A solid state disk implicit fault detection method and related device

Abstract

The solid state disk implicit fault detection method and related equipment provided by the present disclosure can obtain target SMART logs of a target solid state disk, extract target SMART data collected at multiple collection time points from the target SMART logs, construct the target SMART data collected at each collection time point into target time sequence feature data with time sequence characteristics according to a collection time sequence, input the target time sequence feature data into a pre-trained time recurrent neural network model, and obtain an implicit fault detection result output by the time recurrent neural network model. The present disclosure constructs the collected target SMART data into target time sequence feature data with time sequence characteristics, and then uses the time recurrent neural network model to obtain an accurate and effective implicit fault detection result according to the target time sequence feature data with time sequence characteristics, thereby helping to discover the implicit fault of the solid state disk in a timely manner and avoiding data loss.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of storage, and in particular, to a solid state disk implicit fault detection method and related equipment. BACKGROUND

[0002] With the rapid development of computer technology, the processing capability of computer equipment is continuously improved. In order to provide faster data read and write speed to users and provide stronger environmental support for applications, solid state disks begin to be widely used.

[0003] In the application scenario of distributed big data, mass data need to be processed and analyzed, and these data are all mapped to solid state disks. Once the data stored in the solid state disk is lost due to a fault, it will have a difficult-to-repair impact on business services.

[0004] The implicit fault of the solid state disk has already affected the system and application performance, but the existing fault diagnosis software cannot find it, so how to detect the implicit fault of the solid state disk becomes a technical problem to be solved urgently. SUMMARY

[0005] In view of the above problems, the present disclosure provides a solid state disk implicit fault detection method and related equipment to overcome the above problems or at least partially solve the above problems, and the technical solution is as follows:

[0006] A solid state disk implicit fault detection method, comprising:

[0007] obtaining a target SMART log of a target solid state disk;

[0008] extracting target SMART data collected at multiple collection time points from the target SMART log;

[0009] constructing the target SMART data collected at each collection time point into target time sequence feature data with time sequence characteristics according to a collection time sequence;

[0010] inputting the target time sequence feature data into a pre-trained time recurrent neural network model to obtain an implicit fault detection result output by the time recurrent neural network model.

[0011] Optionally, the training process of the time recurrent neural network model comprises:

[0012] obtaining a historical full-amount SMART log of a training solid state disk that has occurred an explicit fault;

[0013] extracting training SMART data collected at multiple collection time points from the historical full-amount SMART log;

[0014] The training SMART data collected at each collection time is constructed into training time sequence feature data with time sequence characteristics according to a collection time sequence, and a sample label is added to the training time sequence feature data, the sample label indicating a time of occurrence of the implicit failure of the training solid state disk;

[0015] The training time sequence feature data to which the sample label is added is input into the time recurrent neural network model for machine learning, and a trained time recurrent neural network model is obtained.

[0016] Optionally, the method further comprises:

[0017] According to the time of occurrence of the explicit failure of the training solid state disk and a preset time threshold, the time of occurrence of the implicit failure of the training solid state disk is determined.

[0018] Optionally, the target SMART data and the training SMART data include at least one of the following data: a bottom layer data read error rate, a start / stop count, a remapping sector number, a power-on time accumulation, a spindle spin retry count, a hard disk calibration retry count, a hard disk power-on count, a report of uncorrectable errors, a command timeout, an offline uncorrectable sector count, a temperature, an ULTRA-ATA access check error rate, and a write error rate.

[0019] Optionally, before the target SMART log of the target solid state disk is obtained, the method further comprises:

[0020] The target solid state disk is subjected to explicit failure detection, and an explicit failure detection result is obtained.

[0021] In a case where the explicit failure detection result is that no explicit failure has occurred, the step of obtaining the target SMART log of the target solid state disk is performed.

[0022] Optionally, the method further comprises:

[0023] In a case where the implicit failure detection result is that an implicit failure has occurred, a preset implicit failure warning is output.

