A data center storage backup fault monitoring method, device, equipment and medium

Abstract

The application relates to the field of data processing, and in particular to a data center storage backup fault monitoring method and device, equipment and a medium. The method comprises the following steps: collecting device state data and current monitoring data of a storage backup system in real time, dynamically judging whether the system is in a normal state, distinguishing the fault type as a device fault or a data fault if an abnormality is detected, automatically capturing and recording a fault frame packet for the data fault, determining a repair method corresponding to the fault frame packet, executing a repair operation based on the repair method, and verifying the repair result. If the repair is successful, a standardized report containing the fault time, type, repair method and result is generated and pushed to an operation and maintenance terminal, closed-loop management and traceability of fault processing are realized, and through automatic monitoring, intelligent diagnosis and standardized repair, the fault response efficiency of the data center storage backup system and the reliability of the backup data are significantly improved.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a method, apparatus, device and medium for monitoring data center storage backup faults. Background Technology

[0002] With the acceleration of digital transformation, data centers have become the central hub for storing and processing core business data for enterprises. To ensure data security and business continuity, storage backup systems need to regularly perform full / incremental backups of massive amounts of data and support rapid recovery (e.g., RTO < 4 hours, RPO < 1 hour). Backup scenarios cover structured data (e.g., databases), unstructured data (e.g., video, logs), and cross-cloud / on-premises hybrid architectures, and must meet compliance requirements (e.g., long-term data retention in the financial and healthcare industries). Against this backdrop, the stability and fault handling capabilities of storage backup systems have become critical challenges.

[0003] Currently, fault monitoring of data center storage backup systems mainly relies on manual inspections. Specifically, maintenance personnel need to manually check device status indicator lights, analyze backup logs, and investigate network transmission records, using experience to determine the type of fault and take corrective measures. However, this manually-led monitoring method depends heavily on human experience, resulting in a high rate of false fault diagnosis and consequently, low reliability of backup data. Summary of the Invention

[0004] To improve the reliability of backup data, this application provides a data center storage backup fault monitoring method, apparatus, device, and medium.

[0005] Firstly, this application provides a data center storage backup fault monitoring method, which adopts the following technical solution:

[0006] A data center storage backup fault monitoring method, comprising:

[0007] Collect device status data and obtain current monitoring data. The device status data is the status data of the device corresponding to the data center storage backup system.

[0008] Based on the device status data and the current monitoring data, determine whether the data center storage backup system is in a normal state;

[0009] If the data center storage backup system is not in normal condition, the fault type of the data center storage backup system is determined, and the fault type is either equipment failure or data failure.

[0010] When the fault type is a data fault, record the fault frame message corresponding to the current monitored data and determine the repair method corresponding to the fault frame message;

[0011] The faulty frame message is repaired based on the repair method, and it is determined whether the faulty frame message is successfully repaired.

[0012] If the repair is successful, a repair report corresponding to the current monitoring data will be generated and sent to the corresponding operation and maintenance terminal.

[0013] By adopting the above technical solution, the system collects real-time device status data and current monitoring data of the storage backup system, dynamically determines whether the system is in a normal state, and distinguishes between device failure and data failure if an anomaly is detected. For data failures, the system automatically captures and records fault frame messages, determines the corresponding repair method, executes repair operations based on the repair method, and verifies the repair results. If the repair is successful, a standardized report containing the fault time, type, repair method, and result is generated and pushed to the operation and maintenance terminal, realizing closed-loop management and traceability of fault handling. At the same time, through automated monitoring, intelligent diagnosis, and standardized repair, the system significantly improves the fault response efficiency and backup data reliability of the data center storage backup system.

[0014] In one possible implementation, the device status data includes device temperature, device rotation speed, and device vibration signals, and the current monitoring data includes the backup software's running logs and network communication data.

[0015] The step of determining whether the data center storage backup system is in a normal state based on the device status data and the current monitoring data includes:

[0016] The vibration signal of the device is processed by a CNN sub-network to obtain the spatial feature vector of the device, and the temperature and rotation speed of the device are processed by an LSTM sub-network to obtain the temporal feature vector.

[0017] The spatial feature vector and the temporal feature vector are input into the encoder, and the intermodal correlation weights are calculated through a multi-head attention mechanism to generate a fused feature vector.

[0018] The fused feature vector is input into a fully connected classification network, and the output device health score is obtained.

[0019] Based on the device health score, it is determined whether the data center storage backup system is in normal condition;

[0020] If it is determined that the data center storage backup system is not in a normal state, then the data center storage backup system is determined to be not in a normal state.

[0021] If it is determined that the data center storage backup system is in a normal state, then based on the current monitoring data, it is determined whether the data center storage backup system is in a normal state.

[0022] In one possible implementation, determining whether the data center storage backup system is in a normal state based on the current monitoring data includes:

[0023] Acquire historical fault data, which includes historical fault frame messages and historical network communication data corresponding to each historical fault frame;

[0024] The historical fault data is analyzed in the time domain and frequency domain to obtain the first time domain relationship between historical fault frame messages and abnormal fault forms, the second time domain relationship between historical fault frames and abnormal network communication data, and the frequency domain feature vector matrix corresponding to the historical fault data.

[0025] Based on the first time-domain relationship, the second time-domain relationship, and the frequency-domain feature vector matrix, the abnormal message format and abnormal network communication data format corresponding to the historical fault frame are determined.

[0026] The current monitoring data is compared with the abnormal message format and the abnormal network communication data format respectively to determine whether the current monitoring data has an abnormal message format and / or an abnormal network communication data format.

[0027] When the current monitoring data contains abnormal message formats or / and abnormal network communication data formats, it is determined that the data center storage backup system is not in a normal state.

[0028] In one possible implementation, the historical fault data is subjected to time-domain and frequency-domain analysis to obtain a first time-domain relationship between historical fault frame messages and abnormal fault forms, a second time-domain relationship between historical fault frames and abnormal network communication data, and a frequency-domain feature vector matrix corresponding to the historical fault data, including:

[0029] The historical fault data is segmented into sliding windows, and the abnormal fault type corresponding to each historical fault message in each historical window is parsed.

