CN114283875A - Method for dynamically predicting NAND Block end-of-life performance - Google Patents
Method for dynamically predicting NAND Block end-of-life performance Download PDFInfo
- Publication number
- CN114283875A CN114283875A CN202111586445.9A CN202111586445A CN114283875A CN 114283875 A CN114283875 A CN 114283875A CN 202111586445 A CN202111586445 A CN 202111586445A CN 114283875 A CN114283875 A CN 114283875A
- Authority
- CN
- China
- Prior art keywords
- block
- nand
- voltage value
- performance
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 23
- 206010011906 Death Diseases 0.000 title claims description 12
- 238000005191 phase separation Methods 0.000 claims description 9
- 238000013500 data storage Methods 0.000 claims description 6
- 238000005299 abrasion Methods 0.000 claims description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000014759 maintenance of location Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Landscapes
- Debugging And Monitoring (AREA)
Abstract
The invention discloses a method for dynamically predicting the performance of a NAND Block at the end of the service life, which collects Block information under various influence factors in real time, analyzes the information, dynamically predicts whether the performance of the Block at the end of the service life meets the design requirement or not, and dynamically calculates the voltage value required by the end of the service life, so that a user can take corresponding measures according to the result of the method to reduce the huge loss caused by a large number of reading errors or other error correction measures at the end of the service life of NAND Flash. Based on the method, the performance of the Block at the end of the service life can be dynamically predicted under the conditions of low resource consumption and full utilization of data, the Block with poor performance is identified in advance, and the probability of operation errors is reduced.
Description
Technical Field
The invention relates to the field of storage, in particular to a method for dynamically predicting the end-of-life performance of NAND Block in NAND Flash.
Background
The NAND Flash gradually decreases Data Retention capacity (Data Retention), read interference resistance, and the like as the wear frequency (PE Cycle) increases during use. The wear count (PE Cycle) is understood as the life of NAND, and the Block of NAND Flash at the end of the operational life may have different predicted errors, such as write failure, read failure (read operation failure or read error too high).
As shown in fig. 1, although a manufacturer guarantees that each Block has the same performance after the NAND Flash is packaged, due to factors such as material, internal circuit, and operating voltage, the Block performance is different, and the Block performance is fixed after the chip is packaged, instead of being mutated along with the increase of the service life, through analysis of a large amount of Block data in NAND flashes of different manufacturers and different batches.
Based on the rule, a series of error correction measures caused by excessive reading errors can be reduced by Block life prediction of NAND Flash, if static prediction modes such as machine learning and the like are adopted for life prediction, firstly, a large amount of data needs to be collected in advance, secondly, the performance characteristics of each Block are inconsistent and cannot cover all blocks, so that the prediction accuracy is low, thirdly, a large amount of resources or special modules are needed when a machine learning algorithm is used in a main control of an SSD chip, and the static prediction method of machine learning is difficult to realize in the SSD due to the reasons.
Disclosure of Invention
The invention aims to provide a method for dynamically predicting the performance of the NAND Block at the end of the service life, which can realize the dynamic prediction of the performance of the Block at the end of the service life under the conditions of low resource consumption and full utilization of data, identify blocks with poor performance in advance and reduce the probability of operation errors.
In order to solve the technical problem, the technical scheme adopted by the invention is as follows: a method for dynamically predicting the end-of-life performance of a NAND Block comprises the following steps:
s01), the NAND data center collects Block operation data in real time, the wear frequency in the selected life cycle is taken as a phase separation point by n, the data storage duration is taken as the phase separation point by a hour at the temperature t, the read interference frequency is taken as the phase separation point by b, the maximum error number FBCmax in different phases, the NAND voltage value and the voltage change direction used when the maximum error number is read, the read frequency and the read failure frequency are collected, and a three-dimensional data set is formed;
s02), when the NAND data center finishes data acquisition, informing the dynamic prediction module to analyze the data;
s03), the dynamic prediction module checks the reading failure times, if the value is more than 2, the Block is marked as a bad Block;
s04), the dynamic prediction module checks whether the maximum error number FBCmax at the initial stage of the Block life is greater than the error number threshold value FBCth, the reading times are 2, the voltage change direction is left, if the maximum error number FBCmax at the initial stage of the Block life is greater than the error number threshold value FBCth, the performance mark EOL _ L is recorded and is added with 1, and when the EOL _ L is greater than 2, the Block is marked as the performance if the Block;
s05), the dynamic prediction module predicts the last-stage voltage value of the Block according to the historical data of the NAND data center and dynamically predicts the last-stage voltage value of the Block according to the abrasion times, the data storage time, the reading interference, the maximum error number, the voltage value and the voltage direction.