[0024] A solid state disk implicit failure detection device, comprising: a target SMART log obtaining unit, a target SMART data extracting unit, a target time sequence feature data constructing unit, and an implicit failure detection result obtaining unit,

[0025] The target SMART log obtaining unit is configured to obtain a target SMART log of a target solid state disk.

[0026] The target SMART data extracting unit is configured to extract target SMART data collected at a plurality of collection times from the target SMART log.

[0027] The target time sequence feature data construction unit is configured to construct the target SMART data collected at each collection time into target time sequence feature data with time sequence features according to a collection time sequence.

[0028] The implicit failure detection result obtaining unit is configured to input the target time sequence feature data into a pre-trained time recurrent neural network model to obtain an implicit failure detection result output by the time recurrent neural network model.

[0029] Optionally, the device further comprises a model training unit, which comprises a historical full-amount SMART log obtaining subunit, a training SMART data extraction subunit, a training time sequence feature data construction subunit, and a model obtaining subunit,

[0030] The historical full-amount SMART log obtaining subunit is configured to obtain a historical full-amount SMART log of a training solid state disk that has experienced an explicit failure.

[0031] The training SMART data extraction subunit is configured to extract training SMART data collected at multiple collection times from the historical full-amount SMART log.

[0032] The training time sequence feature data construction subunit is configured to construct the training SMART data collected at each collection time into training time sequence feature data with time sequence features according to a collection time sequence, and add a sample label to the training time sequence feature data, the sample label indicating a time of occurrence of an implicit failure of the training solid state disk.

[0033] The model obtaining subunit is configured to input the training time sequence feature data with the sample label added thereto into the time recurrent neural network model for machine learning to obtain the time recurrent neural network model that has been trained.

[0034] A computer readable storage medium having a program stored thereon, the program being executed by a processor to implement the solid state disk implicit failure detection method according to any one of the preceding embodiments.

[0035] An electronic device, comprising at least one processor, and at least one memory connected to the processor via a bus; wherein the processor, the memory and the bus complete communication with each other; the processor is configured to call program instructions in the memory to execute the solid state disk implicit failure detection method according to any one of the preceding embodiments.

[0036] By means of the technical scheme, the solid state disk implicit fault detection method and related equipment provided by the present disclosure can obtain target SMART logs of a target solid state disk; target SMART data collected at multiple collection time points is extracted from the target SMART logs; the target SMART data collected at the collection time points is constructed into target time sequence feature data with time sequence characteristics according to a collection time sequence; and the target time sequence feature data is input into a time recurrent neural network model pre-trained, to obtain an implicit fault detection result output by the time recurrent neural network model. The present disclosure constructs the collected target SMART data into target time sequence feature data with time sequence characteristics, and then uses the time recurrent neural network model to obtain an accurate and effective implicit fault detection result according to the target time sequence feature data with time sequence characteristics, thereby helping to discover the implicit fault of the solid state disk in time and avoiding data loss.

[0037] The above description is only a summary of the technical scheme of the present disclosure, in order to enable the technical means of the present disclosure to be more clearly understood, the contents of the specification can be implemented, and in order to enable the above and other purposes, features and advantages of the present disclosure to be more obvious and easy to understand, the specific embodiments of the present disclosure are described below. BRIEF DESCRIPTION OF DRAWINGS

[0038] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are intended to depict only preferred embodiments of the disclosure and therefore should not be considered to narrow its scope. Additionally, like reference numerals are intended to refer to like parts throughout this disclosure. In the drawings:

[0039] Figure 1 An embodiment of the solid state disk implicit fault detection method provided by the present disclosure is shown in a schematic diagram;

[0040] Figure 2 A training process of the time recurrent neural network model provided by the present disclosure is shown in a schematic diagram;

[0041] Figure 3 Another embodiment of the solid state disk implicit fault detection method provided by the present disclosure is shown in a schematic diagram;

[0042] Figure 4 Another embodiment of the solid state disk implicit fault detection method provided by the present disclosure is shown in a schematic diagram;

[0043] Figure 5 A structure of the solid state disk implicit fault detection device provided by the present disclosure is shown in a schematic diagram;

[0044] Figure 6Fig. 1 shows a structural schematic diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0045] Exemplary embodiments of the present disclosure will be described in detail with reference to the drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly and completely understood, and so that the scope of the present disclosure can be conveyed to those skilled in the art.

