Solid-state hard disk life monitoring and early warning system and method based on performance analysis

By connecting to the solid-state drive management system, obtaining historical fault event samples and training a life-decreasing warning model, monitoring the SSD failure event data in real time, solving the comprehensive problem of SSD life monitoring and early warning technology in the existing technology, and realizing the life of SSDs Accurate and timely assessment and early warning ensure data security and system stability.

CN118642921BActive Publication Date: 2024-11-15JIANGSU HUACUN ELECTRONICS TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411111180.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-11-15
Estimated Expiration
2044-08-14

AI Technical Summary

Technical Problem

The existing solid-state drive life monitoring and early warning technology is insufficient in data collection, fault classification, model training and real-time monitoring and early warning, and cannot comprehensively, accurately and timely evaluate and predict the life of solid-state drives.

Method used

通过连接固态硬盘管理系统获取历史故障事件样本,按故障类型聚类得到多类故障事件样本,训练寿命下降预警模型,实时监测固态硬盘故障事件数据,识别寿命下降梯度,并生成寿命预警信号。

Benefits of technology

It realizes comprehensive, accurate and timely assessment and early warning of the life of solid-state drives, ensuring data security and stable system operation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118642921B_ABST
    Figure CN118642921B_ABST
Patent Text Reader

Abstract

The present invention discloses a solid-state hard disk life monitoring and early warning system and method based on performance analysis, and relates to the field of computer storage technology. The system comprises: connecting to a solid-state hard disk management system to obtain historical fault event samples containing data of each stage of the fault, clustering multiple types of fault event samples according to the fault type, and training a life reduction early warning model. Real-time monitoring of solid-state hard disk fault event data, inputting a model to identify the life reduction gradient, and generating a life warning signal if the real-time life reduction gradient is greater than the matching reduction warning gradient. The system solves the technical problem that the existing solid-state hard disk life monitoring and early warning technology is not comprehensive enough in data collection, fault classification, model training, and real-time monitoring and early warning, and cannot comprehensively, accurately, and timely evaluate and predict the life of the solid-state hard disk, and realizes comprehensive, accurate, and timely evaluation and early warning of the life of the solid-state hard disk, achieving the technical effect of ensuring data security and stable system operation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of computer storage technology, and in particular to a solid-state hard disk life monitoring and early warning system and method based on performance analysis. Background Art

[0002] In the use scenario of SSDs, the accuracy of life monitoring and early warning is particularly prominent, and the contradictions in related technical requirements are relatively more prominent. Accurate and efficient SSD life monitoring and early warning can better meet users' needs for data security and system stability, and become a crucial link in solving the reliable use of SSDs. Traditional SSD life monitoring and early warning are often one-sided and extensive, relying only on partial fault data or simple indicator judgments, and lack of comprehensive analysis of SSD failures. There is a lack of in-depth use of data before and after the failure occurs, making it difficult to accurately and meticulously evaluate the life of SSDs. In the establishment of early warning models, there are unscientific situations, resulting in untimely early warnings in some key stages, and unable to cope well with the complex and changeable actual use of SSDs.

[0003] At present, relevant technologies still have the problem that the SSD life monitoring and early warning technologies are not comprehensive enough in terms of data collection, fault classification, model training, and real-time monitoring and early warning, and cannot comprehensively, accurately, and timely evaluate and predict the life of SSDs. Summary of the invention

[0004] This application obtains historical fault event samples containing data of each stage of the fault by connecting to the solid-state drive management system, and obtains multiple types of fault event samples by clustering according to the fault type, so as to train the life reduction warning model. The solid-state drive fault event data is monitored in real time, and the input model is used to identify the life reduction gradient. If the real-time life reduction gradient is greater than the matching reduction warning gradient, a life warning signal is generated, achieving the technical effect of ensuring data security and stable system operation.

[0005] The present application provides a solid state drive life monitoring and early warning system based on performance analysis, the system comprising:

[0006] A historical fault event sample acquisition module, the historical fault event sample acquisition module is used to connect to the solid-state hard disk management system to obtain historical fault event samples, wherein each fault event sample records data before the fault occurs, data when the fault occurs, data when the fault recovers, and data after the fault recovers; a fault event sample output module, the fault event sample output module is used to cluster the historical fault event samples according to the fault type, and output multiple types of fault event samples; a life reduction warning model output module, the life reduction warning model output module is used to train the life reduction warning model according to the multiple types of fault event samples, and output the trained life reduction warning model; a real-time life reduction gradient output module, the real-time life reduction gradient output module is used to monitor the real-time fault event data of the solid-state hard disk in real time, input the real-time fault event data into the life reduction warning model to identify the life reduction gradient, and output the real-time life reduction gradient; a life warning signal generation module, the life warning signal generation module is used to generate a life warning signal when the real-time life reduction gradient is greater than the matching reduction warning gradient of the real-time fault event data.

