A data retention period prediction method and related components

Abstract

The present invention discloses a data retention period prediction method and related components. After determining N sampling storage units in a sample storage data area, firstly, based on the error situation of the minimum unit storage data of each first sampling storage unit, a first error confusion matrix corresponding to each first sampling storage unit is established. By performing machine learning operations on each first error confusion matrix, a data retention period prediction model is obtained. Thus, when it is desired to predict the data retention period of the storage data area to be tested, the data retention period prediction model is used to predict the data retention period of the storage data area to be tested. It can be seen that in this application, a data retention period prediction model is first established based on the sample storage data area, so that the data retention period of each storage data area to be tested can be predicted according to the data retention period prediction model, and the data stored in the storage data area to be tested is processed according to the predicted data retention period to avoid the loss of stored data.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method for predicting data retention period and related components. Background Art

[0002] Solid-state drives typically use NAND flash to store data. Writing and erasing data is essentially the process of charging and discharging the storage medium. Over time, previously written electrons in NAND flash particles can be lost, or they can be accidentally energized, altering the stored data and causing errors when reading the data. Therefore, data has a retention period (i.e., during which it can be successfully read using ECC (Error Checking and Correction) without a flash refresh). This retention period varies from particle to particle, and even from location to location on the same particle. Furthermore, data retention varies with usage; for example, oxidation of the materials used in NAND flash can shorten the retention period.

[0003] Due to data retention limitations in memory, data in flash memory products cannot be permanently stored once written. Conventional technology typically uses a preset data retention period to prevent data loss. Specifically, when the storage time of existing data reaches or is about to reach the preset data retention period, the data in that storage location is promptly refreshed or moved to a new storage location to refresh the data retention period. This prevents the risk of data read failures caused by further data errors due to electronic loss or injection in the data storage location.

[0004] However, in actual applications, if the preset data retention period is too short, data will be frequently moved and rewritten, increasing the workload and power consumption of the memory, thereby affecting the memory's performance. If the preset data retention period is set too long, the risk of data loss will also increase, affecting data security. Therefore, existing technologies often choose a shorter preset data retention period, ensuring data security even if it increases the workload and power consumption of the memory. However, even with the additional workload and power consumption, due to the defects such as the inability to dynamically and differentiate the preset data retention period, it is still impossible to avoid the abnormal situation where some data is lost before the preset data retention period expires.

[0005] In summary, how to accurately determine and predict the data retention period is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a data retention period prediction method and related components. First, a data retention period prediction model is established through a sample storage data area, so that the data retention period of each storage data area to be tested can be predicted according to the data retention period prediction model, and then the data stored in the storage data area to be tested is processed according to the predicted data retention period to avoid the loss of stored data.

[0007] To solve the above technical problems, the present invention provides a method for predicting data retention period, comprising:

[0008] Determine N first sampling storage units in the sample data storage area, where N is a positive integer;

[0009] Establishing a first error confusion matrix corresponding to each of the first sampling storage units based on an error condition of the minimum unit storage data of each of the first sampling storage units;

[0010] Performing machine learning on each of the first error confusion matrices to obtain a data retention period prediction model;

[0011] The data retention period prediction model is used to predict the data retention period of the data storage area to be tested.

[0012] Preferably, performing machine learning on each of the first error confusion matrices to obtain a data shelf life prediction model includes:

[0013] Establishing a machine learning model based on each of the first error confusion matrices;

[0014] The machine learning model is trained using each of the first error confusion matrices to obtain the data retention period prediction model.

[0015] Preferably, the machine learning model includes a regression model or a neural network model.

[0016] Preferably, the step of establishing a first error confusion matrix corresponding to each of the first sampling storage units based on an error condition of minimum unit storage data of each of the first sampling storage units includes:

[0017] error classification of possible data errors in minimum units based on a change relationship of minimum unit programming states of each of the first sampled memory cells;

[0018] counting errors of the minimum unit stored data in each of the first sampling storage units according to the error classification, so as to obtain the error statistics number of each type of error corresponding to each of the first sampling storage units;

[0019] The corresponding first error confusion matrix is ​​established based on the error statistics of each type of error corresponding to each first sampling storage unit.

[0020] Preferably, using the data retention period prediction model to predict the data retention period of the data storage area to be tested includes:

[0021] Determine M second sampling storage units in the data storage area to be tested, where M is a positive integer;

[0022] Establishing a second error confusion matrix corresponding to each second sampling storage unit based on an error condition of the minimum unit storage data of each second sampling storage unit;

[0023] The second error confusion matrix is ​​input into the data retention period prediction model for calculation to obtain the data retention period of the data storage area to be tested.

