CN-122024807-A - Solid state disk life prediction method and equipment based on predictive wear balance

Abstract

The invention provides a solid state disk life prediction method and equipment based on predictive wear leveling, wherein the method comprises the steps of obtaining a wear state snapshot sequence and wear leveling scheduling history record of a solid state disk, executing wear trend state space reconstruction operation on obtained data, iteratively injecting history migration operation information as a state disturbance factor into a state transition matrix to obtain a wear evolution correction track set containing deviation correction factors, adjusting a dynamic wear threshold according to the deviation correction factors to generate an imbalance tolerance threshold set of a storage unit level, comparing a wear index predicted value sequence with the imbalance tolerance threshold to identify a predictive wear imbalance unit, and generating a dynamic wear leveling strategy set based on an identification result. According to the method, the historical wear balancing operation is used as a disturbance factor to be integrated into the wear prediction process, so that the wear unbalanced unit is identified in advance and intervened in a targeted manner.

Inventors

• MA ZHEN
• JIN YOUGANG
• YANG CHUNLIANG
• HU XIAONAN

Assignees

• 贵州大学
• 深圳市大乘科技股份有限公司

Dates

Publication Date

20260512

Application Date

20260410

Claims (10)

1. 1. The solid state disk life prediction method based on predictive wear balance is characterized by comprising the following steps of: Acquiring a wear state snapshot sequence and a wear balance scheduling history record of a solid state disk, wherein the wear state snapshot sequence comprises a plurality of wear state snapshot units, and the wear balance scheduling history record comprises a time stamp of executed data migration operation and migration data quantity; Executing a wear trend state space reconstruction operation on the wear state snapshot sequence and the wear balance scheduling history, and iteratively injecting migration operation information in the wear balance scheduling history as a state disturbance factor into a state transition matrix of a wear trend prediction process to obtain a wear evolution correction track set corresponding to each wear state snapshot unit, wherein the wear evolution correction track set comprises a wear index predicted value sequence and a deviation correction factor of a wear index actual observed value sequence of a storage unit; Performing dynamic wear threshold adjustment processing according to deviation correction factors of all storage units in the wear evolution correction track set, and mapping the accumulated effect of the deviation correction factors into an unbalanced tolerance threshold set of the storage unit level; Comparing the wear index predicted value sequence of each storage unit in the wear evolution correction track set with a corresponding threshold value in the unbalance tolerance threshold value set, and identifying the storage unit with the wear index predicted value exceeding the corresponding threshold value as a predictive wear unbalance unit; generating a dynamic wear leveling strategy set based on the predictive wear leveling units, wherein the dynamic wear leveling strategy set comprises migration operation execution time points and migration data volume allocation schemes aiming at each predictive wear leveling unit.
2. 2. The method according to claim 1, wherein the performing a wear trend state space reconstruction operation on the wear state snapshot sequence and the wear leveling schedule history, and iteratively injecting migration operation information in the wear leveling schedule history as a state disturbance factor into a state transition matrix of a wear trend prediction process to obtain a wear evolution correction track set corresponding to each wear state snapshot unit includes: extracting an original wear index data set corresponding to a current wear state snapshot unit from the wear state snapshot sequence, and extracting a migration operation record set in a preset time window before the current wear state snapshot unit from the wear balance scheduling historical record; Determining the data migration total amount of each storage unit in the preset time window according to a migration data amount distribution scheme in the migration operation record set, and analyzing the abrasion evolution inhibition amount of each storage unit in the preset time window due to migration operation according to the data migration total amount; acquiring a wear evolution prediction track corresponding to a previous wear state snapshot unit, and carrying out association fusion on a wear index prediction value of each storage unit in the wear evolution prediction track and the wear evolution inhibition amount to obtain a preliminary corrected wear prediction value; performing deviation comparison on the preliminary correction wear predicted value and the actual observed value of the wear index of the corresponding storage unit in the snapshot unit of the current wear state to generate a single-step correction deviation value corresponding to each storage unit in the snapshot unit of the current wear state; Embedding the single-step correction deviation value into a state transition matrix of the wear trend prediction process through a state disturbance injection mechanism, generating a deviation correction factor corresponding to each storage unit, and carrying out cooperative integration on the deviation correction factors and the wear evolution prediction track to obtain a wear evolution correction track set corresponding to the snapshot unit of the current wear state; And storing the abrasion evolution correction track set corresponding to the current abrasion state snapshot unit into an abrasion evolution correction track storage area, and synchronously storing the deviation correction factors into a deviation correction factor history record library.