[0007] This application also provides a solid state drive life monitoring and early warning method based on performance analysis, including:

[0008] Connect to the solid-state drive management system to obtain historical fault event samples, wherein each fault event sample records data before the fault occurs, data when the fault occurs, data when the fault recovers, and data after the fault recovers; cluster the historical fault event samples according to the fault type, and output multiple types of fault event samples; train a life reduction warning model based on the multiple types of fault event samples, and output the trained life reduction warning model; monitor the real-time fault event data of the solid-state drive in real time, input the real-time fault event data into the life reduction warning model to identify the life reduction gradient, and output the real-time life reduction gradient; when the real-time life reduction gradient is greater than the matching reduction warning gradient of the real-time fault event data, generate a life warning signal

[0009] The solid-state hard disk life monitoring and early warning system and method based on performance analysis proposed in this application is proposed. By connecting to the solid-state hard disk management system, historical fault event samples containing data of each stage of the fault are obtained, and multiple types of fault event samples are clustered according to the fault type to train the life reduction early warning model. The solid-state hard disk fault event data is monitored in real time, and the input model is used to identify the life reduction gradient. If the real-time life reduction gradient is greater than the matching reduction warning gradient, a life warning signal is generated. This solves the comprehensive deficiencies of the existing solid-state hard disk life monitoring and early warning technology in data collection, fault classification, model training, and real-time monitoring and early warning, and the technical effect of being unable to comprehensively, accurately, and timely evaluate and predict the life of the solid-state hard disk. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0011] Figure 1 A schematic diagram of the structure of a solid state drive life monitoring and early warning system based on performance analysis provided in an embodiment of the present application;

[0012] Figure 2 A flowchart of a solid-state drive life monitoring and early warning method based on performance analysis provided in an embodiment of the present application.

[0013] Explanation of the reference numerals: historical fault event sample acquisition module 10 , fault event sample output module 20 , life reduction warning model output module 30 , real-time life reduction gradient output module 40 , life warning signal generation module 50 . DETAILED DESCRIPTION

[0014] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0015] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0016] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0017] The present application embodiment provides a solid state drive life monitoring and early warning system based on performance analysis, such as Figure 1 As shown, the system comprises:

[0018] The historical fault event sample acquisition module 10 is used to connect to the solid state hard disk management system to obtain historical fault event samples, wherein each fault event sample records data before the fault occurs, data when the fault occurs, data when the fault is recovered, and data after the fault is recovered. Specifically, a connection channel with the solid state hard disk management system is established, and through the connection, historical fault event samples inside the system are obtained, and the sample information includes: the data before the fault occurs records various parameters and indicators of the hard disk in normal operation, including: total number of written bytes, wear leveling count, number of bad blocks, ECC errors, etc., wherein the total number of written bytes reflects the total amount of write operations borne by the solid state hard disk before the fault occurs, and evaluates the write load degree of the hard disk in the normal use stage, the wear leveling count shows the number of adjustment operations performed by the hard disk to balance the wear of each storage unit, and can reflect the execution of the hard disk's wear leveling strategy, the number of bad blocks can intuitively indicate the number of damaged storage areas that already existed before the fault occurs, so as to understand the initial health of the hard disk, and the ECC error rate can reflect the error correction situation during data transmission and storage, and the error rate can indirectly reflect the stability of the data. The data at the time of failure captures the abnormal performance of the hard disk at the moment of failure, including sudden read and write errors, freezes, system errors and other information. The change in the total number of written bytes can reflect whether the write activity at the moment of failure is abnormal. The sudden change in the wear leveling count indicates that the working mode of the hard disk at the moment of failure is abnormal. The increase in the number of bad blocks can intuitively reflect the degree of immediate damage to the storage area caused by the failure. The sharp increase in the ECC error rate indicates that the accuracy of the data is seriously threatened when the failure occurs. The data at the time of failure recovery reflects the process of the hard disk's transition from the failure state to the recovery state, such as the recovery measures taken and the system's response. The recovery trend of the total number of written bytes can determine whether there are abnormal write operations during the recovery process. The adjustment of the wear leveling count can reflect the hard disk's re-balancing strategy for storage units during the recovery process. The change in the number of bad blocks can show the effect of the recovery measures on the damaged area. The gradual decrease or stabilization of the ECC error rate indicates that the data's error correction capability is gradually returning to normal during the recovery process. The data after fault recovery presents the final state of the hard disk after the recovery operation, such as whether it has completely recovered to normal and whether there is any performance loss. Among them, the final value of the total number of written bytes can be compared with the value before the fault to evaluate the write performance after recovery. The final state of the wear leveling count can indicate whether the overall wear leveling effect of the hard disk meets expectations. The remaining number of bad blocks reflects the damaged areas that still exist after recovery. The stable low value of the ECC error rate means that the accuracy of the hard disk data after recovery is guaranteed. The acquisition and integration of data provides a data basis for subsequent analysis and model training.