[0024] Preferably, the step of establishing a second error confusion matrix corresponding to each second sampling storage unit based on an error condition of minimum unit storage data of each second sampling storage unit includes:

[0025] performing error classification on data errors that may occur in minimum units based on a change relationship of minimum unit programming states of each of the second sampled memory cells;

[0026] Counting the error conditions of the minimum unit stored data in each of the second sampling storage units according to the error classification to obtain the error statistics number of each type of error corresponding to each of the second sampling storage units;

[0027] The corresponding second error confusion matrix is ​​established based on the error statistics of each type of error corresponding to each second sampling storage unit.

[0028] Preferably, after establishing the first error confusion matrix corresponding to each of the first sampling storage units based on the error condition of the minimum unit storage data of each of the first sampling storage units, the method further includes:

[0029] Normalizing the first error confusion matrix of each of the first sampling storage units to determine a third error confusion matrix;

[0030] Performing machine learning on each of the first error confusion matrices to obtain a data shelf life prediction model includes:

[0031] Machine learning is performed on each of the third error confusion matrices to obtain a data retention period prediction model.

[0032] Preferably, after normalizing the first error confusion matrix of each of the first sampling storage units to determine a third error confusion matrix, the method further includes:

[0033] Performing nonlinear mapping processing on each element in the third error confusion matrix to generate a fourth error confusion matrix;

[0034] Performing machine learning on each of the third error confusion matrices to obtain a data shelf life prediction model includes:

[0035] Machine learning is performed on each of the fourth error confusion matrices to obtain a data retention period prediction model.

[0036] Preferably, before using the data retention period prediction model to predict the data retention period of the data storage area to be tested, the method further includes:

[0037] Determining whether an inspection instruction or a read instruction for the data storage area to be tested is received;

[0038] If so, the process proceeds to the step of using the data retention period prediction model to predict the data retention period of the data storage area to be tested.

[0039] Preferably, after predicting the data retention period of the data storage area to be tested using the data retention period prediction model, the method further includes:

[0040] Determining whether the data storage period of the tested data storage area is less than a preset shelf life threshold;

[0041] If so, the data stored in the data storage area to be tested is rewritten or transferred to a spare storage location.

[0042] To solve the above technical problems, the present invention provides a data retention period prediction system, comprising:

[0043] a determining unit, configured to determine N first sampling storage units in the sample data storage area, where N is a positive integer;

[0044] a matrix establishing unit, configured to establish a first error confusion matrix corresponding to each of the first sampling storage units based on an error condition of the minimum unit storage data of each of the first sampling storage units;

[0045] a model acquisition unit, configured to perform machine learning on each of the first error confusion matrices to obtain a data retention period prediction model;

[0046] The prediction unit is used to predict the data retention period of the data storage area to be tested by using the data retention period prediction model.

[0047] To solve the above technical problems, the present invention provides a device for predicting data retention period, comprising:

[0048] Memory for storing computer programs;

[0049] A processor is configured to implement the steps of the method for predicting the data retention period as described above when executing the computer program.

[0050] The present application provides a data retention period prediction method and related components. After determining N sampling storage units in a sample storage data area, firstly, based on the error situation of the minimum unit storage data of each first sampling storage unit, a first error confusion matrix corresponding to each first sampling storage unit is established. By performing machine learning operations on each first error confusion matrix, a data retention period prediction model is obtained. Thus, when it is desired to predict the data retention period of the storage data area to be tested, the data retention period prediction model is used to predict the data retention period of the storage data area to be tested. It can be seen that in the present application, a data retention period prediction model is first established through the sample storage data area, so that the data retention period of each storage data area to be tested can be predicted according to the data retention period prediction model, and the data stored in the storage data area to be tested is processed according to the predicted data retention period to avoid the loss of stored data. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the prior art and the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 A schematic flow chart of a method for predicting data retention period provided by the present invention;

[0053] Figure 2 A schematic diagram of the data retention period in the prior art;

[0054] Figure 3 A schematic diagram of a TLC memory cell programming state definition provided by the present invention;

[0055] Figure 4 A schematic diagram of a first error confusion matrix provided by the present invention;

[0056] Figure 5 A schematic diagram of the machine learning model provided by the present invention when it is an SVR model;

[0057] Figure 6 A schematic diagram of a CNN model used as the machine learning model provided by the present invention;

[0058] Figure 7 A schematic diagram of a third error confusion matrix provided by the present invention;