[0046] As shown in Figure 1 Fig. 1, an embodiment of a solid state disk implicit failure detection method provided by the present disclosure can include the following steps.

[0047] S100, obtaining target SMART logs of a target solid state disk.

[0048] The solid state disk (SSD) is also known as a solid state drive, which is made of an array of solid state electronic storage chips. SMART (Self-Monitoring Analysis and Reporting Technology) is an automatic state detection and early warning system that applies self-detection, analysis and reporting technology to the hard disk. The system monitors various information data of the hard disk and generates SMART logs according to the information data.

[0049] Optionally, an embodiment of the present disclosure can obtain the full amount of SMART logs of the target solid state disk from the factory to the current time as the target SMART logs.

[0050] S200, extracting target SMART data collected at a plurality of collection time points from the target SMART logs.

[0051] Optionally, the target SMART data can include at least one of the following data: raw read error rate, start / stop count, relocated sector count, power-on time count (POH), spin up retry count, calibration retry count, power cycle count, reported uncorrectable errors, command timeout, offline uncorrectable sector count, temperature, ULTRA-ATA CRC error rate, and write error rate.

[0052] It can be understood that, in the case that the target SMART data includes only one type of data, the target SMART data of one type of data collected at the plurality of collection time points can be extracted from the target SMART log. For example, assuming that the target SMART data includes only temperature, the temperature data collected at the three collection time points can be 10℃, 24℃, and 31℃, respectively.

[0053] It can be understood that, in the case that the target SMART data includes multiple types of data, the target SMART data of multiple types of data collected at the plurality of collection time points can be extracted from the target SMART log. For example, assuming that the target SMART data includes raw read error rate, power-on time count, and spin up retry count, the raw read error rate collected at the three collection time points can be 22%, 37%, and 45%, respectively, the power-on time count collected at the three collection time points can be 15 minutes, 23 minutes, and 46 minutes, respectively, and the spin up retry count collected at the three collection time points can be 10 times, 50 times, and 100 times, respectively.

[0054] S300, constructing the target SMART data collected at the plurality of collection time points into target time-series characteristic data having time-series characteristics according to the collection time sequence.

[0055] Specifically, embodiments of this disclosure can construct target SMART data of the same data type collected at each acquisition time into target time-series feature data with time-series characteristics. For ease of understanding, an example is provided below: Assume the target SMART data includes the underlying data read error rate and cumulative power-on time. The underlying data read error rate collected at the third time is 23% and the cumulative power-on time is 13 minutes; the underlying data read error rate collected at the fourth time is 38% and the cumulative power-on time is 21 minutes; and the underlying data read error rate collected at the fifth time is 46% and the cumulative power-on time is 39 minutes. Therefore, embodiments of this disclosure can construct the underlying data read error rates of 23%, 38%, and 46% collected sequentially at the third, fourth, and fifth times into first target time-series feature data with time-series characteristics, and construct the cumulative power-on times of 13 minutes, 21 minutes, and 39 minutes collected sequentially at the third, fourth, and fifth times into second target time-series feature data with time-series characteristics.

[0056] Specifically, in this embodiment of the disclosure, the target SMART data collected at each collection time can be constructed into target temporal feature data with temporal characteristics according to the order of collection time.

[0057] S400. Input the target time-series feature data into the pre-trained time recurrent neural network model to obtain the latent fault detection results output by the time recurrent neural network model.

[0058] Optionally, the time-recurrent neural network model can be a long short-term memory (LSTM) network model.

[0059] Optional, such as Figure 2 As shown, a training process for the time-recurrent neural network model provided in this embodiment may include:

[0060] S01. Obtain the full historical SMART logs of the training SSD that has experienced a visible failure.