3. 3. The method according to claim 2, wherein the determining the total data migration amount of each storage unit in the preset time window according to the migration data amount allocation scheme in the migration operation record set, and resolving the wear evolution suppression amount of each storage unit due to the migration operation in the preset time window according to the total data migration amount, includes: Analyzing each migration operation record in the migration operation record set, extracting a source storage unit identifier and the data quantity migrated from the source storage unit contained in each migration operation record, and carrying out aggregation accumulation on the migration data quantity corresponding to each source storage unit identifier to obtain the data migration total quantity of each storage unit in the preset time window; Acquiring a storage unit abrasion model of the solid state disk, wherein the storage unit abrasion model defines the inhibition relation of migration operation to the storage unit abrasion evolution process; mapping and matching the data migration total amount of each storage unit with the inhibition relation in the storage unit abrasion model to obtain the abrasion evolution inhibition total amount of each storage unit in the preset time window; Acquiring the duration of the preset time window, and carrying out distribution processing on the total abrasion evolution inhibition amount of each storage unit according to the duration of the time window to obtain the abrasion evolution inhibition rate of each storage unit in unit time; Taking the wear evolution inhibition rate as a wear evolution inhibition amount of each storage unit, wherein the wear evolution inhibition amount is used for representing the correction rate of migration operation on the predicted value of the wear index of the storage unit in unit time; and arranging the abrasion evolution inhibition amounts of the storage units according to the storage address sequence to generate an abrasion evolution inhibition amount vector, wherein the vector length of the abrasion evolution inhibition amount vector is equal to the total number of the storage units.
4. 4. The method of claim 3, wherein the obtaining a wear evolution prediction track corresponding to a previous wear state snapshot unit, and the performing association fusion on the wear index prediction value of each storage unit in the wear evolution prediction track and the wear evolution inhibition amount to obtain a preliminary corrected wear prediction value, includes: Reading a wear evolution prediction track corresponding to a previous wear state snapshot unit from a wear evolution prediction track storage area, wherein the wear evolution prediction track comprises a wear index prediction value sequence from a preset starting time point to a previous wear state snapshot unit acquisition time point of each storage unit; for each storage unit, extracting a wear index predicted value corresponding to the last time point in the wear index predicted value sequence as a reference predicted wear value; Calculating the accumulated wear evolution inhibition amount generated by migration operation in the time interval according to the time interval between the previous wear state snapshot unit and the current wear state snapshot unit and the wear evolution inhibition amount of each storage unit; performing difference operation on the reference predicted wear value of each storage unit and the wear evolution inhibition accumulation amount to obtain a preliminary corrected wear predicted value of each storage unit; And arranging the preliminary correction wear prediction values of the storage units according to the storage address sequence to generate a preliminary correction wear prediction value vector, wherein the vector dimension of the preliminary correction wear prediction value vector is equal to the total number of the storage units.
5. 5. The method according to claim 1, wherein the performing the dynamic wear threshold adjustment process according to the deviation correction factor of each storage unit in the wear evolution correction track set, mapping the cumulative effect of the deviation correction factors to the imbalance tolerance threshold set at the storage unit level, includes: extracting a deviation correction factor sequence of each storage unit from the abrasion evolution correction track set, wherein the deviation correction factor sequence comprises deviation correction factors corresponding to each abrasion state snapshot unit; Inputting the deviation correction factor sequences of the storage units into a deviation pattern identifier, wherein the deviation pattern identifier performs time sequence structural analysis on the deviation correction factor sequences, and identifies oscillation pattern characteristics and drift pattern characteristics existing in the deviation correction factor sequences, the oscillation pattern characteristics comprise fluctuation amplitude distribution information and fluctuation period length information of the deviation correction factors in the time dimension, and the drift pattern characteristics comprise accumulated offset direction information and offset speed information of the deviation correction factors in the time dimension; Classifying the oscillation intensity of each storage unit according to the fluctuation amplitude distribution information in the oscillation mode characteristics to obtain an oscillation intensity grade identifier, and judging the drift direction of each storage unit according to the accumulated offset direction information in the drift mode characteristics to obtain a drift direction symbol identifier; Combining and encoding the oscillation intensity grade identifier and the drift direction symbol identifier to generate a threshold adjustment mode code of each storage unit; A basic wear balance trigger threshold value of the solid state disk is obtained, differential reconstruction is carried out on the basic wear balance trigger threshold value according to the threshold value adjustment mode code, a storage unit with a drift direction sign being positive and an oscillation intensity level sign being higher than a preset oscillation level threshold value is inquired according to the oscillation intensity level sign of the storage unit, a preset threshold value lifting coefficient mapping table is obtained, a corresponding threshold value lifting coefficient is obtained, and the basic wear balance trigger threshold value is multiplied by the threshold value lifting coefficient to obtain an imbalance tolerance threshold value after lifting; for a storage unit with a negative drift direction sign and an oscillation intensity level sign higher than a preset oscillation level threshold, inquiring a preset threshold reduction coefficient mapping table according to the oscillation intensity level sign to obtain a corresponding threshold reduction coefficient, and multiplying the basic wear balance trigger threshold by the threshold reduction coefficient to obtain a reduced imbalance tolerance threshold; and for the storage units with oscillation intensity level marks not exceeding a preset oscillation level threshold, taking the basic wear balance trigger threshold as an unbalance tolerance threshold of the storage unit, arranging the unbalance tolerance thresholds of the storage units according to a storage address sequence, and generating an unbalance tolerance threshold set of the storage unit level.