[0059] Figure 8 A schematic diagram of a fourth error confusion matrix provided by the present invention;

[0060] Figure 9 A flow chart of another embodiment of a method for predicting data retention period provided by the present invention;

[0061] Figure 10 A schematic diagram of the structure of a data retention period prediction system provided by the present invention;

[0062] Figure 11 This is a structural diagram of a data retention period prediction device provided by the present invention. DETAILED DESCRIPTION

[0063] The core of the present invention is to provide a data retention period prediction method and related components. First, a data retention period prediction model is established through a sample storage data area, so that the data retention period of each storage data area to be tested can be predicted according to the data retention period prediction model, and then the data stored in the storage data area to be tested is processed according to the predicted data retention period to avoid the loss of stored data.

[0064] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0065] The applicant has taken into account that the data stored in each data storage area in the memory has its own corresponding data retention period. When the data is stored in the data storage area until it approaches the end of its own data retention period, the probability of error or loss of the data will increase accordingly. Therefore, it is necessary to predetermine the data retention period of the data so that the data can be rewritten or stored in a new storage location before the data storage time reaches its data retention period to extend the reliable storage period of the data.

[0066] However, in the prior art, the data retention period of each storage data area is usually set manually. The setting of the data retention period includes two implementation methods. The first is the statistical threshold method, that is, a relatively safe time threshold is set based on multiple experimental statistics as a unified given data retention period for the storage data areas of all storage locations. Obviously, the set time threshold cannot be too long (for example, consumer-grade flash memory is generally set to 180 days, while enterprise-grade flash memory is set to 90 days or even shorter). For storage data areas with shorter actual data retention periods, a shorter given data retention period can avoid data loss. For storage data areas with longer actual data retention periods, data transfer or rewriting and other processing will be frequently performed, increasing the workload of the memory. Moreover, if one wants to ensure a confidence level of 99.5% or even higher in this way, then at least 99.5% of the storage data areas will have a large difference between the actual data retention period and the given data retention period. Figure 2 As shown, Figure 2 Schematic diagram of data retention period in the prior art. In addition, a fixed given data retention period cannot guarantee that the data retention period will change accordingly as the storage medium of the memory changes with the usage status.

[0067] Another approach is to use a relational expression, such as a first-order or second-order one, to express the relationship between data retention and wear or usage time. While this can partially alleviate the problem of fixed data retention periods that cannot change with environmental changes, it still does not address the differences between different storage media and cannot adapt to the varying data retention periods of different data storage areas.

[0068] Please refer to Figure 1 , Figure 1 A flow chart of a method for predicting data retention period provided by the present invention includes:

[0069] S11: Determine N first sampling storage units in the sample data storage area, where N is a positive integer;

[0070] In order to solve the above technical problems, the present application predicts the data retention period of the storage data area. Specifically, N first sampling storage units in the sample storage data area are first determined. It should be noted that, as a preferred embodiment, when determining multiple first sampling storage units in the sample storage data area, it is possible but not limited to setting the first and last two layers of storage data pages or data blocks in the sample storage data block as the first sampling storage unit.

[0071] For a storage data region containing multiple storage cells (e.g., a storage data region containing multiple storage data blocks, each storage data block including multiple storage data pages, and each storage data page being a storage cell), an error confusion matrix for the multiple storage cells can be obtained and then concatenated into a new matrix or three-dimensional array using several of the error confusion matrices. For example, a current 96-layer 3DTLC (Trinary-Level Cell) flash memory is typically constructed by stacking two 48-layer stacks. For a 48-layer stack, data errors are often more pronounced in the first and last layers. Therefore, the storage data pages in the 1st, 48th, 49th, and 96th layers of the storage data region can be selected as sampling storage cells. The corresponding at least four error confusion matrices can be concatenated into at least a new 16×16 matrix or an 8×8×4 3D array to perform machine learning or data retention period prediction. Of course, more layers of storage data pages or other different layers of data pages can also be selected as the first sampling storage cell, which will not be discussed further here.

[0072] S12: establishing a first error confusion matrix corresponding to each first sampling storage unit based on an error condition of the minimum unit storage data of each first sampling storage unit;

[0073] The applicant took into account that each storage unit includes multiple minimum units, each minimum unit stores different data, and each minimum unit may cause errors in the stored data due to the loss or injection of electrons. Since the data stored in the storage unit is in binary form of 0 / 1, the stored data can be divided into two categories when it changes (i.e., 0 / 1 inversion), i.e., 0 is inverted to 1, or 1 is inverted to 0. Therefore, the number of erroneous bits of data stored in the minimum unit, the number of 0 inversions to 1, or the number of 1 inversions to 0 can be used as data error indicators to respond to the management of data stored in the storage data area. For example, RBER (Raw Bit Error Rate) is usually used as an important indicator to evaluate the severity of data errors in the storage data area.