[0030] Based on the abnormal fault type corresponding to each historical fault message in each historical window, the first temporal relationship between historical fault messages and abnormal fault types is determined through correlation analysis.

[0031] Identify the abnormal network communication data within each historical window, calculate the cross-correlation of the historical window sequence, identify the temporal coupling relationship between historical fault frames and abnormal network communication data, and form a second temporal relationship;

[0032] Frequency domain analysis is performed on the abnormal fault types and abnormal network communication data in the historical fault data to obtain the first frequency domain feature vector corresponding to the abnormal fault type and the second frequency domain feature vector corresponding to the abnormal network communication data.

[0033] The first frequency domain feature vector and the second frequency domain feature vector are integrated to obtain the frequency domain feature vector matrix corresponding to the historical fault data.

[0034] In one possible implementation, the historical fault data further includes historical fault frame data, and determining the repair method corresponding to the fault frame message includes:

[0035] The fault frame message is parsed to extract key feature vectors, and the key feature vectors are compared with the historical feature vectors corresponding to the historical fault frame data to obtain a comprehensive similarity. The historical fault frame data includes historical feature vectors and historical repair methods.

[0036] When there is a comprehensive similarity greater than the first similarity threshold, the comprehensive similarity greater than the first similarity threshold is determined as the first target similarity, and the historical repair method corresponding to the first target similarity is obtained and determined as the repair method corresponding to the fault frame message;

[0037] When there is no comprehensive similarity greater than the first similarity threshold, and there are multiple comprehensive similarities greater than the second similarity threshold, the comprehensive similarity greater than the second similarity threshold is determined as the second target similarity. Based on the random forest algorithm, the historical repair methods corresponding to each of the multiple second target similarities are combined and applied to obtain the optimal strategy combination. The optimal strategy combination is determined as the repair method corresponding to the fault frame message, wherein the second similarity threshold is less than the first similarity threshold.

[0038] When the overall similarity is not greater than the second similarity threshold, a fault propagation graph model is constructed based on each historical repair method. The key feature vector is input into the fault propagation graph model, and the expected success rate of each repair method output by the fault propagation graph model is obtained. The repair method with the highest expected success rate is selected as the repair method corresponding to the fault frame message.

[0039] In one possible implementation, repairing the faulty frame message based on the repair method includes:

[0040] Determine whether the repair method meets the automatic repair conditions;

[0041] If the repair method meets the automatic repair conditions, a repair instruction is generated based on the repair method, and the repair instruction is sent to the repair terminal corresponding to the fault frame message;

[0042] If the repair method does not meet the automatic repair conditions, a basic repair method is obtained, a basic repair instruction corresponding to the basic repair method is generated, and the basic repair instruction is sent to the repair terminal corresponding to the fault frame message.

[0043] In one possible implementation, when the repair method meets the automatic repair conditions, determining whether the fault frame message has been successfully repaired includes:

[0044] Automatically acquire repair completion signals, and after acquiring the repair completion signal, obtain repair monitoring data and repair fault frame messages;

[0045] The repair monitoring data is compared with the abnormal message format and the abnormal network communication data format to determine whether the repair monitoring data has an abnormal message format or / and an abnormal network communication data format.

[0046] If it is determined that the repair monitoring data does not contain abnormal message formats or / and abnormal network communication data formats, the repair fault frame message is parsed to obtain the repair key feature vector, and based on the historical feature vector, it is determined whether the fault frame message has been successfully repaired.

[0047] Secondly, this application provides a data center storage backup fault monitoring device, which adopts the following technical solution:

[0048] A data center storage backup fault monitoring device, comprising:

[0049] The acquisition module is used to collect device status data and obtain current monitoring data. The device status data is the status data of the device corresponding to the data center storage backup system.

[0050] The status determination module is used to determine whether the data center storage backup system is in a normal state based on the device status data and the current monitoring data.

[0051] The type determination module is used to determine the fault type of the data center storage backup system if the data center storage backup system is not in a normal state, wherein the fault type is equipment failure or data failure.

[0052] The recording module is used to record the fault frame message corresponding to the current monitoring data when the fault type is a data fault, and to determine the repair method corresponding to the fault frame message;

[0053] The repair module is used to repair the fault frame message based on the repair method and determine whether the fault frame message has been successfully repaired.

[0054] The sending module is used to generate a repair report corresponding to the current monitoring data if the repair is successful, and send the repair report to the corresponding operation and maintenance terminal.

[0055] Thirdly, this application provides an electronic device that adopts the following technical solution:

[0056] An electronic device comprising:

[0057] At least one processor;

[0058] Memory;

[0059] At least one application, wherein the at least one application is stored in memory and configured to be executed by at least one processor, the at least one application being configured to: perform the method described in any of the first aspects above.

[0060] Fourthly, this application provides a computer-readable storage medium, which adopts the following technical solution:

[0061] A computer-readable storage medium includes: a computer program stored thereon that can be loaded by a processor and execute the method described in any of the first aspects above.

[0062] In summary, this application includes the following beneficial technical effects:

[0063] By collecting real-time device status data and current monitoring data from the storage backup system, and dynamically determining whether the system is in a normal state, if an anomaly is detected, the system distinguishes between device failure and data failure. For data failures, the system automatically captures and records fault frame messages, determines the corresponding repair method, executes repair operations based on that method, and verifies the repair results. If the repair is successful, a standardized report containing the fault time, type, repair method, and result is generated and pushed to the operations and maintenance terminal, achieving closed-loop management and traceability of fault handling. Simultaneously, through automated monitoring, intelligent diagnosis, and standardized repair, the system significantly improves the fault response efficiency and backup data reliability of the data center storage backup system. Attached Figure Description

[0064] Figure 1 This is a flowchart illustrating a data center storage backup fault monitoring method provided in an embodiment of this application;

[0065] Figure 2 This is a block diagram of a data center storage backup fault monitoring device provided in an embodiment of this application;

[0066] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0067] The following is in conjunction with the appendix Figure 1 To be continued Figure 3 This application will be described in further detail.