[0061] Among them, the historical full SMART log is the SMART log of the training solid-state drive from the time it was manufactured until the moment of the explicit failure.

[0062] S02. Extract training SMART data collected at multiple collection times from the historical full SMART log.

[0063] Optionally, the training SMART data can include at least one of a bottom data read error rate, a start / stop count, a remap sector count, a power-on time accumulation, a spindle spin retry count, a hard disk calibration retry count, a hard disk power-on count, an uncorrectable error report, a command timeout, an offline uncorrectable sector count, a temperature, an ULTRA-ATA access check error rate, and a write error rate.

[0064] It can be understood that the training SMART data can refer to the description of the target SMART data, which will not be described here.

[0065] S03, according to the time sequence of collection, the training SMART data collected at each collection time is constructed into training time sequence feature data with time sequence features, and sample labels are added to the training time sequence feature data, the sample labels indicating the occurrence time of the implicit failure of the training solid state disk.

[0066] Optionally, the disclosure embodiment can determine the occurrence time of the implicit failure of the training solid state disk according to the occurrence time of the explicit failure of the training solid state disk and a preset time threshold.

[0067] The preset time threshold can be set according to actual needs.

[0068] S04, the training time sequence feature data with added sample labels is input into a time recurrent neural network model for machine learning, and a trained time recurrent neural network model is obtained.

[0069] It can be understood that the disclosure embodiment can obtain the historical full SMART logs of at least one training solid state disk that has occurred an explicit failure, and the training SMART data extracted from each historical full SMART log is used to construct the training time sequence feature data with time sequence features and add sample labels. Each training time sequence feature data with added sample labels is divided into a training set and a test set. The training time sequence feature data in the training set is input into the time recurrent neural network model for supervised learning and model parameter tuning, and the training time sequence feature data in the test set is used to verify the robustness and accuracy of the time recurrent neural network model, and finally a trained time recurrent neural network model is obtained.

[0070] The disclosure embodiment constructs the training SMART data into training time sequence feature data with time sequence features and adds sample labels, so that the time recurrent neural network can learn the time flow change features associated with the implicit failure in the SMART data of the solid state disk, thereby accurately detecting the implicit failure of the target solid state disk.

[0071] Optionally, based on Figure 1 The method shown in FIG. 8, such as Figure 3As shown in the method, before the step S100, the method further includes:

[0072] S001, performing dominant fault detection on the target solid state disk to obtain a dominant fault detection result. If the dominant fault detection result is that no dominant fault occurs, the step S100 is performed.

[0073] Optionally, the dominant fault detection on the target solid state disk can be performed by hard disk inspection or application load inclination. It can be understood that if the dominant fault detection result is that a dominant fault occurs, a preset dominant fault alarm can be output.

[0074] According to the embodiment of the present disclosure, by performing the dominant fault detection on the target solid state disk in advance, it can be determined whether the target solid state disk needs to be detected for the recessive fault, so as to avoid waste of relevant computer resources.

[0075] Optionally, based on the dominant fault detection result, the recessive fault detection on the target solid state disk can be performed. Figure 1 As shown in the method, the recessive fault detection on the target solid state disk can be performed by the following steps. Figure 4 As shown in the method, another embodiment of the recessive fault detection method of the solid state disk provided by the embodiment of the present disclosure can further include the following steps.

[0076] S500, if the recessive fault detection result is that a recessive fault occurs, outputting a preset recessive fault warning.

[0077] The preset recessive fault report can include the recessive fault detection result and a pre-warning suggestion. Optionally, if the recessive fault detection result is that no recessive fault occurs, the embodiment of the present disclosure can not perform any processing.

[0078] According to the embodiment of the present disclosure, by outputting the preset recessive fault warning, the maintenance personnel can be reminded in time to maintain the solid state disk with the recessive fault, so as to ensure normal operation of the relevant server cluster.

[0079] This disclosure provides a method for detecting latent faults in solid-state drives (SSDs). The method involves obtaining the target SMART log of the target SSD; extracting target SMART data collected at multiple acquisition times from the target SMART log; constructing target time-series feature data with temporal characteristics from the target SMART data collected at each acquisition time according to the acquisition time sequence; and inputting the target time-series feature data into a pre-trained recurrent neural network (RNN) model to obtain the latent fault detection result output by the RNN model. This disclosure, by constructing target time-series feature data with temporal characteristics from the collected target SMART data and then using a RNN model based on this data, obtains accurate and effective latent fault detection results, thereby helping to promptly detect latent faults in SSDs and prevent data loss.

[0080] Although the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous.

[0081] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0082] Corresponding to the above method embodiments, this disclosure also provides a solid-state drive latent fault detection device, the structure of which is as follows: Figure 5 As shown, it may include: a target SMART log acquisition unit 100, a target SMART data extraction unit 200, a target time series feature data construction unit 300, and a hidden fault detection result acquisition unit 400.

[0083] The target SMART log acquisition unit 100 is used to acquire the target SMART log of the target solid-state drive.

[0084] The target SMART data extraction unit 200 is used to extract target SMART data collected at multiple acquisition times from the target SMART log.

[0085] The target time-series feature data construction unit 300 is used to construct target time-series feature data with time-series features from the target SMART data collected at each collection time according to the collection time sequence.

[0086] The latent fault detection result acquisition unit 400 is used to input the target time series feature data into a pre-trained time recurrent neural network model to obtain the latent fault detection result output by the time recurrent neural network model.

[0087] Optionally, the solid state disk implicit failure detection apparatus can further include a model training unit. The model training unit includes a historical full SMART log obtaining subunit, a training SMART data extraction subunit, a training time sequence feature data construction subunit, and a model obtaining subunit.

[0088] The historical full SMART log obtaining subunit is configured to obtain a historical full SMART log of a training solid state disk that has occurred an explicit failure.

[0089] The training SMART data extraction subunit is configured to extract training SMART data collected at a plurality of collection time points from the historical full SMART log.

[0090] The training time sequence feature data construction subunit is configured to construct the training SMART data collected at the plurality of collection time points into training time sequence feature data having time sequence features according to a collection time sequence, and add a sample label to the training time sequence feature data, the sample label indicating a time point of occurrence of an implicit failure of the training solid state disk.

[0091] The model obtaining subunit is configured to input the training time sequence feature data having the sample label into a time recurrent neural network model for machine learning, and obtain a trained time recurrent neural network model.

[0092] Optionally, the solid state disk implicit failure detection apparatus can further include an implicit failure occurrence time point determination unit.

[0093] The implicit failure occurrence time point determination unit is configured to determine a time point of occurrence of an implicit failure of a training solid state disk according to a time point of occurrence of an explicit failure of the training solid state disk and a preset time threshold.

[0094] Optionally, the target SMART data and the training SMART data include at least one of a bottom layer data read error rate, a start / stop count, a remapping sector number, a power-on time accumulation, a spindle spin retry count, a hard disk calibration retry count, a hard disk power-on count, an uncorrectable error report, a command timeout, an offline uncorrectable sector count, a temperature, an ULTRA-ATA access check error rate, and a write error rate.

[0095] Optionally, the solid state disk implicit failure detection apparatus can further include an explicit failure detection result obtaining unit.

[0096] The explicit failure detection result obtaining unit is configured to perform explicit failure detection on a target solid state disk, obtain an explicit failure detection result, and trigger the target SMART log obtaining unit 100 in a case where the explicit failure detection result is that no explicit failure has occurred.

[0097] Optionally, the solid state disk implicit failure detection device can further comprise an implicit failure alarm output unit.

[0098] The implicit failure alarm output unit is configured to output a preset implicit failure warning when the implicit failure detection result indicates that an implicit failure occurs.

[0099] The solid state disk implicit failure detection device provided by the present disclosure can obtain target SMART logs of a target solid state disk, extract target SMART data collected at multiple collection time points from the target SMART logs, construct the target SMART data collected at the multiple collection time points into target time sequence feature data having time sequence features according to a collection time sequence, input the target time sequence feature data into a time recurrent neural network model that is pre-trained, and obtain an implicit failure detection result output by the time recurrent neural network model. The present disclosure can help to discover implicit failures of a solid state disk in a timely manner and avoid data loss by constructing the collected target SMART data into target time sequence feature data having time sequence features and then using a time recurrent neural network model to obtain accurate and effective implicit failure detection results according to the target time sequence feature data having time sequence features.

[0100] As to the device in the above-mentioned embodiments, the specific manners in which the modules perform operations have been described in detail in the embodiments of the method, and thus will not be described here in detail.

[0101] The solid state disk implicit failure detection device comprises a processor and a memory, and the target SMART log obtaining unit 100, the target SMART data extracting unit 200, the target time sequence feature data constructing unit 300, and the implicit failure detection result obtaining unit 400 are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions.

[0102] The processor comprises a core, and the core retrieves the corresponding program units from the memory. The core can be set to one or more, and the core parameters can be adjusted to construct the collected target SMART data into target time sequence feature data having time sequence features, and then use a time recurrent neural network model to obtain accurate and effective implicit failure detection results according to the target time sequence feature data having time sequence features, thereby helping to discover implicit failures of a solid state disk in a timely manner and avoiding data loss.

[0103] The present disclosure provides a computer readable storage medium having a program stored thereon, and the program is executed by a processor to implement the solid state disk implicit failure detection method.

[0104] The embodiment of the present disclosure provides a processor used for running a program, wherein the program performs the solid state disk implicit fault detection method when running.

[0105] As shown in Figure 6 The embodiment of the present disclosure provides an electronic device 500, which comprises at least one processor 501, at least one memory 502 connected with the processor 501, and a bus 503; wherein the processor 501 and the memory 502 complete mutual communication through the bus 503; the processor 501 is used for calling program instructions in the memory 502, so as to execute the solid state disk implicit fault detection method described above. The electronic device in the present embodiment can be a server, a PC, a PAD, a mobile phone and the like.

[0106] The present disclosure further provides a computer program product adapted to execute the program initialized with the steps of the solid state disk implicit fault detection method when executed on an electronic device.

[0107] The present disclosure is described with reference to the flowcharts and / or block diagrams of the method, device (system), electronic device and computer program product according to the embodiments of the present disclosure. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor or other programmable devices to produce a machine, so that the instructions executed by the computer or other programmable devices produce a device for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The device for implementing the functions specified in one flow or multiple flows and / or blocks Figure 1 The device for implementing the functions specified in one flow or multiple flows and / or blocks

[0108] In a typical configuration, the electronic device comprises one or more processors (CPU), memories and buses. The electronic device can further comprise input / output interfaces, network interfaces and the like.

[0109] The memory can comprise a non-permanent memory in a computer readable medium, random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM), and the memory comprises at least one memory chip. The memory is an example of the computer readable medium.

[0110] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition herein, computer-readable media does not include transitory computer-readable media, such as modulated data signals and carriers.

[0111] In the description of the present disclosure, it should be understood that the orientation or positional relationship indicated by the terms "upper", "lower", "front", "back", "left" and "right" and the like is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the indicated position or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present disclosure.

[0112] It should be noted that in this paper, the relationship terms such as first and second are only used 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 the entities or operations. It should also be noted that the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, product or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, product or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, product or device including the element.

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

[0114] The above merely provides an example of the present disclosure, but is not intended to limit the present disclosure. The present disclosure can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present disclosure should be included in the scope of claims of the present disclosure.

Claims

1. A method for detecting latent faults in a solid-state drive, characterized in that, The method comprises the following steps: obtaining a target SMART log of a target solid state disk; extracting target SMART data collected at multiple collection time points from the target SMART log; constructing the target SMART data collected at the multiple collection time points into target time sequence feature data with time sequence characteristics according to a collection time sequence; inputting the target time sequence feature data into a pre-trained time recurrent neural network model to obtain a hidden fault detection result output by the time recurrent neural network model; wherein the training process of the time recurrent neural network model comprises: obtaining a historical full-amount SMART log of a training solid state disk that has occurred a dominant fault; extracting training SMART data collected at multiple collection time points from the historical full-amount SMART log; constructing the training SMART data collected at the multiple collection time points into training time sequence feature data with time sequence characteristics according to a collection time sequence, and adding a sample label to the training time sequence feature data, the sample label indicating a hidden fault occurrence time of the training solid state disk; inputting the training time sequence feature data with the sample label into the time recurrent neural network model for machine learning to obtain the trained time recurrent neural network model.

2. The method of claim 1, wherein, Further comprising: determining the hidden fault occurrence time of the training solid state disk according to the dominant fault occurrence time of the training solid state disk and a preset time threshold.

3. The method of claim 1, wherein, The target SMART data and the training SMART data comprise at least one of the following data: a bottom layer data read error rate, a start / stop count, a remapping sector number, a power-on time accumulation, a spindle spin retry count, a hard disk calibration retry count, a hard disk power-on count, an uncorrectable error report, a command timeout, an offline uncorrectable sector count, a temperature, an ULTRA-ATA access check error rate, and a write error rate.

4. The method of claim 1, wherein, Before the step of obtaining the target SMART log of the target solid state disk, the method further comprises: performing dominant fault detection on the target solid state disk to obtain a dominant fault detection result; in a case where the dominant fault detection result is that no dominant fault has occurred, performing the step of obtaining the target SMART log of the target solid state disk.

5. The method of claim 1, wherein, Further comprising: in a case where the hidden fault detection result is that a hidden fault has occurred, outputting a preset hidden fault warning.

6. A solid state drive implicit failure detection apparatus, comprising: The method comprises the following steps: a target SMART log obtaining unit, a target SMART data extracting unit, a target time sequence feature data constructing unit, a hidden fault detection result obtaining unit, and a model training unit, the target SMART log obtaining unit is configured to obtain a target SMART log of a target solid state disk; the target SMART data extracting unit is configured to extract target SMART data collected at multiple collection time points from the target SMART log; the target time sequence feature data constructing unit is configured to construct the target SMART data collected at the multiple collection time points into target time sequence feature data with time sequence characteristics according to a collection time sequence; the hidden fault detection result obtaining unit is configured to input the target time sequence feature data into a pre-trained time recurrent neural network model to obtain a hidden fault detection result output by the time recurrent neural network model; The implicit failure detection result obtaining unit is configured to input the target time sequence feature data into a pre-trained time recurrent neural network model to obtain an implicit failure detection result output by the time recurrent neural network model. The model training unit comprises a historical full-amount SMART log obtaining subunit, a training SMART data extracting subunit, a training time sequence feature data constructing subunit, and a model obtaining subunit. The historical full-amount SMART log obtaining subunit is configured to obtain a historical full-amount SMART log of a training solid state disk that has occurred an explicit failure. The training SMART data extracting subunit is configured to extract training SMART data collected at a plurality of collection time points from the historical full-amount SMART log. The training time sequence feature data constructing subunit is configured to construct the training SMART data collected at the plurality of collection time points into training time sequence feature data with time sequence features according to a collection time sequence, and add a sample label to the training time sequence feature data, the sample label indicating a time point of occurrence of an implicit failure of the training solid state disk. The model obtaining subunit is configured to input the training time sequence feature data with the sample label added thereto into the time recurrent neural network model for machine learning to obtain the trained time recurrent neural network model.

7. A computer-readable storage medium having stored thereon a program, characterized in that, The program is executed by the processor to implement the solid state disk implicit failure detection method in any one of claims 1 to 5.

8. An electronic device, comprising at least one processor, and at least one memory connected to the processor via a bus; wherein, The processor, the memory and the bus complete communication among each other. The processor is configured to invoke program instructions in the memory to execute the solid state disk implicit failure detection method in any one of claims 1 to 5. The processor, the memory and the bus complete communication among each other. The processor is configured to invoke program instructions in the memory to execute the solid state disk implicit failure detection method in any one of claims 1 to 5.

Images