[0019] The fault event sample output module 20 is used to cluster the historical fault event samples according to the fault type and output multiple types of fault event samples. Specifically, the fault event sample output module 20 receives the historical fault event samples obtained from the historical fault event sample acquisition module 10, and uses specific algorithms and logics to conduct in-depth analysis and classification of these samples based on the key feature of the type of fault. In the clustering process, the fault characteristics, manifestations, related parameters and other key elements contained in each sample are compared. For example, similar types of fault events are grouped together according to whether the fault is caused by hardware damage, software conflict, power problem or other specific reasons. Through classification operations, multiple clearly distinguished fault event sample categories are finally output. The clearly classified multiple types of fault event samples provide a clear and targeted data foundation for subsequent more in-depth model training and analysis, so that subsequent processing can be more focused on specific types of faults, improving the processing efficiency and accuracy of the entire system. The fault event sample output module 20 provides strong support for the further optimization and accurate prediction of the entire solid-state hard disk life monitoring and early warning system through accurate clustering of historical fault event samples.

[0020] The life reduction warning model output module 30 is used to train the life reduction warning model according to the multiple types of fault event samples and output the trained life reduction warning model. Specifically, the life reduction warning model output module 30 receives multiple types of fault event samples generated by clustering by the fault event sample output module 20. Each sample contains a series of characteristic data such as the frequency of fault occurrence, the severity of the fault (which can be quantified by the degree of impact on hard disk performance, etc.), the time required for fault recovery, and the performance index after fault recovery. For each sample, the life reduction warning model output module 30 calculates a comprehensive evaluation index. The comprehensive evaluation index is obtained by multiplying the frequency of fault occurrence, the severity of the fault, the time required for fault recovery, and the performance index after fault recovery by their respective corresponding weight coefficients and then adding them together. The sample data is trained using a machine learning algorithm. Adjusting internal parameters aims to reduce the error between the predicted value and the actual value. For example, the mean square error is used as an error evaluation indicator. For each training iteration, the mean square error is calculated, that is, the difference between the predicted life reduction value and the actual life reduction value is first found, and then squared. These square differences of all samples are summed and divided by the number of samples. By continuously adjusting the model parameters, the mean square error is gradually reduced until it reaches a satisfactory level or reaches a pre-set number of training times. After repeated iterations and optimizations, the life reduction warning model is finally output. The model can accurately predict the life reduction of the solid-state drive based on the input new fault event feature data.

[0021] A real-time life decline gradient output module 40 is used to monitor the real-time fault event data of the solid state drive in real time, input the real-time fault event data into the life decline warning model to identify the life decline gradient, and output the real-time life decline gradient. Specifically, the real-time life reduction gradient output module 40 continuously monitors the operating status of the solid-state hard disk in real time to obtain real-time fault event data, which includes but is not limited to the read and write operation status of the hard disk, data transmission speed, error frequency and other information. Once the real-time fault event data is obtained, the module will quickly input it into the previously trained life reduction warning model. After receiving these data, the real-time life reduction gradient output module 40 analyzes and processes the data. The real-time life reduction gradient output module 40 compares and matches the features in the real-time data with the features of various fault modes and life reduction trends learned during previous training, identifies the degree and trend of the hard disk life reduction reflected by the current data, and thus calculates the life reduction gradient. The real-time life reduction gradient output module 40 outputs the calculated real-time life reduction gradient. The real-time gradient information can timely reflect the health status and life loss rate of the solid-state hard disk at the current moment, and provides a key basis for subsequent life warnings and corresponding maintenance decisions. The real-time life reduction gradient output module 40 realizes instant evaluation and feedback of the life status of the solid-state hard disk through real-time monitoring, data input model analysis and gradient calculation output.

[0022] The life warning signal generating module 50 is used to generate a life warning signal when the real-time life decline gradient is greater than the matching decline warning gradient of the real-time fault event data. Specifically, during the operation of the system, the real-time life decline gradient output module 40 continuously outputs the real-time life decline gradient, and determines the decline warning gradient matching it according to the real-time fault event data. When the life warning signal generation module 50 receives the relevant information of the real-time life decline gradient and the matching decline warning gradient, it immediately compares and judges. If the real-time life decline gradient is greater than the matching decline warning gradient, it means that the life loss rate of the solid-state hard disk exceeds the preset safety threshold, and it will face serious life problems in the near future. The life warning signal generation module 50 responds quickly and starts the corresponding mechanism to generate a life warning signal. The signal can be a specific sound and light prompt, a system pop-up notification, or a specific instruction code sent to a related monitoring system. The generated life warning signal will be promptly transmitted to relevant personnel or systems so that they can quickly take corresponding measures, such as backing up data, arranging hard disk replacement or maintenance plans, etc., so as to minimize data loss or system failures that may be caused by the exhaustion of the hard disk life. The life warning signal generation module 50 generates an effective warning signal in time by accurately comparing and judging the real-time life decline gradient and the matching decline warning gradient, which plays an important role in ensuring the stable operation and data security of the solid-state hard disk.