Further, step S05) constructing a polynomial equation prediction model, calculating each parameter by using the voltage value as a solution and the influence factor and the error number as variables according to the historical data to form a polynomial equation, and substituting the influence factor and the error number in the time of calculating the voltage value into the polynomial equation for calculation and solution;
the polynomial equation is constructed as follows:
Vt = aXret^3 + bXpe^2 + cXfbc + dXret^2*Xpe + eXret*Xpe + f,
wherein a, b, c, d, e and f are the coefficients to be calculated, Xret, Xpe and Xfbc are the storage time, the wear times and the maximum error number of the collected data, and Vt is the predicted voltage value;
when calculating the voltage value, the influence factors and the error number are substituted into the polynomial equation to calculate and solve.
Further, step S05) calculates the difference between the two FBCs and the voltage value change value each time the influence factor changes, and calculates the average value of the two values, wherein the change value and the average value are regarded as the FBC change value influenced by each voltage value change of the Block in a certain voltage change direction, and the voltage value required by a certain kind of influence factor at the end stage is calculated according to the values.
Further, the Block life initial stage in step S04) means that the number of Block wearing times is within 3000.
Further, n =500 or n = 1000.
Further, t =85 ℃, a =2 hours.
Further, b = 500.
The invention has the beneficial effects that: by the method described in the patent, the Block information under various influence factors can be collected in real time, the information is analyzed to dynamically predict whether the performance of the Block at the end of the service life meets the design requirement or not, and the voltage value required by the end of the service life is dynamically calculated, so that a user can take corresponding measures according to the result of the method to reduce a large number of read errors occurring at the end of the service life of the NAND Flash or huge loss caused by other error correction measures.
Drawings
FIG. 1 is a schematic diagram of the maximum number of read errors during the full life of Block;
FIG. 2 is a flow chart of the present invention.
Detailed Description
The invention is further described with reference to the following figures and specific embodiments.
Example 1
The embodiment discloses a method for dynamically predicting the end-of-life performance of a NAND Block, which comprises the following steps as shown in FIG. 2:
s01), the NAND data center collects Block operation data in real time, the wear frequency in the selected life period is 500 as a phase separation point, the data storage duration is 2 hours at the temperature of 85 ℃ as the phase separation point, the read interference frequency is 500 as the phase separation point, the maximum error number FBCmax in different phases, the NAND voltage value and the voltage change direction used when the maximum error number is read, the read frequency and the read failure frequency are collected, and a three-dimensional data set is formed;
s02), when the NAND data center finishes data acquisition, informing the dynamic prediction module to analyze the data;
s03), the dynamic prediction module checks the reading failure times, if the value is more than 2, the Block is marked as a bad Block;
s04), the dynamic prediction module checks whether the maximum error number FBCmax at the initial stage of the Block life is greater than the error number threshold value FBCth, the reading times are 2, the voltage change direction is left, if the maximum error number FBCmax at the initial stage of the Block life is greater than the error number threshold value FBCth, the performance mark EOL _ L is recorded and is added with 1, and when the EOL _ L is greater than 2, the Block is marked as the performance if the Block;
s05), the dynamic prediction module predicts the last-stage voltage value of the Block according to the historical data of the NAND data center and dynamically predicts the last-stage voltage value of the Block according to the abrasion times, the data storage time, the reading interference, the maximum error number, the voltage value and the voltage direction.
In this embodiment, step S05) constructs a polynomial equation prediction model, calculates each parameter with the voltage value as a solution and the influence factor and the error number as variables according to the historical data to form a polynomial equation, and substitutes the influence factor and the error number where the voltage value is located into the polynomial equation for calculation and solution when calculating the voltage value;
the polynomial equation is constructed as follows:
Vt = aXret^3 + bXpe^2 + cXfbc + dXret^2*Xpe + eXret*Xpe + f,
wherein a, b, c, d, e and f are the coefficients to be calculated, Xret, Xpe and Xfbc are the storage time, the wear times and the maximum error number of the collected data, and Vt is the predicted voltage value;
when calculating the voltage value, the influence factors and the error number are substituted into the polynomial equation to calculate and solve.