[0074] However, in recent years, with the development of flash memory, from SLC (Single-Level Cell) to MLC (Multi-Level Cell), TLC, and even QLC (quadruple-Level Cell), each minimum unit (cell) of the data storage area can store multiple bits of information. For example, taking 3D TLC flash memory as an example, each minimum unit of the data storage area can store 3 bits of information, such as Figure 3 As shown, Figure 3This is a schematic diagram of the definition of the programming state of a TLC memory cell provided by the present invention. In TLC, the range of the charge carried by a memory cell is divided into 8 segments, that is, the smallest unit can have 8 programming states, and different storage charge ranges represent different programming states, namely Figure 3 ER, A, B, C, D, E, F and G correspond to the programming state of a minimum unit respectively, and each programming state corresponds to a code from 3'b000 to 3'b111, so a minimum unit of a TLC flash memory can store 3 bits of information.

[0075] When each minimum unit can store multiple bits of information, if the number of 0 / 1 reversals is still used as the main or only data error evaluation indicator, the error measurement will be too rough and cannot reflect the differences in data errors, and may even lead to unscientific evaluation conclusions. For example, Figure 3 The programming state of the smallest cell in the data structure is D, meaning the corresponding stored data is 3'b010, which differs from the C state (3'b000), E state (3'b011), and A state (3'b110) by only one bit. However, when a data error occurs, the probability of the D state changing to the C state or the D state changing to the E state is much higher than the D state changing to the A state. In fact, the probability of the D state changing to the C state is also higher than the D state changing to the E state (because the D state to C state is a loss of electrons, while the D state to E state requires additional electron injection). In other words, although there is only a single bit of 0 / 1 flip when changing from D state to C state, D state to E state, or D state to A state, the D state to A state error is clearly the most serious of the three types of errors, followed by the D state to E state error, and the D state to C state error is the least serious.

[0076] Therefore, in order to fully reflect the differences in data errors in new flash memory media such as MLC or TLC, this application can use the changes in the storage power of the smallest unit to represent the transition between different programming states instead of the original simple error classification of 0 / 1 reversal. For example, in the above example, the change in storage power state from D state to C state, E state or A state represents three different error categories, which are named DC, DE and DA errors respectively. It is not difficult to deduce that according to the above definition, there are 56 types of errors in TLC. For another example, using the same principle, it can be defined that there are a total of 12 types of errors in MLC, which will not be repeated here.

[0077] Based on this, when a storage unit is collected as a first sampling storage unit and its first error confusion matrix is ​​determined, the first error confusion matrix can be established based on the number of errors corresponding to each type of error that may occur in the first sampling storage unit. Here, TLC is still used as an example, and the first error confusion matrix is ​​as follows:

[0078]

[0079] Among them, e ij represents the number of errors in transitioning from programming state i to programming state j. The diagonal elements can be set to 0, that is, e ii =0, here the programming state ER is abbreviated as R, please refer to Figure 4 , Figure 4 A schematic diagram of a first error confusion matrix provided by the present invention, Figure 4 Taking TLC as an example, the first error confusion matrix is ​​presented in tabular form. The 56 error types present in TLC can be divided into three categories: 1-bit, 2-bit, and 3-bit errors. Here, 1, 2, and 3 represent the number of 0 / 1 inversions within the 3-bit data within a minimum unit. The severity of these three categories of 1-bit, 2-bit, and 3-bit errors increases in order of severity, as indicated by the gradually darkening colors in the figure. As can be seen, when generating the first error confusion matrix for each first sampling storage unit based on the error conditions of the minimum unit stored data within each first sampling storage unit, if the minimum unit stored data within that first sampling storage unit exhibits the corresponding error type, the eigenvalues ​​of the first error confusion matrix will change accordingly.

[0080] S13: Performing machine learning on each first error confusion matrix to obtain a data retention period prediction model;

[0081] S14: Predicting the data retention period of the data storage area to be tested using the data retention period prediction model.