[0068] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0069] This application provides a data center storage backup fault monitoring method, such as... Figure 1 As shown, the method provided in this application embodiment is executed by an electronic device, which can be a server or a terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smartphone, tablet, laptop, desktop computer, etc., but is not limited to these. The terminal device and the server can be directly or indirectly connected via wired or wireless communication. This application embodiment does not impose any limitations on this connection. The method includes steps S101-S106, wherein:

[0070] Step S101: Collect device status data and obtain current monitoring data.

[0071] The device status data refers to the status data of the devices (such as hard drives, tape drives, switches, etc.) corresponding to the data center storage backup system. Specifically, the device status data can include physical indicators such as device temperature, rotation speed, vibration signal, voltage and current, reflecting the health of the device hardware.

[0072] The current monitoring data is related to the operation of the storage backup system, including the backup software's operation logs (recording backup task execution status, error information, etc.), network communication data (such as transmission bandwidth, latency, packet loss rate), and data verification results, which are used to monitor the status of system software and data transmission.

[0073] Specifically, electronic devices collect device status data in real time through sensors (such as temperature sensors and vibration sensors) deployed on storage backup devices; at the same time, they interact with backup software and network devices using data acquisition programs to obtain operation logs and network communication data in real time in order to obtain current monitoring data.

[0074] Step S102: Based on the device status data and current monitoring data, determine whether the data center storage backup system is in normal condition.

[0075] In this context, "normal state" means that the equipment, software, and data transmission of the data center storage backup system are all operating normally as expected, without any hardware failures, software errors, or data loss.

[0076] After obtaining equipment status data and current monitoring data, pre-trained machine learning models can be used to analyze these data. Specifically, convolutional neural networks (CNNs) can be used to process equipment vibration signals and extract spatial features; long short-term memory networks (LSTMs) can be used to analyze the temporal changes in equipment temperature and rotational speed. The extracted features are then fused and input into a classification model to calculate the system's health score, which is compared with a preset normal threshold to determine whether the system is in a normal state.

[0077] Step S103: If the data center storage backup system is not in normal condition, determine the fault type of the data center storage backup system.

[0078] The fault types are equipment faults or data faults. Equipment faults refer to abnormalities in the hardware devices of the storage backup system (such as hard drive failure, fan failure, or controller failure), causing the system to malfunction. Data faults refer to problems that occur during the storage, transmission, or processing of backup data, such as data loss, corruption, verification errors, or incomplete backup task execution resulting in incomplete data.

[0079] When the electronic device determines that the system is in an abnormal state, the characteristics of the abnormal data are further analyzed. If hardware-related indicators (such as excessively high temperature or abnormal rotation speed) in the device status data exceed the normal range, it is determined to be a device malfunction; if the current monitoring data contains backup data verification errors or backup task failure records, it is determined to be a data malfunction.

[0080] Furthermore, if it is determined that the data center storage backup system is in normal condition, no further operations will be performed.

[0081] Step S104: When the fault type is a data fault, record the fault frame message corresponding to the current monitored data and determine the repair method corresponding to the fault frame message.

[0082] A fault frame message is a data packet or data fragment containing error information generated during data transmission or storage due to data failure. It records the data status and related parameters at the time of the failure.

[0083] Once the fault type is determined to be a data fault, the complete data frame message related to the fault can be extracted from the current monitoring data, including error code, data content, transmission time, and other information, and stored in the fault log. Based on the characteristics of the fault frame message (such as error type and the range of data involved), the corresponding repair method is matched from the predefined repair strategy library. For example, if it is a verification error, the data can be retrieved again from the most recent complete backup copy.

[0084] Furthermore, when the fault type is equipment fault, an alarm message containing equipment fault information is generated and sent to the terminal device of the operation and maintenance personnel corresponding to the data center storage backup system to remind the operation and maintenance personnel to repair the equipment.

[0085] Step S105: Repair the faulty frame message based on the repair method, and determine whether the faulty frame message has been successfully repaired.

[0086] After obtaining the repair method corresponding to the faulty frame message, the repair operation can be performed according to the determined repair method. Specifically, a data retransmission request is sent to the data source, or the correct data is read from the backup storage to overwrite the faulty data. After the repair is completed, the repaired data is verified again (e.g., the hash value is calculated and compared with the hash value of the original correct data). If the verification passes, the repair is considered successful; if the verification fails, the repair is considered to have failed.

[0087] Step S106: If the repair is successful, generate a repair report corresponding to the current monitoring data and send the repair report to the corresponding operation and maintenance terminal.

[0088] The repair report is a document that records the data fault repair process and results. It includes information such as the fault occurrence time, fault type, repair method, repair time, and data status after repair, which is used by operation and maintenance personnel to understand the fault handling situation.

[0089] Operation and maintenance terminals are devices used by operation and maintenance personnel to receive fault information and manage the system, such as operation and maintenance workstations and mobile terminals, so that operation and maintenance personnel can keep abreast of the system status and take appropriate measures.

[0090] After confirming successful repair, a structured repair report is automatically generated, summarizing key information about the fault handling. Then, the repair report is sent to a pre-defined maintenance terminal via network communication protocols (such as HTTP or SMTP), such as sending it to the maintenance personnel's email address or pushing a message on the maintenance management platform to remind the maintenance personnel that the fault has been successfully repaired.

[0091] Furthermore, if the repair fails, an alarm signal for the current monitoring data is generated and sent to the corresponding operation and maintenance terminal.

[0092] This application provides a data center storage backup fault monitoring method. It collects real-time device status data and current monitoring data from the storage backup system, dynamically determining whether the system is in a normal state. If an anomaly is detected, the method distinguishes between device faults and data faults. For data faults, it automatically captures and records fault frame messages, determines the corresponding repair method, performs repair operations based on the method, and verifies the repair results. If the repair is successful, a standardized report containing the fault time, type, repair method, and result is generated and pushed to the operation and maintenance terminal, achieving closed-loop management and traceability of fault handling. Simultaneously, through automated monitoring, intelligent diagnosis, and standardized repair, it significantly improves the fault response efficiency and backup data reliability of the data center storage backup system.