[0023] The solid-state hard disk life monitoring and early warning system based on performance analysis according to the embodiment of the present invention is used to solve the technical problem that the existing solid-state hard disk life monitoring and early warning technology is not comprehensive enough in data collection, fault classification, model training, and real-time monitoring and early warning, and cannot comprehensively, accurately, and timely evaluate and predict the life of the solid-state hard disk, so as to achieve a comprehensive, accurate, and timely evaluation and early warning of the life of the solid-state hard disk, and achieve the technical effect of ensuring data security and stable operation of the system. The solid-state hard disk life monitoring and early warning system based on performance analysis includes: a historical fault event sample acquisition module 10, a fault event sample output module 20, a life reduction early warning model output module 30, a real-time life reduction gradient output module 40, and a life warning signal generation module 50.

[0024] The specific configuration of the fault event sample output module 20 will be described in detail below. The fault event sample output module 20 may further include: performing secondary fault event analysis on the multiple types of fault event samples and constructing a fault association graph. Specifically, a secondary fault event analysis is carried out for the classified multiple types of fault event samples, taking full account of possible chain reactions and mutual influences. Taking power failure as an example, after the power failure occurs, subsequent problems are triggered, such as common controller errors. For each type of fault event sample, the secondary fault events that may be caused are studied, and a fault association graph is constructed based on the analysis. The fault association graph shows the relationship between different types of initial faults and possible derived secondary faults. It also includes identifying the secondary fault event triggering probability of the multiple types of fault event samples, probabilistically identifying the fault association graph according to the secondary fault event triggering probability, and outputting an identified fault association identification graph. Specifically, data analysis and statistical methods are used to identify the triggering probability of secondary fault events in multiple types of fault event samples. For example, by studying a large amount of historical data, the probability of the occurrence of a secondary fault event, a controller error, after a power failure is determined. According to the identified triggering probability of the secondary fault event, accurate probability identification is performed on the constructed fault association graph. After identification, an identification fault association identification graph of the possibility of each secondary fault event is output. It also includes connecting the identification fault association identification graph with the life reduction warning model. Specifically, the identification fault association identification graph is closely connected with the life reduction warning model. Due to the influence of cascading failures, when the life reduction warning model receives new fault event data, it can refer to the identification fault association identification graph to more comprehensively consider the potential secondary fault possibility. For example, when a power failure is detected, the model will make early warnings and assessments on possible secondary faults based on the triggering probability of the controller error in the identification graph, take countermeasures, and enhance the stability and reliability of the system. Through secondary fault event analysis, accurate probability identification, and effective connection with the life reduction warning model, it can better cope with the challenges brought by cascading failures and improve the prediction and response capabilities of solid-state hard disk failures.

[0025] The specific configuration of the fault event sample output module 20 will be described in detail below. The fault event sample output module 20 may further include: inputting the real-time fault event data into the life reduction warning model, performing secondary fault identification in the identification fault association identification map according to the matching class fault event of the real-time fault event data, and outputting multiple secondary fault events. Specifically, the real-time fault event data is input into the life reduction warning model, the model analyzes the data, determines the class fault event that matches it, and carries out the identification of secondary faults in the identification fault association identification map according to the matching class fault event. For example, when a specific fault (such as a power short circuit fault) occurs, the position in the identification fault association identification map is determined. Since there is a dependency relationship between faults, the initial fault will transmit the impact along the associated path. Taking the power short circuit fault as an example, it will cause voltage instability, and then transmit the impact to the storage controller through the associated line in the identification fault association identification map, causing the storage controller to work abnormally, and be identified as a secondary fault event. Through transmission and analysis, multiple similar secondary fault events are output. It also includes obtaining multiple life reduction gradients corresponding to the multiple secondary fault events. Specifically, for the identified multiple secondary failure events, the corresponding life decline gradients are obtained, and the gradient reflects the individual impact of each secondary failure event on the overall life of the solid state drive. It also includes feeding back the real-time life decline gradients with the trigger probabilities corresponding to the multiple life decline gradients as weights, and outputting the optimized real-time life decline gradients. Specifically, the trigger probability corresponding to each secondary fault event is obtained, and the trigger probability is used as a weight to feedback-adjust the real-time life degradation gradient. The higher the trigger probability, the more likely the secondary fault event is to occur, and the greater its weight in the feedback adjustment. For example, if the trigger probability of a secondary fault event is 80%, and its corresponding life degradation gradient is large, then a larger weight will be given in the calculation to more significantly affect the adjustment of the real-time life degradation gradient. After the above-mentioned weighted calculation and feedback adjustment based on the trigger probability, the optimized real-time life degradation gradient is output. The optimized gradient comprehensively considers the impact of the initial fault and multiple secondary fault events that may occur, and can more accurately and comprehensively reflect the actual life degradation of the solid-state hard disk, providing a more reliable basis for subsequent maintenance decisions and early warning measures. By inputting real-time fault event data into the model, identifying secondary faults in the fault association identification diagram, obtaining the life degradation gradient and performing feedback adjustment with the trigger probability as the weight, an accurate assessment of the life degradation of the solid-state hard disk is achieved, especially fully considering the transmission impact of the fault in the association diagram.