6. 6. The method of claim 5, wherein the inputting the bias correction factor sequence of each memory cell into a bias pattern identifier, the bias pattern identifier performing a time-series structure analysis on the bias correction factor sequence, identifying oscillation pattern features and drift pattern features present in the bias correction factor sequence, comprises: Inputting the deviation correction factor sequences of all the storage units into a time domain segmentation layer of a deviation mode identifier, wherein the time domain segmentation layer divides the deviation correction factor sequences into a plurality of continuous time sequence sub-segments according to a preset time window length, and each time sequence sub-segment corresponds to a deviation correction factor set in a time window; Inputting the time sequence sub-segments into an oscillation characteristic extraction layer of a deviation mode identifier, performing extreme point detection on each time sequence sub-segment by the oscillation characteristic extraction layer, extracting local maximum value points and local minimum value points of deviation correction factors in each time sequence sub-segment, generating oscillation period parameters according to the frequency of the alternate occurrence of the local maximum value points and the local minimum value points, generating oscillation amplitude parameters according to the difference between the local maximum value points and the local minimum value points, and collecting the oscillation period parameters and the oscillation amplitude parameters into oscillation mode characteristics; Inputting the time sequence sub-segments into a drift feature extraction layer of a deviation mode identifier, performing endpoint value capturing on each time sequence sub-segment by the drift feature extraction layer, acquiring a start deviation correction factor value and an end deviation correction factor value of each time sequence sub-segment, generating a drift direction indication according to the difference direction of the end deviation correction factor value and the start deviation correction factor value, generating a drift speed indication according to the ratio of the absolute value of the difference between the end deviation correction factor value and the start deviation correction factor value to the time window length, and collecting the drift direction indication and the drift speed indication as drift mode features; Inputting the oscillation mode characteristics of adjacent time sequence sub-segments into an oscillation mode association analysis unit, wherein the oscillation mode association analysis unit calculates the variation of oscillation period parameters and the variation of oscillation amplitude parameters between the adjacent time sequence sub-segments to generate an oscillation evolution track describing the evolution rule of the oscillation mode on a time axis; Inputting the drift pattern characteristics of adjacent time sequence sub-segments into a drift pattern association analysis unit, wherein the drift pattern association analysis unit calculates the connection relation of drift direction indication and the connection relation of drift speed indication between the adjacent time sequence sub-segments, and generates a drift evolution track describing the evolution rule of a drift pattern on a time axis; And inputting the oscillation evolution track and the drift evolution track into a mode fusion layer of a deviation mode identifier, and aligning the oscillation evolution track with the drift evolution track in a time sequence by the mode fusion layer to generate a fusion mode descriptor, wherein the fusion mode descriptor comprises corresponding relation information of oscillation mode characteristics and drift mode characteristics on different time windows.
7. 7. The method of claim 6, wherein the classifying the oscillating intensity of each memory cell according to the oscillating amplitude distribution information in the oscillating pattern feature to obtain an oscillating intensity class identifier, and determining the drift direction of each memory cell according to the accumulated drift direction information in the drift pattern feature to obtain a drift direction symbol identifier, includes: Extracting oscillation amplitude parameters of each storage unit on all time sequence sub-segments from the fusion mode descriptor to form an oscillation amplitude parameter set, carrying out oscillation amplitude accumulation calculation on the oscillation amplitude parameter set to obtain oscillation amplitude accumulation amounts of each storage unit, and determining oscillation intensity grade identifiers of each storage unit according to the falling positions of the oscillation amplitude accumulation amounts in a preset oscillation grade division interval; Extracting drift direction indications of each storage unit on all time sequence sub-segments from the fusion mode descriptor to form a drift direction indication sequence, executing drift direction continuity analysis on the drift direction indication sequence, and if the number of continuous occurrence times of the same direction in the drift direction indication sequence exceeds a preset number threshold value, using the direction as a drift direction symbol mark; If the number of continuous occurrence times of any direction in the drift direction indication sequence exceeds a preset number threshold value, determining a dominant drift direction as a drift direction symbol mark according to the total number of occurrence times of each direction; According to the oscillation intensity level identifiers and the drift direction symbol identifiers of all the storage units, matching inquiry is carried out in a preset threshold adjustment strategy mapping table, the threshold adjustment strategy mapping table comprises adjustment strategy codes corresponding to different oscillation intensity level identifiers and different drift direction symbol identifier combinations, and the matched adjustment strategy codes are obtained; Performing association binding on the adjustment strategy codes and the storage addresses of the storage units to generate threshold adjustment mode codes of the storage units; and establishing an association relation between the threshold value adjustment mode codes of the storage units and the imbalance tolerance threshold values of the corresponding storage units.
8. 8. The method of claim 1, wherein comparing the wear indicator predictor sequence for each memory cell in the wear-evolution correction trajectory set with a corresponding threshold in the imbalance tolerance threshold set identifies a memory cell for which the wear indicator predictor exceeds the corresponding threshold as a predictive wear imbalance cell, comprising: Extracting a wear index predicted value sequence of each storage unit from the wear evolution correction track set, wherein the wear index predicted value sequence comprises wear index predicted values corresponding to a plurality of time points, acquiring an imbalance tolerance threshold value corresponding to each storage unit from the imbalance tolerance threshold value set, expanding the imbalance tolerance threshold value of each storage unit along a time axis, and generating a threshold value reference sequence aligned with the wear index predicted value sequence in time; Inputting the abrasion index predicted value sequence of each storage unit and the corresponding threshold value reference sequence into a track crossing analyzer, performing space superposition comparison on the abrasion index predicted value sequence and the threshold value reference sequence by the track crossing analyzer, detecting crossing points formed by crossing the abrasion index predicted value sequence from the lower part to the upper part of the threshold value reference sequence, and recording crossing time stamps corresponding to each crossing point and differences between the abrasion index predicted value and the unbalanced tolerance threshold value when crossing as crossing deviation amounts; Determining the exceeding duration of the predicted value of each storage unit according to the time interval between adjacent crossing points in the abrasion index predicted value sequence of each storage unit, and marking the storage unit with the exceeding duration of the predicted value exceeding the preset duration threshold as a continuous exceeding unit; Performing cross point density analysis on the continuous standard exceeding units, calculating the occurrence frequency of each continuous standard exceeding unit in a preset time window, and marking the continuous standard exceeding unit with the cross point occurrence frequency exceeding the preset frequency threshold as a high-frequency traversing unit according to the comparison result of the cross point occurrence frequency and the preset frequency threshold; Performing cross deviation accumulation analysis on the high-frequency crossing units, accumulating the cross deviation of all the crossing points of each high-frequency crossing unit to obtain cross deviation accumulation, and marking the high-frequency crossing units with the cross deviation accumulation exceeding a preset accumulation threshold as depth crossing units; and identifying the depth traversing units as predictive wear unbalanced units, and packaging the intersection distribution characteristics, the intersection deviation accumulation amount and the exceeding duration of each depth traversing unit to generate a predictive wear unbalanced unit characteristic record.
9. 9. The method according to claim 8, wherein the inputting the wear indicator predictor sequence of each storage unit into the track crossing analyzer, the track crossing analyzer performs a spatial overlapping comparison of the wear indicator predictor sequence and the threshold reference sequence, detects crossing points formed by crossing the wear indicator predictor sequence from below to above the threshold reference sequence, records crossing time stamps corresponding to each crossing point and differences between the wear indicator predictor and the unbalance tolerance threshold value at crossing as crossing deviation amounts, and includes: Inputting the abrasion index predicted value sequence of each storage unit and the corresponding threshold reference sequence into a sequence alignment layer of the track intersection analyzer, wherein the sequence alignment layer pairs each time point in the abrasion index predicted value sequence with a threshold value of the same time point in the threshold reference sequence to generate a prediction-threshold value pairing set with aligned time points; Inputting the prediction-threshold pairing set into a crossing detection layer of a track crossing analyzer, wherein the crossing detection layer crosses prediction-threshold pairing aligned with time points, and sequentially comparing the magnitude relation between the wear index predicted value and the unbalance tolerance threshold value at each time point to generate a magnitude relation mark sequence, wherein each mark in the magnitude relation mark sequence is used for indicating that the wear index predicted value at the corresponding time point is larger than the unbalance tolerance threshold value or smaller than or equal to the unbalance tolerance threshold value; inputting the size relation mark sequence into a crossing point positioning layer of a track crossing analyzer, wherein the crossing point positioning layer scans the change of adjacent marks in the size relation mark sequence, when the adjacent marks change from marks representing less than or equal to marks representing greater than the marks, determining an interval between two time points corresponding to the adjacent marks as a crossing interval, and extracting the crossing point position of a wear index predicted value sequence and a threshold value reference sequence in the crossing interval as a crossing point; Inputting all the identified crossing points into a crossing point characteristic recording layer of a track crossing analyzer, acquiring a starting time point and an ending time point of a crossing section where each crossing point is located by the crossing point characteristic recording layer, calculating accurate time stamps of the crossing points according to the starting time point and the ending time point, calculating a difference value between a wear index predicted value at the crossing point and an unbalance tolerance threshold value as crossing deviation amount, and storing the crossing time stamps and the crossing deviation amount in a correlated mode.
10. 10. An electronic device, comprising: A memory in which a computer program is stored; a processor for loading the computer program to implement the solid state disk life prediction method based on predictive wear leveling according to any one of claims 1-9.