Step S05 adopts another method: calculating the difference between two FBCs and the voltage value change value when each influence factor changes, respectively obtaining the average value of the two values, regarding the change value and the average value as the FBC change value influenced by each voltage value change of the Block in a certain voltage change direction, and calculating the required voltage value under certain influence factors in the last stage according to the values.
In this embodiment, the Block life initial stage in step S04) means that the Block wear count is within 3000.
By the method described in the embodiment, the Block information under each influence factor can be collected in real time, the information is analyzed to dynamically predict whether the performance of the Block at the end of the service life meets the design requirement or not, and the voltage value required by the end of the service life is dynamically calculated, so that a user can take corresponding measures according to the result of the method to reduce a large number of read errors occurring at the end of the service life of the NAND Flash or huge loss caused by other error correction measures.
The foregoing description is only for the basic principle and the preferred embodiments of the present invention, and modifications and substitutions by those skilled in the art are included in the scope of the present invention.
Claims (7)
1. A method for dynamically predicting the end-of-life performance of a NAND Block is characterized by comprising the following steps: the method comprises the following steps:
s01), the NAND data center collects Block operation data in real time, the wear frequency in the selected life cycle is taken as a phase separation point by n, the data storage duration is taken as the phase separation point by a hour at the temperature t, the read interference frequency is taken as the phase separation point by b, the maximum error number FBCmax in different phases, the NAND voltage value and the voltage change direction used when the maximum error number is read, the read frequency and the read failure frequency are collected, and a three-dimensional data set is formed;
s02), when the NAND data center finishes data acquisition, informing the dynamic prediction module to analyze the data;
s03), the dynamic prediction module checks the reading failure times, if the value is more than 2, the Block is marked as a bad Block;
s04), the dynamic prediction module checks whether the maximum error number FBCmax at the initial life stage of the Block is greater than the error number threshold value FBCth, the reading times are 2, the voltage change direction is left, if the maximum error number FBCmax and the reading times are all 2, the performance mark EOL _ L is recorded and added with 1, and when the EOL _ L is greater than 2, the Block is marked as a low-performance Block;
s05), the dynamic prediction module predicts the last-stage voltage value of the Block according to the historical data of the NAND data center and dynamically predicts the last-stage voltage value of the Block according to the abrasion times, the data storage time, the reading interference, the maximum error number, the voltage value and the voltage direction.
2. The method of dynamically predicting the end-of-life performance of a NAND Block of claim 1, wherein: step S05), a polynomial equation prediction model is built, voltage values are used as solutions according to historical data, influence factors and error numbers are used as variables to calculate all parameters to form a polynomial equation, and the influence factors and the error numbers where the voltage values are located are substituted into the polynomial equation to calculate and solve when the voltage values are calculated;
the polynomial equation that can be constructed is of the form:
Vt = aXret^3 + bXpe^2 + cXfbc + dXret^2*Xpe + eXret*Xpe + f
wherein a, b, c, d, e and f are the coefficients to be calculated, Xret, Xpe and Xfbc are the storage time, the wear times and the maximum error number of the collected data, and Vt is the predicted voltage value;
when calculating the voltage value, the influence factors and the error number are substituted into the polynomial equation to calculate and solve.
3. The method of dynamically predicting the end-of-life performance of a NAND Block of claim 1, wherein: step S05) calculates the difference between two FBCs and the voltage value variation value each time the influence factor varies, and calculates the average value of the two values, wherein the variation value and the average value are regarded as the FBC variation value influenced by each voltage value variation of the Block in a certain voltage variation direction, and the voltage value required by a certain kind of influence factor at the end stage is calculated according to the values.
4. The method of dynamically predicting the end-of-life performance of a NAND Block of claim 1, wherein: the Block life initial stage in step S04) means that the number of Block wearing times is within 3000.
5. The method of dynamically predicting the end-of-life performance of a NAND Block of claim 1, wherein: n =500 or n = 1000.
6. The method of dynamically predicting the end-of-life performance of a NAND Block of claim 1, wherein: t =85 ℃, a =2 hours.