[0026] The specific configuration of the fault event sample output module 20 will be described in detail below. The fault event sample output module 20 can further include: obtaining the associated fault event set of the matching fault event. Specifically, after determining the matching fault event, all possible fault events associated with it are obtained to form an associated fault event set. For example, when a matching fault event such as a hard disk read / write head failure occurs, through the analysis of the fault dependency graph, it is found that the possible fault events associated with it include data transmission errors, storage chip damage, etc., which constitute an associated fault event set. It also includes identifying the transfer probability set of the matching fault event to the associated fault event set one by one. Specifically, the probability of this matching fault event transferring to each event in the associated fault event set is identified to form a transfer probability set. Taking the hard disk read / write head failure as an example, the probability of transferring to a data transmission error may be 60%, and the probability of transferring to a storage chip damage may be 40%. These probability values ​​constitute the transfer probability set. It also includes outputting the events in the transfer probability set that are greater than the preset transfer probability as secondary fault events. Specifically, a preset transfer probability is set, such as 50%, and events greater than the preset value in the transfer probability set are screened out. For example, in the above example, since the probability of 60% of the hard disk read / write head failure transferring to a data transmission error is greater than the preset transfer probability of 50%, the data transmission error will be output as a secondary fault event. When an initial fault occurs, the secondary fault events that may follow are predicted and determined through the association relationship and transfer possibility between faults represented in the fault dependency graph. The initial fault is like a trigger point, and its influence is transmitted along the associated path in the fault dependency graph, resulting in an increased possibility of failure in other associated parts. By analyzing and screening the transfer probability, secondary fault events with a higher probability of occurrence are accurately output, providing an important basis for subsequent early warning and processing. By obtaining a set of associated fault events, identifying a set of transfer probabilities and screening based on the preset transfer probability, accurate prediction and output of secondary fault events are achieved, fully considering the transmission and evolution of faults in the dependency graph.

[0027] The specific configuration of the life reduction warning model output module 30 will be described in detail below. The life reduction warning model output module 30 can further include: according to the multiple types of fault event samples, obtaining the stage feature vector and life reduction gradient label corresponding to each type of fault event sample. Specifically, the multiple types of fault event samples obtained by clustering are analyzed, and the corresponding stage feature vector is extracted from each type of fault event sample. The feature vector contains key information of the fault at different stages, such as the system state before the fault occurs, the specific performance when the fault occurs, and the various indicators in the fault recovery process. The life reduction gradient label is set for each type of fault event sample to clarify the impact of this type of fault on the life of the hard disk. It also includes obtaining a life reduction warning model according to the stage feature vector and the life reduction gradient label. Specifically, the life reduction warning model is constructed based on the extracted stage feature vector and the life reduction gradient label. The model can predict the life reduction according to the newly input fault feature through the intrinsic relationship between the feature and the label. It also includes obtaining the sensitivity factor corresponding to each type of fault event sample according to the stage feature vector and the life reduction gradient label. Specifically, according to the same stage feature vector and life decline gradient label, further analysis is performed to obtain the sensitive factors corresponding to each type of fault event sample. The sensitive factors can reflect which features are more critical and sensitive for judging life decline. It also includes configuring multiple types of decline warning gradients according to the sensitive factors, and storing the multiple types of decline warning gradients in the life decline warning model. Specifically, multiple types of decline warning gradients are configured according to the obtained sensitive factors. Different sensitive factors will lead to different warning gradient settings. The carefully configured multiple types of decline warning gradients are stored in the life decline warning model, so that the model can accurately give the corresponding warning level according to the specific fault situation when making predictions and warnings. Through the step-by-step analysis and processing of multiple types of fault event samples, key information is obtained, and the warning gradients are configured and stored in the model, so that the life decline warning model can play a more accurate and effective role.