[0093] In one possible implementation of this application embodiment, in step S102 above, the device status data includes device temperature, device rotation speed, and device vibration signal, and the current monitoring data includes the backup software's running log and network communication data.

[0094] This includes determining whether the data center storage backup system is in a normal state based on device status data and current monitoring data, including:

[0095] The vibration signal of the equipment is processed by a CNN sub-network to obtain the spatial feature vector of the equipment, and the temperature and rotation speed of the equipment are processed by an LSTM sub-network to obtain the temporal feature vector.

[0096] Spatial and temporal feature vectors are input into the encoder, and intermodal correlation weights are calculated through a multi-head attention mechanism to generate a fused feature vector.

[0097] The fused feature vectors are input into a fully connected classification network, and the output device health score is obtained.

[0098] Based on the equipment health score, determine whether the data center storage backup system is in normal condition;

[0099] If it is determined that the data center storage backup system is not in a normal state, then it is determined that the data center storage backup system is not in a normal state.

[0100] If it is determined that the data center storage backup system is in normal condition, then based on the current monitoring data, determine whether the data center storage backup system is in normal condition.

[0101] Among them, equipment vibration signal is the signal formed by the mechanical vibration generated by the equipment during operation. Different equipment faults (such as bearing wear and poor gear meshing) will produce vibration signals of specific patterns, which are important evidence reflecting the operating status of the equipment.

[0102] After obtaining the equipment vibration signal collected from the sensor, the vibration signal can be preprocessed, such as by filtering and normalization, to remove noise interference and unify the data format. Then, the processed vibration signal is input into a pre-trained CNN sub-network model. The CNN sub-network model automatically extracts local features from the vibration signal through convolutional layers, and then reduces the data dimensionality through pooling layers, finally outputting a spatial feature vector that reflects the spatial structure of the vibration signal. Simultaneously, the equipment temperature sequence and equipment rotation speed sequence corresponding to the current period are obtained, and the time-series data of equipment temperature and rotation speed are organized and input into the LSTM sub-network in chronological order. The LSTM sub-network, with its unique memory units, learns and extracts the characteristics of temperature and rotation speed data in the time dimension, such as trends and periodicity, and outputs a vector containing temporal features.

[0103] Furthermore, the obtained spatial and temporal feature vectors are input into the encoder module. The encoder utilizes a multi-head attention mechanism to analyze these two feature vectors from multiple perspectives. During the calculation process, matrix operations and similarity metrics are used to calculate the correlation between the equipment vibration signal features and the equipment temperature and rotational speed features, resulting in multiple attention weights. These weights reflect the relative importance of different modal data in describing the equipment state. Finally, the data are merged and weighted to generate a fused feature vector that integrates spatial and temporal information, enabling this vector to more comprehensively reflect the equipment's operating state.

[0104] Furthermore, the generated fused feature vector is directly fed into a fully connected classification network. Each layer of neurons in the fully connected classification network sequentially calculates and transforms the input fused feature vector, mapping the high-dimensional feature vector to a specific-dimensional output space through multiple layers of linear and nonlinear operations. Finally, the network outputs a numerical value, namely the device health score. This score is derived from the multi-faceted state information of the device contained in the fused feature vector, through the mapping relationship learned by the network, and can intuitively reflect the current health level of the device. A pre-set normal threshold range for the device health score is obtained, for example, a score of 80-100 is considered normal. After obtaining the device health score, it is compared with the set threshold. If the score is within the normal threshold range, the data center storage backup system is determined to be in a normal state; if the score is below the lower limit of the normal threshold, the system is determined to be in an abnormal state.

[0105] When it is determined that the data center storage backup system is not in a normal state, the system is directly identified as abnormal. If the system is determined to be in a normal state, further analysis is performed on the backup software's operation logs and current monitoring data such as network communication data to determine whether the data center storage backup system is truly operating normally.

[0106] One possible implementation of this application embodiment involves determining whether a data center storage backup system is in a normal state based on current monitoring data, including:

[0107] Acquire historical fault data, which includes historical fault frame messages and historical network communication data corresponding to each historical fault frame;

[0108] Time-domain and frequency-domain analyses are performed on historical fault data to obtain the first time-domain relationship between historical fault frame messages and abnormal fault forms, the second time-domain relationship between historical fault frames and abnormal network communication data, and the frequency-domain feature vector matrix corresponding to historical fault data.

[0109] Based on the first time-domain relationship, the second time-domain relationship, and the frequency domain feature vector matrix, the abnormal message format and abnormal network communication data format corresponding to the historical fault frames are determined.

[0110] The current monitoring data is compared with the abnormal message format and the abnormal network communication data format to determine whether the current monitoring data contains abnormal message format and / or abnormal network communication data format.

[0111] When the current monitoring data contains abnormal message formats or / and abnormal network communication data formats, it is determined that the data center storage backup system is not in a normal state.

[0112] Historical fault records corresponding to the current data center storage backup system are retrieved from the data center storage backup database. Historical fault frame messages and historical network communication data corresponding to each historical fault event are extracted and categorized by timestamp and fault type to construct a historical fault dataset. Further, the historical fault frame messages are segmented using a sliding window method, and statistical characteristics (such as error code frequency and peak intensity) within each window are calculated and analyzed for their temporal correlation with the fault type. For example, if a certain type of error code occurs frequently within the 5 minutes prior to the fault, a corresponding relationship is established. A Fast Fourier Transform (FFT) is performed on the historical network communication data (such as delay curves) to extract the main frequency components and energy distribution, constructing a frequency domain feature vector matrix. For example, this identifies the characteristic frequencies of periodic delay fluctuations.

[0113] Based on the first and second time-domain relationships, cluster analysis is performed on historical fault frame messages to identify common error code combinations, data corruption patterns, etc., forming abnormal message templates. Simultaneously, the frequency characteristic patterns of abnormal network communication data (such as high-frequency jitter and low-frequency oscillation) are determined using a frequency domain feature vector matrix, and these patterns are associated with specific fault types.

[0114] The system performs pattern matching between fault frame messages in the current monitoring data and abnormal message templates, for example, checking for specific error code sequences or data segment corruption characteristics. Simultaneously, it performs time-domain and frequency-domain analysis on the current network communication data, extracting features and comparing them with the frequency feature patterns of abnormal network communication data, calculating similarity based on methods such as cosine similarity or Euclidean distance. If a fault frame matching the abnormal message format exists in the current monitoring data, or if the frequency domain features of the network communication data are highly similar to the abnormal network communication data (similarity exceeding a preset threshold, such as 80%), the data center storage backup system is determined to be in an abnormal state.

[0115] One possible implementation of this application involves performing time-domain and frequency-domain analysis on historical fault data to obtain a first time-domain relationship between historical fault frame messages and abnormal fault forms, a second time-domain relationship between historical fault frames and abnormal network communication data, and a frequency-domain feature vector matrix corresponding to the historical fault data, including:

[0116] The historical fault data is divided into sliding windows, and the abnormal fault type corresponding to each historical fault message in each historical window is parsed.

[0117] Based on the abnormal fault type corresponding to each historical fault message in each historical window, the first temporal relationship between historical fault messages and abnormal fault types is determined through correlation analysis.

[0118] Identify the abnormal network communication data within each historical window, calculate the cross-correlation of the historical window sequence, identify the temporal coupling relationship between historical fault frames and abnormal network communication data, and form a second temporal relationship;

[0119] Frequency domain analysis was performed on the abnormal fault forms and abnormal network communication data in the historical fault data to obtain the first frequency domain feature vector corresponding to the abnormal fault forms and the second frequency domain feature vector corresponding to the abnormal network communication data.

[0120] The first frequency domain feature vector and the second frequency domain feature vector are integrated to obtain the frequency domain feature vector matrix corresponding to the historical fault data.

[0121] Specifically, the size (e.g., 30 seconds) and step size (e.g., 10 seconds) of the set sliding window can be obtained to perform time-series segmentation of historical fault data. Within each window, key fields of the fault message (e.g., error code, timestamp, device ID) are parsed and matched with a preset fault classification dictionary to automatically label the abnormal fault type corresponding to each message (e.g., error code "ERR-003" corresponds to "disk read / write error"). Furthermore, the fault messages within each historical window are aggregated and statistically analyzed to calculate the frequency and time distribution characteristics of different abnormal fault types (e.g., first occurrence time, peak time), and these statistical patterns are organized into a rule base to form the first time-domain relationship.

[0122] Network communication data (such as latency per second) is extracted from historical windows and compared with preset thresholds to mark anomalies. Then, for each fault frame, the time difference between it and network anomalies within the preceding and following 5 minutes is calculated. Patterns are identified through cross-correlation analysis to form a temporal coupling relationship of "network jitter before the fault," thus obtaining a second temporal relationship. Further, the anomalous fault forms in the historical fault data are reconstructed in the temporal domain (e.g., converting error code sequences into numerical sequences), and converted to the frequency domain using Fast Fourier Transform (FFT). The main frequency components and their energies are extracted to form a first frequency domain feature vector. Simultaneously, anomalous network communication data (such as latency curves) undergoes the same processing to obtain a second frequency domain feature vector. The first and second frequency domain feature vectors are aligned according to fault type and timestamp and merged into a unified feature matrix. For example, for each historical fault sample, the first half of the matrix stores the frequency domain features of the fault form, and the second half stores the frequency domain features of the network anomaly. Each row of the frequency domain feature vector matrix corresponds to a fault sample, and each column corresponds to a frequency domain feature dimension.

[0123] In one possible implementation of this application embodiment, the historical fault data further includes historical fault frame data, and determining the repair method corresponding to the fault frame message includes:

[0124] The fault frame message is parsed to extract key feature vectors, and the key feature vectors are compared with the historical feature vectors corresponding to the historical fault frame data to obtain the comprehensive similarity. The historical fault frame data includes historical feature vectors and historical repair methods.

[0125] When there is a comprehensive similarity greater than the first similarity threshold, the comprehensive similarity greater than the first similarity threshold is determined as the first target similarity, and the historical repair method corresponding to the first target similarity is obtained and determined as the repair method corresponding to the fault frame message;

[0126] When there is no comprehensive similarity greater than the first similarity threshold, but there are multiple comprehensive similarities greater than the second similarity threshold, the comprehensive similarity greater than the second similarity threshold is determined as the second target similarity. Based on the random forest algorithm, the historical repair methods corresponding to the multiple second target similarities are combined to obtain the optimal strategy combination. The optimal strategy combination is determined as the repair method corresponding to the fault frame message. The second similarity threshold is less than the first similarity threshold.

[0127] When the overall similarity is not greater than the second similarity threshold, a fault propagation graph model is constructed based on each historical repair method. The key feature vector is input into the fault propagation graph model, and the expected success rate of each repair method output by the fault propagation graph model is obtained. The repair method with the highest expected success rate is selected as the repair method corresponding to the fault frame message.

[0128] Among them, the key feature vector is a set of features extracted from the current fault frame message that can characterize the fault type, such as error code, data segment corruption location, timestamp, etc.

[0129] Specifically, the fault frame message is parsed, and key features (such as error code sequences, data checksums, and transmission protocol versions) are extracted and converted into standardized vectors. Further, the historical fault frame database is traversed, and the similarity between the feature vector of each historical record and the current vector is calculated. For example, by calculating the cosine of the angle between the vectors, a similarity score between 0 and 1 is obtained, ultimately generating a similarity list between the current fault and all historical faults. It is worth noting that the method of extracting the parsed message features is common knowledge and is not limited in this embodiment.

[0130] Furthermore, after obtaining the overall similarity score, records with similarity scores greater than the first threshold are filtered out. If such records exist, the corresponding fault repair method is directly extracted from the historical database (for example, when the overall similarity score is 0.95, the historical repair method is "resynchronize backup task"), and it is determined as the repair solution for the current fault.

[0131] When no high-similarity matches exist, historical fault records with a comprehensive similarity between the first and second similarity thresholds are selected. For each record, its repair method (such as "clearing the cache" or "adjusting network parameters") is extracted, and these methods are used as candidate strategies. Furthermore, a random forest algorithm can be used to train a model based on historical data, learn the effectiveness weights of different repair methods, and generate a combined strategy (such as clearing the cache first, and adjusting parameters if ineffective) as the repair solution for the current fault. Further, when there is only one historical fault record with a comprehensive similarity between the first and second similarity thresholds, the corresponding fault repair method is directly extracted from the historical database and determined as the repair solution for the current fault.

[0132] When all historical similarities fall below the second similarity threshold, the fault propagation graph model can be activated. Built upon historical data, the model records the transition relationships between different fault states and the effectiveness of different repair methods. Specifically, the key feature vector of the current fault is input into the fault propagation graph model. By simulating the fault evolution path, the model predicts the success probability of each possible repair method and selects the repair method with the highest expected success rate (such as "restore from backup") as the solution for the current fault. For example, for a "data checksum error" fault, the model might predict an 85% success rate for "restore from backup" and a 30% success rate for "recalculate checksum".

[0133] One possible implementation of this application embodiment repairs faulty frame messages based on a repair method, including:

[0134] Determine whether the repair method meets the conditions for automatic repair;

[0135] If the repair method meets the automatic repair conditions, a repair instruction is generated based on the repair method and sent to the repair terminal corresponding to the fault frame message.

[0136] If the repair method does not meet the automatic repair conditions, the basic repair method is obtained, the basic repair instruction corresponding to the basic repair method is generated, and the basic repair instruction is sent to the repair end corresponding to the fault frame message.

[0137] Among them, automatic repair conditions refer to a set of preset rules and standards used by electronic devices in the data center storage backup system to assess whether automatic repair can be performed directly without human intervention when determining the fault repair method. These include operational risk assessment (such as success rate > 90%), time window limits (such as non-peak business periods), and system resource status (such as CPU utilization < 80%).

[0138] The basic repair methods are predefined, general operations applicable to various fault scenarios, such as log collection, service status checks, and temporary data isolation.

[0139] Specifically, the system retrieves the attributes of the current repair method (such as operation type, scope of impact, and historical success rate) from the data center storage backup system knowledge base and compares them with the automatic repair condition database. For example, if the repair method is "restart backup service," it checks whether the current service has a historical restart success rate > 95%; if the repair method is "data retransmission," it checks whether the network bandwidth is sufficient (current utilization < 70%) and whether it is outside of peak business hours (such as 2 AM). If all conditions are met, it is determined that "automatic repair conditions are met."

[0140] The repair method is converted into an executable instruction. After the instruction is generated, the electronic device sends it to the corresponding repair terminal through a secure channel (such as a TLS-encrypted REST API) and initiates timeout monitoring (e.g., triggering an alarm if there is no response after 120 seconds). The repair terminal is the hardware device or software component (such as a backup server or storage controller) responsible for performing the actual repair operation.

[0141] Furthermore, when the repair method does not meet the conditions for automatic repair, the operation that best matches the current fault type (such as "collect system logs") is selected from the basic repair library to obtain the basic modification method, and the basic repair instruction corresponding to the basic repair method is generated. Even further, after the instruction is sent, a manual intervention process is triggered simultaneously: an alarm message is sent to the operations and maintenance team, and a fault work order is generated in the management interface, attaching the basic repair results.

[0142] One possible implementation of this application embodiment, when the repair method meets the automatic repair conditions, determines whether the fault frame message has been successfully repaired, including:

[0143] Automatically acquire repair completion signals, and after acquiring the repair completion signal, obtain repair monitoring data and repair fault frame messages;

[0144] The repair monitoring data is compared with the abnormal message format and the abnormal network communication data format respectively to determine whether the repair monitoring data has an abnormal message format or / and an abnormal network communication data format.

[0145] If it is determined that there are no abnormal message formats or / and abnormal network communication data formats in the repair monitoring data, the fault frame message is parsed to obtain the key feature vector for repair, and based on the historical feature vector, it is determined whether the fault frame message has been successfully repaired.

[0146] Among them, the fault repair frame message is the repaired data frame, which contains the correction content of the original fault location.

[0147] Specifically, after sending the repair command, a timeout timer (e.g., 120 seconds) is started to listen for the HTTP 200 response or MQTT message from the repair end. Upon receiving a success signal, metrics such as network bandwidth, latency, and packet loss rate after repair are obtained from the monitoring system (e.g., Prometheus); the latest operation logs of the backup software are extracted from the log system (e.g., ELKStack); and the repaired fault frame messages (e.g., retransmitted data blocks) are obtained from the storage system.

[0148] Furthermore, the repair monitoring data is compared with a pre-defined anomaly pattern database. Specifically, the repaired packet headers are parsed to check for error code sequences from historical faults (e.g., through regular expression matching), and current network traffic is analyzed in the frequency domain (e.g., using FFT transform) to check for frequency characteristics matching historical anomalies (e.g., 50Hz jitter). If no match is found, the process proceeds to the next step of feature verification. If a match is found, an alarm is sent to the operations and maintenance team.

[0149] Furthermore, the repaired faulty frame packets undergo in-depth analysis to determine whether the repair was successful. Specifically, key fields such as the version number and sequence number are extracted from the protocol header. The extracted feature vectors are then compared with normal templates in the historical feature library using cosine similarity calculation (with a threshold set to 0.95), and multi-dimensional verification is performed. More specifically, multi-dimensional verification includes: checking whether the timestamp of the repaired packet conforms to the normal time sequence; and verifying whether the length and number of segments of the data frame are consistent with the standard template. If the similarity exceeds the threshold, the repair is considered successful; otherwise, an alarm message is sent to the operations and maintenance team.

[0150] Furthermore, when the repair method does not meet the conditions for automatic repair, it is determined that the fault frame message repair has failed, and an alarm message is sent to the operation and maintenance team.