7. The method of dynamically predicting the end-of-life performance of a NAND Block of claim 1, wherein: b = 500.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111586445.9A CN114283875A (en) | 2021-12-23 | 2021-12-23 | Method for dynamically predicting NAND Block end-of-life performance |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111586445.9A CN114283875A (en) | 2021-12-23 | 2021-12-23 | Method for dynamically predicting NAND Block end-of-life performance |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114283875A true CN114283875A (en) | 2022-04-05 |
Family
ID=80874208
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111586445.9A Pending CN114283875A (en) | 2021-12-23 | 2021-12-23 | Method for dynamically predicting NAND Block end-of-life performance |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114283875A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115101113A (en) * | 2022-07-08 | 2022-09-23 | 山东华芯半导体有限公司 | Method for analyzing variation of QLC threshold distribution under various conditions |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009230660A (en) * | 2008-03-25 | 2009-10-08 | Nec Infrontia Corp | Apparatus for monitoring life of flash memory storage |
CN108228371A (en) * | 2016-12-15 | 2018-06-29 | 发那科株式会社 | Machine learning device and method, life predication apparatus, numerical control device |
CN108766496A (en) * | 2018-05-23 | 2018-11-06 | 武汉忆数存储技术有限公司 | A kind of method and device in online dynamic prediction flash chip service life |
CN109119108A (en) * | 2018-08-15 | 2019-01-01 | 杭州阿姆科技有限公司 | A method of improving the Nand service life |
CN112256462A (en) * | 2020-10-29 | 2021-01-22 | 浪潮(北京)电子信息产业有限公司 | Method, device and medium for estimating service life of NAND Flash memory |
-
2021
- 2021-12-23 CN CN202111586445.9A patent/CN114283875A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009230660A (en) * | 2008-03-25 | 2009-10-08 | Nec Infrontia Corp | Apparatus for monitoring life of flash memory storage |
CN108228371A (en) * | 2016-12-15 | 2018-06-29 | 发那科株式会社 | Machine learning device and method, life predication apparatus, numerical control device |
CN108766496A (en) * | 2018-05-23 | 2018-11-06 | 武汉忆数存储技术有限公司 | A kind of method and device in online dynamic prediction flash chip service life |
CN109119108A (en) * | 2018-08-15 | 2019-01-01 | 杭州阿姆科技有限公司 | A method of improving the Nand service life |
CN112256462A (en) * | 2020-10-29 | 2021-01-22 | 浪潮(北京)电子信息产业有限公司 | Method, device and medium for estimating service life of NAND Flash memory |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115101113A (en) * | 2022-07-08 | 2022-09-23 | 山东华芯半导体有限公司 | Method for analyzing variation of QLC threshold distribution under various conditions |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109634527B (en) | Flash memory service life prediction method realized in SSD | |
CN114114049B (en) | Lithium ion battery life prediction method based on sample migration | |
WO2016092811A1 (en) | Power storage element state estimation device and power storage element state estimation method | |
EP3839537B1 (en) | Battery management apparatus | |
US9361222B2 (en) | Electronic system with storage drive life estimation mechanism and method of operation thereof | |
WO2022165955A1 (en) | Flash memory abnormality detection method and apparatus, and computer device and storage medium | |
CN114283875A (en) | Method for dynamically predicting NAND Block end-of-life performance | |
KR20200111016A (en) | Apparatus for estimating state of battery | |
JP2020180820A (en) | Battery evaluation system, battery evaluation method, and program | |
CN111176575A (en) | SSD (solid State disk) service life prediction method, system, terminal and storage medium based on Prophet model | |
US20230305065A1 (en) | Apparatus and method for diagnosing state of battery | |
US20220099750A1 (en) | Method and Apparatus for Parameterizing an Electrochemical Battery Model for a Battery Based on Data from a Plurality of Batteries | |
KR102683174B1 (en) | Method for compressing and restoring time series data | |
EP4145157A1 (en) | Battery system state of health monitoring system | |
CN117805649A (en) | Method for identifying abnormal battery cells based on SOH quantized battery capacity attenuation | |
GB2583510A (en) | A method and apparatus for detecting defective cells within a battery | |
CN115550195A (en) | Traffic suppression prediction method, electronic device, and storage medium | |
CN112309480A (en) | Failure early warning device and failure early warning method of flash memory device | |
CN116400249A (en) | Detection method and device for energy storage battery | |
CN118011254A (en) | SVR-based battery health state prediction method, system and equipment | |
CN111459706B (en) | NAND flash memory error rate prediction method and system | |
CN111312326A (en) | Flash memory life testing method and device, power acquisition terminal and storage medium | |
CN109815534A (en) | A kind of service life of flash memory prediction technique and device based on decision Tree algorithms | |
CN113030760B (en) | Battery life prediction method, device and storage medium | |
CN111736768B (en) | Method, system and storage medium for predicting remaining service time of server hard disk |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |