rate data of the solid-state drive, and the construction of a data-driven read / write rate compensation model based on a convolutional neural network with a fused attention mechanism include: S221. Based on the simulated read / write rate data of the solid-state drive, determine the theoretical read / write rate data of the solid-state drive under the same conditions, and calculate the difference between the simulated read / write rate data and the theoretical read / write rate data to obtain the read / write rate error; S222. Through the convolutional neural network, use the read / write rate error between the simulated read / write rate data and the theoretical read / write rate data to extract local features, and use the channel attention mechanism to weight the extracted local features to generate a weighted feature map; S223. According to the weighted feature map, construct a data-driven read / write rate compensation model through a fully connected layer.

9. A method for analyzing the read and write speeds of a solid-state drive according to claim 8, characterized in that, The process of using the convolutional neural network to extract local features using the read / write rate error between the simulated read / write rate data and the theoretical read / write rate data, and using the channel attention mechanism to weight the extracted local features to generate a weighted feature map includes: S2221. Input the read / write rate error into the convolutional neural network, and extract local features from the input read / write rate error data through convolutional operations; S2222. Perform non-linear mapping on the local features after convolutional operations through an activation function to obtain a non-linearly mapped feature map; S2223. Using the channel attention mechanism, perform weighted processing on the feature map after non-linear mapping to obtain a weighted feature map.

10. A solid-state drive read and write speed analysis system for implementing the solid-state drive read and write speed analysis method described in any one of claims 1-9, characterized in that, Including: A preliminary prediction module, which is used to collect the status information and read / write rate of the solid-state drive, analyze the influencing indicators of the solid-state drive's read / write rate using the extension analytic hierarchy process, and perform a preliminary prediction of the solid-state drive's read / write rate through the fuzzy closeness calculation method; A compensation model construction module, which is used to perform joint simulation on the solid-state drive based on digital twin technology and construct a read / write rate compensation model for the solid-state drive in combination with a recurrent neural network; A read / write rate correction module, which is used to correct the error in the preliminary prediction result of the solid-state drive's read / write rate using the read / write rate compensation model to obtain the final read / write rate of the solid-state drive.

Images