[0082] Machine learning is performed on the first error confusion matrix of each determined first sampling storage unit to obtain a data retention period prediction model. After the data retention period prediction model is determined, regardless of whether the storage data area to be tested is the sample storage data area, its data retention period can be predicted by the data retention period prediction model to determine the predicted data retention period of the storage data block to be tested. The machine learning model can be, but is not limited to, an SVR (support vector regression) model, a CNN (convolutional neural network) model, an FCN (fully connected network) model, a decision tree, a random forest or a logistic regression model. The SVR model is to convert the first error confusion matrix into a vector eigenvalue by establishing and solving an objective function that satisfies the constraints, and then input it into the SVR model. For example, a Lagrangian function is established, a slack variable is introduced to solve the weight, and the offset is solved by transforming the dual problem to obtain an objective function of the form f(x) = w·x+b, thereby predicting the data retention period.

[0083] Please refer to Figure 5 and Figure 6 , Figure 5 This is a schematic diagram of the machine learning model provided by the present invention when it is an SVR model. Figure 6 This is a schematic diagram of a CNN model used in the machine learning model provided by the present invention.

[0084] As for the CNN model, taking the LeNet-5 type as an example, when predicting the data retention period of the storage data area to be tested, there are a total of 7 layers, among which the first and third layers are convolution layers. The purpose of the convolution operation is to extract the different eigenvalues ​​of the input first error confusion matrix to ensure parameter sharing; the second and fourth layers are pooling layers, which maintain the robustness to rotation and translation after the eigenvalue input and reduce the calculation dimension; the fifth and sixth layers are fully connected layers, which are used to map the distributed feature representation calculated previously to a low-dimensional vector space; the seventh layer is an activation function layer, which is used to map the low-dimensional vector space feature representation calculated previously to the sample label space, for example, using the Softmax function, to achieve the prediction of the data retention period.

[0085] The method of the present invention can be divided into two stages in actual use: the first stage is sample sampling, model building, model training, and parameter tuning; the second stage is to use the trained and optimized data retention period prediction model to make predictions and provide a reference basis for storage management strategies (the data retention period of other storage data areas can be predicted based on the trained data retention period prediction model).

[0086] In addition, in order to further ensure the effectiveness of data retention period prediction model training, it is necessary to control the proportion of each type of samples in the process of determining the first sampling storage unit, and pay attention to balanced distribution. For example, according to the actual remaining data retention period of the first sampling storage unit, the samples are divided into multiple categories with 10 days as a segment to ensure that the proportion of samples in each category is balanced.

[0087] As a preferred embodiment, when performing machine learning on each first error confusion matrix to obtain a data retention period prediction model, it can be but is not limited to establishing a machine learning model based on each first error confusion matrix; and using each first error confusion matrix to train the machine learning model to obtain a data retention period prediction model.

[0088] When determining the data retention period prediction model, a machine learning model can be established based on each first error confusion matrix, and then the machine learning model can be trained so that the machine learning model can deduce the error conditions of the data stored in the storage data area to be tested, thereby obtaining a data retention period prediction model, which is convenient for predicting the data retention period of different storage data areas to be tested.

[0089] As a preferred embodiment, when using a data retention period prediction model to predict the data retention period of the storage data area to be tested, it can be specifically, but not limited to, determining M second sampling storage units in the storage data area to be tested, where M is a positive integer; establishing a second error confusion matrix corresponding to each second sampling storage unit based on the error situation of the minimum unit storage data of each second sampling storage unit; and inputting the second error confusion matrix into the data retention period prediction model for calculation to obtain the data retention period of the storage data area to be tested.

[0090] When determining the data retention period of a test data storage region, multiple second sampling storage cells are identified within the test data storage region, and second error confusion matrices corresponding to each second sampling storage cell are obtained. Several of these error confusion matrices are then combined to form a new matrix or three-dimensional array. This newly combined matrix or three-dimensional array is then input into a data retention period prediction model. This allows the data situation of the test data storage region to be deduced to determine the data retention period of the test data storage region. For the method of combining several error confusion matrices into a new matrix or three-dimensional array, please refer to the above-described method embodiment, and the present invention will not be further described herein.

[0091] As a preferred embodiment, establishing a second error confusion matrix corresponding to each second sampling storage unit based on the error conditions of the minimum unit storage data of each second sampling storage unit can specifically, but is not limited to, classifying possible data errors in the minimum unit based on the change relationship of the minimum unit programming state of each second sampling storage unit; counting the error conditions of the minimum unit storage data in each second sampling storage unit according to the error classification to obtain the error statistics of each type of error corresponding to each second sampling storage unit; and establishing a corresponding second error confusion matrix based on the error statistics of each type of error corresponding to each second sampling storage unit.

[0092] In this embodiment, when predicting the data retention period of the data storage area to be tested based on the data retention period prediction model, the specific process of determining the second error confusion matrix is ​​the same as the process of determining the first error confusion matrix, and this application will not repeat it here.

[0093] In summary, in this application, a data retention period prediction model is first established through a sample storage data area, so that the data retention period of each storage data area to be tested can be predicted according to the data retention period prediction model, and the data stored in the storage data area to be tested can be processed according to the predicted data retention period to avoid the loss of stored data.

[0094] Based on the above embodiment:

[0095] As a preferred embodiment, after establishing the first error confusion matrix corresponding to each of the first sampling storage units based on the error condition of the minimum unit storage data of each of the first sampling storage units, the method further includes:

[0096] Normalizing the first error confusion matrix of each first sampling storage unit to determine a third error confusion matrix;

[0097] Performing machine learning on each of the first error confusion matrices to obtain a data shelf life prediction model includes:

[0098] Machine learning is performed on each of the third error confusion matrices to obtain a data retention period prediction model.

[0099] The applicant considers that the cardinality of the first error confusion matrix statistics is different due to the different sizes of the first sampling storage units. Therefore, the present application adopts a data normalization method to normalize the values ​​of the elements in the first error confusion matrix to a unified value space, namely:

[0100]

[0101] Where τ is the number of the smallest unit in the first sampling storage unit. E' is the third error confusion matrix, e' ijare the eigenvalues ​​in the third error confusion matrix.

[0102] Please refer to Figure 7 , Figure 7 This is a schematic diagram of the third error confusion matrix provided by the present invention. It should be noted that the eigenvalues ​​in the third error confusion matrix are Figure 7 The color of the corresponding position indicates the error ratio. Dark color indicates a high error ratio, and vice versa. Figure 4 There is no corresponding relationship between the three major categories of wrong color definitions.

[0103] As a preferred embodiment, after normalizing the first error confusion matrix of each of the first sampling storage units to determine a third error confusion matrix, the method further includes:

[0104] Performing nonlinear mapping processing on each element in the third error confusion matrix to generate a fourth error confusion matrix;

[0105] Performing machine learning on each of the third error confusion matrices to obtain a data shelf life prediction model includes:

[0106] Machine learning is performed on each of the fourth error confusion matrices to obtain a data retention period prediction model.

[0107] In addition, the applicant considers that when the cardinality of the first sampling storage unit is large, the eigenvalues ​​in the third error confusion matrix are much less than 1. Since the data retention period prediction model established later uses gradient descent to obtain the optimal parameter values, in order to prevent the problem of gradient disappearance when the model parameters are adjusted or to facilitate the accurate transmission of the gradient, the present application performs nonlinear mapping processing on the third error confusion matrix, namely:

[0108] E″=[e″ ij ] 8×8 , i, j∈{R,A,B,C,D,E,F,G};

[0109]

[0110] Where E' is the fourth error confusion matrix, e' ij is each eigenvalue in the fourth error confusion matrix, α is the amplification factor, which is usually 1. However, when some pure decimal expressions are not accurate (such as fp16, using a 16-bit half-precision floating point format) or integer data formats are calculated, it is often necessary to further project it to a non-pure decimal value interval. In this case, α can be 128, 256, etc., and this application does not limit this.

[0111] Please refer to Figure 8 , Figure 8This is a schematic diagram of the fourth error confusion matrix provided by the present invention. It should be noted that the eigenvalues ​​in the fourth error confusion matrix are Figure 8 The color of the corresponding position indicates the error ratio. Dark color indicates a high error ratio, and vice versa. Figure 4 There is no corresponding relationship between the three major categories of wrong color definitions.

[0112] As a preferred embodiment, before using the data retention period prediction model to predict the data retention period of the data storage area to be tested, the method further includes:

[0113] Determine whether an inspection instruction or a read instruction for the storage data area to be tested is received;

[0114] If so, the process proceeds to the step of using the data retention period prediction model to predict the data retention period of the data storage area to be tested.

[0115] In this embodiment, before determining the data retention period of the storage data area to be tested, it is necessary to first determine whether an inspection instruction or a read instruction for the storage data area to be tested is received. If the current system wants to inspect the storage data area to be tested or read the data, the data retention period of the storage data area to be tested is predicted to avoid repeated prediction of the data retention period of the storage data area to be tested, which leads to waste of resources and occupation of system resources. Moreover, by predicting the data retention period only when the storage data area to be tested is inspected or read, the current data retention period can be predicted in time to ensure the effectiveness of the processing decision of the storage data area to be tested.

[0116] As a preferred embodiment, after predicting the data retention period of the data storage area to be tested using the data retention period prediction model, the method further includes:

[0117] Determine whether the data retention period of the data storage area to be tested is less than a preset shelf life threshold;

[0118] If so, the data stored in the data storage area to be tested is rewritten or transferred to a spare storage location.

[0119] In this embodiment, after determining the predicted data retention period of the storage data area to be tested, when processing the data in the storage data area to be tested, it is first determined whether the predicted data retention period is less than the preset shelf life threshold. If it is less than, the probability of errors in the data stored in the storage data area to be tested is relatively high. At this time, the data stored in the storage data area to be tested is rewritten or transferred to a spare storage location to update the data retention period of the data stored in the storage data area to be tested, so as to avoid errors in the data stored in the storage data area to be tested.

[0120] It should be noted that when transferring the data stored in the storage data area to be tested to the backup storage location, it can be, but is not limited to, adding the data stored in the storage data area to be tested to the forced space garbage collection queue, so that through the garbage collection operation, the data stored in the storage data area to be tested is transferred to a new storage location, i.e., the backup storage location, before the data stored in the storage data area to be tested reaches its data retention period.

[0121] In addition, if the predicted data retention period is not less than the preset shelf life threshold, the data stored in the test storage data area is relatively safe and does not need to be transferred. The next inspection instruction or read instruction sending time can be set according to the predicted data retention period, which increases the security of the test storage data area while reducing the workload of the system when transferring data. Figure 9 , Figure 9 A flowchart of another embodiment of a data retention period prediction method provided by the present invention is provided. After setting the time for sending the next inspection instruction or read instruction, the data retention period of the storage data area to be tested can be predicted after receiving the next inspection instruction or read instruction.

[0122] Please refer to Figure 10 , Figure 10 This is a schematic diagram of the structure of a data retention period prediction system provided by the present invention, the system comprising:

[0123] A determining unit 101 is configured to determine N first sampling storage units in a sample data storage area, where N is a positive integer;

[0124] A matrix establishing unit 102 is configured to establish a first error confusion matrix corresponding to each of the first sampling storage units based on an error condition of the minimum unit storage data of each of the first sampling storage units;

[0125] A model acquisition unit 103 is configured to perform machine learning on each of the first error confusion matrices to obtain a data retention period prediction model;

[0126] The prediction unit 104 is configured to predict the data retention period of the data storage area to be tested by using the data retention period prediction model.

[0127] For an introduction to a data retention period prediction system provided by the present invention, please refer to the above method embodiment, and the present invention will not be described in detail here.

[0128] Please refer to Figure 11 , Figure 11 This is a schematic diagram of the structure of a data retention period prediction device provided by the present invention, the device comprising:

[0129] Memory 111, used for storing computer programs;

[0130] The processor 112 is configured to implement the steps of the above-mentioned method for predicting the data retention period when executing a computer program.

[0131] For an introduction to a data retention period prediction device provided by the present invention, please refer to the above method embodiment, and the present invention will not be described in detail here.

[0132] The computer-readable storage medium in the present invention stores a computer program, and when the computer program is executed by a processor, the steps of the method for predicting the data retention period as described above are implemented.

[0133] For an introduction to the computer-readable storage medium provided by the present invention, please refer to the above method embodiment, and the present invention will not go into details here.

[0134] It should also be noted that, in this specification, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

[0135] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for predicting data retention period, characterized in that: include: Determine N first sampling storage units in the sample data storage area, where N is a positive integer; Establishing a first error confusion matrix corresponding to each of the first sampling storage units based on an error condition of the minimum unit storage data of each of the first sampling storage units; Performing machine learning on each of the first error confusion matrices to obtain a data retention period prediction model; Using the data retention period prediction model to predict the data retention period of the data storage area to be tested; The establishing of a first error confusion matrix corresponding to each of the first sampling storage units based on an error condition of the minimum unit storage data of each of the first sampling storage units includes: Determining a change relationship of a programming state in a minimum unit based on a change in the storage power of each bit in the minimum unit of each of the first sampling storage units, so as to classify data errors that may occur in the minimum unit; counting errors of the minimum unit stored data in each of the first sampling storage units according to the error classification, so as to obtain the error statistics number of each type of error corresponding to each of the first sampling storage units; Establishing the corresponding first error confusion matrix based on the error statistics of each type of error corresponding to each of the first sampling storage units; After predicting the data retention period of the data storage area to be tested using the data retention period prediction model, the method further includes: Determining whether the data storage period of the tested data storage area is less than a preset shelf life threshold; If so, rewriting or transferring the data stored in the data storage area to be tested to a spare storage location; If not, the data stored in the data storage area to be tested is retained.