[0028] The specific configuration of the real-time life reduction gradient output module 40 will be described in detail below. The real-time life reduction gradient output module 40 may further include: the life reduction warning model performs category judgment on the real-time fault event data to obtain matching class fault event samples. Specifically, when the real-time fault event data is input into the life reduction warning model, the model performs category judgment on the data, compares and analyzes the characteristics of the real-time fault event data with the characteristics of the existing various types of fault event samples, so as to determine the matching class fault event samples. For example, if the real-time fault event data shows that the read and write operations frequently fail and the speed drops significantly when the fault occurs, the model may classify it as a fault category related to the read and write channel through analysis and judgment, and obtain the corresponding matching class fault event samples. It also includes, according to the matching class fault event samples, obtaining the matching decline warning gradient. Specifically, according to the obtained matching class fault event samples, further determine the corresponding matching decline warning gradient, and the gradient reflects the degree of influence of this type of fault on the life of the hard disk. It also includes, the life reduction warning model analyzes the real-time fault event data to obtain data recovery integrity, recovery time change, and read and write performance loss. Specifically, the life reduction warning model will conduct an in-depth analysis of real-time fault event data. In the event of a fault, it will pay special attention to the completeness of data recovery, that is, the proportion of successfully recovered data to the total amount of damaged data; the change in recovery time, including the difference between the time from the occurrence of the fault to the start of recovery and the time spent on complete recovery and the normal situation; and the loss of read and write performance, such as the difference between the read and write speed and accuracy and the normal state. It also includes obtaining the real-time life reduction gradient based on the data recovery completeness, the change in recovery time, and the read and write performance loss. Specifically, the real-time life reduction gradient is calculated by comprehensively considering the key factors such as data recovery completeness, recovery time change and read-write performance loss. Taking a fault caused by power fluctuation as an example, some data may be lost under the fault, and the recovery completeness is low; the recovery process takes a long time, beyond the normal range; the read and write speeds are significantly reduced, and the read and write resistance is weakened. A real-time life reduction gradient is calculated to reflect the great impact of the fault on the life of the hard disk. Through category judgment of real-time fault event data, matching gradient acquisition, detailed analysis and comprehensive calculation, the real-time life reduction gradient can be accurately obtained, especially after considering key factors such as read and write resistance under faults, which provides an important basis for timely and effective early warning and maintenance.

[0029] Below, the specific configuration of the real-time life decline gradient output module 40 will be described in detail. The real-time life decline gradient output module 40 can further include: the life decline warning model analyzes the real-time fault event data to obtain a real-time stage feature vector. Specifically, the life decline warning model receives real-time fault event data, and the data covers each stage from before the fault occurs, when it occurs, when it recovers, and after the recovery. During the analysis process, the model will extract multiple key stage feature vectors respectively. It also includes, wherein the stage feature vector includes the first data recovery completeness, the first recovery time change, and the first read-write performance loss between the data before the fault occurs and the data when the fault occurs, the second data recovery completeness, the second recovery time change, and the second read-write performance loss between the data when the fault occurs and the data when the fault is recovered, and the third data recovery completeness, the third recovery time change, and the third read-write performance loss between the data when the fault is recovered and the data after the fault is recovered. Specifically, for the data before the failure and the data at the time of the failure, the model will calculate the first data recovery completeness, that is, measure the completeness of the data that can be recovered from before the failure to the time of the failure; at the same time, it will also pay attention to the change of the first recovery time to understand the dynamic situation of the time spent on recovery at this stage; and the first read and write performance loss to evaluate the degree of impact on the read and write operation performance at this stage. For the data at the time of the failure and the data at the time of failure recovery, the model will also calculate the second data recovery completeness to clarify the completeness of the data recovery at this stage; determine the change of the second recovery time, monitor the change of the time required for recovery; and measure the second read and write performance loss to know the damage of the read and write performance at this stage. For the data at the time of failure recovery and the data after failure recovery, the model will calculate the third data recovery completeness to reflect the final data integrity of the recovery; examine the change of the third recovery time to grasp the final change trend of the recovery time; and evaluate the third read and write performance loss to clarify the status of the read and write performance after recovery. It also includes outputting the data recovery completeness according to the first data recovery completeness, the second data recovery completeness and the third data recovery completeness. Specifically, based on the data from the above three stages, in terms of data recovery completeness, according to the first, second and third data recovery completeness, a specific algorithm or weight distribution is used to output the final data recovery completeness result, thereby comprehensively reflecting the overall completeness of data recovery in the entire failure and recovery process. It also includes outputting the recovery time change according to the first recovery time change, the second recovery time change and the third recovery time change. Specifically, in terms of recovery time change, based on the first, second and third recovery time changes, a comprehensive analysis and calculation are performed to output the change in recovery time during the entire recovery process, so as to clearly understand the dynamic characteristics of the recovery time. It also includes outputting the read-write performance loss according to the first read-write performance loss, the second read-write performance loss and the third read-write performance loss.Specifically, in terms of read and write performance loss, according to the first, second and third read and write performance losses, after reasonable integration and calculation, the read and write performance loss situation in the whole process is output, which provides an important basis for evaluating the decline in system performance. Through analysis and calculation, the life degradation warning model can accurately evaluate the impact of fault events on the system, and provide strong support for subsequent optimization and improvement.