[0151] The above embodiments describe a data center storage backup fault monitoring method from the perspective of process flow. The following embodiments describe a data center storage backup fault monitoring device from the perspective of virtual module or virtual unit. For details, please refer to the following embodiments.

[0152] See Figure 2 The data center storage backup fault monitoring device 20 may specifically include: a data acquisition module 201, a status determination module 202, a type determination module 203, a recording module 204, a repair module 205, and a sending module 206, wherein:

[0153] A data center storage backup fault monitoring device 20 includes:

[0154] The acquisition module 201 is used to acquire device status data and obtain current monitoring data. The device status data is the status data of the device corresponding to the data center storage backup system.

[0155] The status determination module 202 is used to determine whether the data center storage backup system is in a normal state based on the device status data and the current monitoring data.

[0156] The type determination module 203 is used to determine the fault type of the data center storage backup system if the data center storage backup system is not in a normal state. The fault type is either equipment failure or data failure.

[0157] The recording module 204 is used to record the fault frame message corresponding to the current monitored data and determine the repair method corresponding to the fault frame message when the fault type is a data fault.

[0158] Repair module 205 is used to repair fault frame messages based on repair methods and determine whether the fault frame messages have been successfully repaired.

[0159] The sending module 206 is used to generate a repair report corresponding to the current monitoring data if the repair is successful, and send the repair report to the corresponding operation and maintenance terminal.

[0160] In one possible implementation of this application, the device status data includes device temperature, device rotation speed, and device vibration signal; the current monitoring data includes the backup software's running logs and network communication data.

[0161] Specifically, when determining whether the data center storage backup system is in a normal state based on device status data and current monitoring data, the status determination module 202 is used for:

[0162] The vibration signal of the equipment is processed by a CNN sub-network to obtain the spatial feature vector of the equipment, and the temperature and rotation speed of the equipment are processed by an LSTM sub-network to obtain the temporal feature vector.

[0163] Spatial and temporal feature vectors are input into the encoder, and intermodal correlation weights are calculated through a multi-head attention mechanism to generate a fused feature vector.

[0164] The fused feature vectors are input into a fully connected classification network, and the output device health score is obtained.

[0165] Based on the equipment health score, determine whether the data center storage backup system is in normal condition;

[0166] If it is determined that the data center storage backup system is not in a normal state, then it is determined that the data center storage backup system is not in a normal state.

[0167] If it is determined that the data center storage backup system is in normal condition, then based on the current monitoring data, determine whether the data center storage backup system is in normal condition.

[0168] In one possible implementation of this application embodiment, when the status determination module 202 determines whether the data center storage backup system is in a normal state based on current monitoring data, it is specifically used for:

[0169] Acquire historical fault data, which includes historical fault frame messages and historical network communication data corresponding to each historical fault frame;

[0170] Time-domain and frequency-domain analyses are performed on historical fault data to obtain the first time-domain relationship between historical fault frame messages and abnormal fault forms, the second time-domain relationship between historical fault frames and abnormal network communication data, and the frequency-domain feature vector matrix corresponding to historical fault data.

[0171] Based on the first time-domain relationship, the second time-domain relationship, and the frequency domain feature vector matrix, the abnormal message format and abnormal network communication data format corresponding to the historical fault frames are determined.

[0172] The current monitoring data is compared with the abnormal message format and the abnormal network communication data format to determine whether the current monitoring data contains abnormal message format and / or abnormal network communication data format.

[0173] When the current monitoring data contains abnormal message formats or / and abnormal network communication data formats, it is determined that the data center storage backup system is not in a normal state.

[0174] In one possible implementation of this application embodiment, when the state determination module 202 performs time-domain analysis and frequency-domain analysis on historical fault data to obtain a first time-domain relationship between historical fault frame messages and abnormal fault forms, a second time-domain relationship between historical fault frames and abnormal network communication data, and a frequency-domain feature vector matrix corresponding to the historical fault data, it is specifically used for:

[0175] The historical fault data is divided into sliding windows, and the abnormal fault type corresponding to each historical fault message in each historical window is parsed.

[0176] Based on the abnormal fault type corresponding to each historical fault message in each historical window, the first temporal relationship between historical fault messages and abnormal fault types is determined through correlation analysis.

[0177] Identify the abnormal network communication data within each historical window, calculate the cross-correlation of the historical window sequence, identify the temporal coupling relationship between historical fault frames and abnormal network communication data, and form a second temporal relationship;

[0178] Frequency domain analysis was performed on the abnormal fault forms and abnormal network communication data in the historical fault data to obtain the first frequency domain feature vector corresponding to the abnormal fault forms and the second frequency domain feature vector corresponding to the abnormal network communication data.

[0179] The first frequency domain feature vector and the second frequency domain feature vector are integrated to obtain the frequency domain feature vector matrix corresponding to the historical fault data.

[0180] In one possible implementation of this application embodiment, the historical fault data further includes historical fault frame data. When determining the repair method corresponding to the fault frame message, the recording module 204 is specifically used for:

[0181] The fault frame message is parsed to extract key feature vectors, and the key feature vectors are compared with the historical feature vectors corresponding to the historical fault frame data to obtain the comprehensive similarity. The historical fault frame data includes historical feature vectors and historical repair methods.

[0182] When there is a comprehensive similarity greater than the first similarity threshold, the comprehensive similarity greater than the first similarity threshold is determined as the first target similarity, and the historical repair method corresponding to the first target similarity is obtained and determined as the repair method corresponding to the fault frame message;

[0183] When there is no comprehensive similarity greater than the first similarity threshold, but there are multiple comprehensive similarities greater than the second similarity threshold, the comprehensive similarity greater than the second similarity threshold is determined as the second target similarity. Based on the random forest algorithm, the historical repair methods corresponding to the multiple second target similarities are combined to obtain the optimal strategy combination. The optimal strategy combination is determined as the repair method corresponding to the fault frame message. The second similarity threshold is less than the first similarity threshold.