Description

Solid state disk life prediction method and equipment based on predictive wear balance Technical Field The invention relates to the field of data processing and solid state storage, in particular to a solid state disk life prediction method and equipment based on predictive wear balance. Background With the continuous increase of the storage density of the solid state disk, the wear balance and the life prediction of the storage unit become key technologies for guaranteeing the reliability of equipment. The existing methods are generally based on real-time collected wear state data, triggering data migration when the degree of wear is detected to exceed a preset threshold, or inputting the wear state data into a predictive model to calculate future wear trends. However, in such a method, the trigger decision of wear balanced scheduling and wear trend prediction are mutually fractured, the prediction model fails to take the influence of the historical scheduling operation on the wear evolution into consideration, so that the prediction result has deviation from the real track of the scheduling interference, the equalization strategy formulated based on the prediction information is difficult to accurately match the actual wear development situation, and the accuracy of scheduling opportunity and the effectiveness of resource allocation are affected. Disclosure of Invention The invention provides a solid state disk life prediction method and equipment based on predictive wear balance. In a first aspect, an embodiment of the present invention provides a solid state disk life prediction method based on predictive wear leveling, where the method includes: Acquiring a wear state snapshot sequence and a wear balance scheduling history record of the solid state disk, wherein the wear state snapshot sequence comprises a plurality of wear state snapshot units, and the wear balance scheduling history record comprises a time stamp of executed data migration operation and migration data volume; Executing the operation of reconstructing the wear trend state space on the wear state snapshot sequence and the wear balanced scheduling history, and iteratively injecting migration operation information in the wear balanced scheduling history as a state disturbance factor into a state transition matrix of the wear trend prediction process to obtain a wear evolution correction track set corresponding to each wear state snapshot unit, wherein the wear evolution correction track set comprises deviation correction factors of a wear index predicted value sequence and a wear index actual observation value sequence of a storage unit; Performing dynamic wear threshold adjustment processing according to deviation correction factors of all storage units in the wear evolution correction track set, and mapping the accumulated effect of the deviation correction factors into an unbalanced tolerance threshold set of the storage unit level; comparing the wear index predicted value sequence of each storage unit in the wear evolution correction track set with a corresponding threshold value in the unbalance tolerance threshold value set, and identifying the storage unit with the wear index predicted value exceeding the corresponding threshold value as a predictive wear unbalance unit; And generating a dynamic wear balancing strategy set based on the predictive wear unbalanced units, wherein the dynamic wear balancing strategy set comprises migration operation execution time points and migration data volume allocation schemes aiming at each predictive wear unbalanced unit. In a second aspect, an embodiment of the present invention provides an electronic device, including: A memory in which a computer program is stored; And the processor is used for loading a computer program to realize the solid state disk life prediction method based on predictive wear balance. The method comprises the steps of taking a wear balanced scheduling history record as a state transition matrix of a state disturbance factor iteration injection wear trend prediction process, enabling a deviation correction factor to adaptively reflect the actual influence of historical migration operation on wear evolution, realizing dynamic embedding and recursion correction of scheduling intervention effects, mapping the accumulated effect of the deviation correction factor into an unbalanced tolerance threshold set of a storage unit level, enabling wear triggering conditions to be configured differently according to the evolution characteristics of each unit deviation factor, comparing a predicted track subjected to state disturbance correction with a tolerance threshold of personalized configuration, identifying a predictive wear unbalanced unit, fusing the corrected predicted information with the personalized threshold to avoid erroneous judgment and leakage, generating a dynamic wear balancing strategy comprising migration execution time points and migration data amount distributi