[0030] In the above, refer to Figure 1 The solid state drive life monitoring and early warning system based on performance analysis according to an embodiment of the present invention is described in detail. Figure 2 A solid state drive life monitoring and early warning method based on performance analysis according to an embodiment of the present invention is described.

[0031] Connect to the solid-state drive management system to obtain historical fault event samples, wherein each fault event sample records data before the fault occurs, data when the fault occurs, data when the fault recovers, and data after the fault recovers; cluster the historical fault event samples according to the fault type, and output multiple types of fault event samples; train a life degradation warning model based on the multiple types of fault event samples, and output the trained life degradation warning model; monitor the real-time fault event data of the solid-state drive in real time, input the real-time fault event data into the life degradation warning model to identify the life degradation gradient, and output the real-time life degradation gradient; when the real-time life degradation gradient is greater than the matching degradation warning gradient of the real-time fault event data, generate a life degradation warning signal.

[0032] In a possible implementation, the solid state drive life monitoring and early warning method based on performance analysis further includes:

[0033] According to the multiple types of fault event samples, the stage feature vectors and life degradation gradient labels corresponding to each type of fault event samples are obtained; according to the stage feature vectors and the life degradation gradient labels, a life degradation warning model is obtained; according to the stage feature vectors and the life degradation gradient labels, the sensitive factors corresponding to each type of fault event samples are obtained; according to the sensitive factors, multiple types of degradation warning gradients are configured, and the multiple types of degradation warning gradients are stored in the life degradation warning model.

[0034] In a possible implementation, the solid state drive life monitoring and early warning method based on performance analysis further includes:

[0035] The life degradation warning model performs category judgment on the real-time fault event data to obtain matching fault event samples; obtains a matching degradation warning gradient based on the matching fault event samples; the life degradation warning model analyzes the real-time fault event data to obtain data recovery completeness, recovery time change, and read-write performance loss; obtains a real-time life degradation gradient based on the data recovery completeness, the recovery time change, and the read-write performance loss.

[0036] In a possible implementation, the solid state drive life monitoring and early warning method based on performance analysis further includes:

[0037] The life reduction warning model analyzes the real-time fault event data to obtain a real-time stage feature vector; wherein the stage feature vector includes a first data recovery completeness, a first recovery time change, and a first read-write performance loss between the data before the fault occurs and the data when the fault occurs, a second data recovery completeness, a second recovery time change, and a second read-write performance loss between the data when the fault occurs and the data when the fault is recovered, and a third data recovery completeness, a third recovery time change, and a third read-write performance loss between the data when the fault is recovered and the data after the fault is recovered; according to the first data recovery completeness, the second data recovery completeness, and the third data recovery completeness, the data recovery completeness is output; according to the first recovery time change, the second recovery time change, and the third recovery time change, the recovery time change is output; according to the first read-write performance loss, the second read-write performance loss, and the third read-write performance loss, the read-write performance loss is output.

[0038] In a possible implementation, the solid state drive life monitoring and early warning method based on performance analysis further includes:

[0039] Perform secondary fault event analysis on the multiple types of fault event samples to construct a fault association graph; identify the secondary fault event triggering probability of the multiple types of fault event samples, perform probability identification on the fault association graph according to the secondary fault event triggering probability, and output an identified fault association identification graph; connect the identified fault association identification graph with the life reduction warning model.

[0040] In a possible implementation, the solid state drive life monitoring and early warning method based on performance analysis further includes:

[0041] The real-time fault event data is input into the life degradation warning model, and secondary fault identification is performed in the identification fault association identification map according to the matching fault events of the real-time fault event data, and multiple secondary fault events are output; multiple life degradation gradients corresponding to the multiple secondary fault events are obtained; the real-time life degradation gradient is fed back with the trigger probabilities corresponding to the multiple life degradation gradients as weights, and the optimized real-time life degradation gradient is output.

[0042] In a possible implementation, the solid state drive life monitoring and early warning method based on performance analysis further includes:

[0043] Acquire a set of associated fault events of the matching fault event; identify a set of transition probabilities in which the matching fault events are transferred to the set of associated fault events one by one; and output events in the set of transition probabilities that are greater than a preset transition probability as secondary fault events.

[0044] The solid-state hard disk life monitoring and early warning system based on performance analysis provided in the embodiment of the present invention can execute the solid-state hard disk life monitoring and early warning method based on performance analysis provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0045] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0046] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A solid state drive life monitoring and early warning system based on performance analysis, characterized in that: The system comprises: A historical fault event sample acquisition module, which is used to connect to the solid-state hard disk management system to obtain historical fault event samples, wherein each fault event sample records data before the fault occurs, data when the fault occurs, data when the fault recovers, and data after the fault recovers; A fault event sample output module, the fault event sample output module is used to cluster the historical fault event samples according to the fault type and output multiple types of fault event samples; A life reduction warning model output module, the life reduction warning model output module is used to train the life reduction warning model according to the multiple types of fault event samples, and output the trained life reduction warning model; A real-time life decline gradient output module, the real-time life decline gradient output module is used to monitor the real-time fault event data of the solid state drive in real time, input the real-time fault event data into the life decline warning model to perform life decline gradient identification, and output the real-time life decline gradient; A life warning signal generating module, the life warning signal generating module is used to generate a life warning signal when the real-time life decline gradient is greater than the matching decline warning gradient of the real-time fault event data; The life reduction warning model training is performed according to the multiple types of fault event samples, including: According to the multiple types of fault event samples, obtaining a stage feature vector and a life decline gradient label corresponding to each type of fault event sample; Acquire a life reduction warning model according to the stage feature vector and the life reduction gradient label; According to the stage feature vector and the life-decreasing gradient label, the sensitivity factor corresponding to each type of fault event sample is obtained; configuring multiple types of decline warning gradients according to the sensitive factors, and storing the multiple types of decline warning gradients in the life decline warning model; After outputting multiple types of fault event samples, including: Performing secondary fault event analysis on the multiple types of fault event samples to construct a fault association graph; Identify the secondary fault event triggering probability of the multiple types of fault event samples, perform probability identification on the fault association graph according to the secondary fault event triggering probability, and output an identified fault association identification graph; Connecting the identification fault association identification diagram with the life reduction warning model; After the identification fault association identification diagram is connected with the life reduction warning model, it includes: Inputting the real-time fault event data into the life reduction warning model, performing secondary fault identification in the identification fault association identification map according to matching class fault events of the real-time fault event data, and outputting a plurality of secondary fault events; Acquire a plurality of life reduction gradients corresponding to the plurality of secondary failure events; Feedback is performed on the real-time life-decreasing gradient using the trigger probabilities corresponding to the multiple life-decreasing gradients as weights, and an optimized real-time life-decreasing gradient is output; Performing secondary fault identification in the identification fault association identification map and outputting multiple secondary fault events include: Obtaining a set of associated fault events of the matching fault event; Identify the transition probability sets in which the matching fault events are transferred to the associated fault event sets in a one-to-one correspondence; Events in the transition probability set that are greater than a preset transition probability are output as secondary fault events.

2. The system according to claim 1, characterized in that The real-time fault event data is input into the life reduction warning model, and the system includes: In the life reduction warning model, the real-time fault event data is classified to obtain matching fault event samples; According to the matching fault event sample, obtaining a matching decline warning gradient; The life reduction warning model analyzes the real-time fault event data to obtain data recovery completeness, recovery time change, and read and write performance loss; A real-time lifespan reduction gradient is obtained according to the data recovery completeness, the recovery time change, and the read / write performance loss.

3. The system according to claim 1, characterized in that The life reduction warning model analyzes the real-time fault event data, and the system includes: The life reduction warning model analyzes the real-time fault event data to obtain a real-time stage feature vector; The stage feature vector includes the first data recovery completeness, the first recovery time change, and the first read / write performance loss between the data before the fault occurs and the data when the fault occurs, the second data recovery completeness, the second recovery time change, and the second read / write performance loss between the data when the fault occurs and the data when the fault is recovered, and the third data recovery completeness, the third recovery time change, and the third read / write performance loss between the data when the fault is recovered and the data after the fault is recovered; Outputting data recovery integrity according to the first data recovery integrity, the second data recovery integrity and the third data recovery integrity; Outputting a recovery time change according to the first recovery time change, the second recovery time change and the third recovery time change; Output a read / write performance loss according to the first read / write performance loss, the second read / write performance loss, and the third read / write performance loss.

4. A solid state drive life monitoring and early warning method based on performance analysis, characterized in that: The method is applied to the solid state drive life monitoring and early warning system based on performance analysis according to claim 1, and the method comprises: Connect to the solid-state drive management system to obtain historical fault event samples, where each fault event sample records data before the fault occurs, data when the fault occurs, data when the fault recovers, and data after the fault recovers; Clustering the historical fault event samples according to fault types, and outputting multiple types of fault event samples; Performing life reduction warning model training according to the multiple types of fault event samples, and outputting the trained life reduction warning model; real-time monitoring of real-time fault event data of the solid-state hard disk, inputting the real-time fault event data into the life reduction warning model to identify the life reduction gradient, and outputting the real-time life reduction gradient; When the real-time life decline gradient is greater than the matching decline warning gradient of the real-time fault event data, a life warning signal is generated.