2. The method for predicting data retention period according to claim 1, wherein: Performing machine learning on each of the first error confusion matrices to obtain a data shelf life prediction model includes: Establishing a machine learning model based on each of the first error confusion matrices; The machine learning model is trained using each of the first error confusion matrices to obtain the data retention period prediction model.

3. The method for predicting data retention period according to claim 2, wherein: The machine learning model includes a regression model or a neural network model.

4. The method for predicting data retention period according to claim 1, wherein: The data retention period prediction model is used to predict the data retention period of the data storage area to be tested, including: Determine M second sampling storage units in the data storage area to be tested, where M is a positive integer; Establishing a second error confusion matrix corresponding to each second sampling storage unit based on an error condition of the minimum unit storage data of each second sampling storage unit; The second error confusion matrix is ​​input into the data retention period prediction model for calculation to obtain the data retention period of the data storage area to be tested.

5. The method for predicting data retention period according to claim 4, wherein: The establishing of a second error confusion matrix corresponding to each second sampling storage unit based on an error condition of the minimum unit storage data of each second sampling storage unit includes: performing error classification on data errors that may occur in minimum units based on a change relationship of minimum unit programming states of each of the second sampled memory cells; Counting the error conditions of the minimum unit stored data in each of the second sampling storage units according to the error classification to obtain the error statistics number of each type of error corresponding to each of the second sampling storage units; The corresponding second error confusion matrix is ​​established based on the error statistics of each type of error corresponding to each second sampling storage unit.

6. The method for predicting data retention period according to claim 1, wherein: After establishing a first error confusion matrix corresponding to each of the first sampling storage units based on an error condition of the minimum unit storage data of each of the first sampling storage units, the method further includes: Normalizing the first error confusion matrix of each of the first sampling storage units to determine a third error confusion matrix; Performing machine learning on each of the first error confusion matrices to obtain a data shelf life prediction model includes: Machine learning is performed on each of the third error confusion matrices to obtain a data retention period prediction model.

7. The method for predicting data retention period according to claim 6, wherein: After normalizing the first error confusion matrix of each of the first sampling storage units to determine a third error confusion matrix, the method further includes: Performing nonlinear mapping processing on each element in the third error confusion matrix to generate a fourth error confusion matrix; Performing machine learning on each of the third error confusion matrices to obtain a data shelf life prediction model includes: Machine learning is performed on each of the fourth error confusion matrices to obtain a data retention period prediction model.

8. The method for predicting data retention period according to claim 1, wherein: Before using the data retention period prediction model to predict the data retention period of the data storage area to be tested, the method further includes: Determining whether an inspection instruction or a read instruction for the data storage area to be tested is received; If so, the process proceeds to the step of using the data retention period prediction model to predict the data retention period of the data storage area to be tested.

9. A data retention period prediction system, characterized in that: include: a determining unit, configured to determine N first sampling storage units in the sample data storage area, where N is a positive integer; a matrix establishing unit, configured to establish a first error confusion matrix corresponding to each of the first sampling storage units based on an error condition of the minimum unit storage data of each of the first sampling storage units; a model acquisition unit, configured to perform machine learning on each of the first error confusion matrices to obtain a data retention period prediction model; A prediction unit, configured to predict the data retention period of the data storage area to be tested using the data retention period prediction model; The matrix establishment unit is specifically configured to determine a change relationship of a programming state in a minimum unit based on a change in the storage power of each bit in the minimum unit of each first sampling storage unit, so as to classify data errors that may occur in the minimum unit; counting errors of the minimum unit stored data in each of the first sampling storage units according to the error classification, so as to obtain the error statistics number of each type of error corresponding to each of the first sampling storage units; Establishing the corresponding first error confusion matrix based on the error statistics of each type of error corresponding to each of the first sampling storage units; After the prediction unit uses the data retention period prediction model to predict the data retention period of the storage data area to be tested, the prediction system is also used to: determine whether the data retention period of the storage data area to be tested is less than a preset shelf life threshold; if so, rewrite or transfer the data stored in the storage data area to be tested to a spare storage location; if not, retain the data stored in the storage data area to be tested.

10. A device for predicting data retention period, characterized in that: include: memory for storing computer programs; A processor, configured to implement the steps of the method for predicting the data retention period as claimed in any one of claims 1 to 8 when executing the computer program.

Images