[0184] When the overall similarity is not greater than the second similarity threshold, a fault propagation graph model is constructed based on each historical repair method. The key feature vector is input into the fault propagation graph model, and the expected success rate of each repair method output by the fault propagation graph model is obtained. The repair method with the highest expected success rate is selected as the repair method corresponding to the fault frame message.

[0185] In one possible implementation of this application embodiment, when the repair module 205 repairs the fault frame message based on the repair method, it is specifically used for:

[0186] Determine whether the repair method meets the conditions for automatic repair;

[0187] If the repair method meets the automatic repair conditions, a repair instruction is generated based on the repair method and sent to the repair terminal corresponding to the fault frame message.

[0188] If the repair method does not meet the automatic repair conditions, the basic repair method is obtained, the basic repair instruction corresponding to the basic repair method is generated, and the basic repair instruction is sent to the repair end corresponding to the fault frame message.

[0189] In one possible implementation of this application embodiment, when the repair method meets the automatic repair conditions, the repair module 205, when determining whether the fault frame message has been successfully repaired, is specifically used for:

[0190] Automatically acquire repair completion signals, and after acquiring the repair completion signal, obtain repair monitoring data and repair fault frame messages;

[0191] The repair monitoring data is compared with the abnormal message format and the abnormal network communication data format respectively to determine whether the repair monitoring data has an abnormal message format or / and an abnormal network communication data format.

[0192] If it is determined that there are no abnormal message formats or / and abnormal network communication data formats in the repair monitoring data, the fault frame message is parsed to obtain the key feature vector for repair, and based on the historical feature vector, it is determined whether the fault frame message has been successfully repaired.

[0193] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0194] See Figure 3 This application also describes an electronic device from the perspective of a physical device, such as... Figure 3 As shown, Figure 3 The illustrated electronic device 300 includes a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, for example, via a bus 302. Optionally, the electronic device 300 may also include a transceiver 304. It should be noted that in practical applications, the transceiver 304 is not limited to one type, and the structure of this electronic device 300 does not constitute a limitation on the embodiments of this application.

[0195] Processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 301 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0196] Bus 302 may include a pathway for transmitting information between the aforementioned components. Bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 302 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0197] The memory 303 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0198] The memory 303 is used to store application code that executes the solution of this application, and its execution is controlled by the processor 301. The processor 301 is used to execute the application code stored in the memory 303 to implement the content shown in the foregoing method embodiments.

[0199] Among them, electronic devices include, but are not limited to: mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers, and can also be servers, etc. Figure 3 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0200] This application provides a computer-readable storage medium storing a computer program that, when run on a computer, enables the computer to execute the corresponding content in the aforementioned method embodiments.

[0201] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0202] The above are only some embodiments of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for monitoring data center storage backup failures, the method comprising: The method comprises the following steps: Collecting device state data and obtaining current monitoring data, wherein the device state data is the state data of the device corresponding to the data center storage backup system; Based on the device state data and the current monitoring data, it is determined whether the data center storage backup system is in a normal state; If the data center storage backup system is not in a normal state, the fault type of the data center storage backup system is determined, which is device failure or data failure; When the fault type is data failure, the current monitoring data corresponding to the fault frame packet is recorded, and the repair method corresponding to the fault frame packet is determined; Based on the repair method, the fault frame packet is repaired, and it is determined whether the repair is successful; If the repair is successful, a repair report corresponding to the current monitoring data is generated, and the repair report is sent to the corresponding operation and maintenance terminal; The device state data includes device temperature, device speed and device vibration signal, and the current monitoring data includes backup software running log and network communication data, Wherein, based on the device state data and the current monitoring data, it is determined whether the data center storage backup system is in a normal state, which comprises: The device vibration signal is processed by the CNN subnetwork to obtain the spatial feature vector of the device, and the device temperature and the device speed are processed by the LSTM subnetwork to obtain the time sequence feature vector; The spatial feature vector and the time sequence feature vector are input into the encoder, the inter-modal correlation weight is calculated by the multi-head attention mechanism, and the fusion feature vector is generated; The fusion feature vector is input into the full connection classification network, and the output device health score is obtained; Based on the device health score, it is determined whether the data center storage backup system is in a normal state; If it is determined that the data center storage backup system is in a normal state, it is determined whether the data center storage backup system is in a normal state based on the current monitoring data; Based on the current monitoring data, it is determined whether the data center storage backup system is in a normal state, which comprises: Obtain historical fault data, which includes historical fault frame packets and historical network communication data corresponding to each historical fault frame; The historical fault data is analyzed in time domain and frequency domain to obtain the first time domain relationship between historical fault frame packets and abnormal fault forms, the second time domain relationship between historical fault frames and abnormal network communication data, and the frequency domain feature vector matrix corresponding to the historical fault data; Based on the first time domain relationship, the second time domain relationship and the frequency domain feature vector matrix, the abnormal packet form corresponding to the historical fault frame and the abnormal network communication data form are determined; The current monitoring data is compared with the abnormal packet form and the abnormal network communication data form respectively; If the current monitoring data contains a fault frame matching the abnormal packet form, or the frequency domain feature of the network communication data exceeds the preset threshold compared with the abnormal network communication data form, it is determined that the data center storage backup system is not in a normal state.

2. The data center storage backup failure monitoring method of claim 1, wherein, The time domain analysis and frequency domain analysis of the historical fault data obtain a first time domain relationship between the historical fault frame packet and the abnormal fault form, a second time domain relationship between the historical fault frame and the abnormal network communication data, and a frequency domain feature vector matrix corresponding to the historical fault data, including: The historical fault data is segmented by a sliding window, and the abnormal fault form corresponding to each historical fault packet in each historical window is parsed; Based on the abnormal fault form corresponding to each historical fault packet in each historical window, a first time domain relationship between the historical fault packet and the abnormal fault form is determined through correlation analysis; The abnormal network communication data in each historical window is determined, and the cross-correlation of the historical window sequence is calculated to identify the time sequence coupling relationship between the historical fault frame and the abnormal network communication data, forming a second time domain relationship; The abnormal fault form and the abnormal network communication data in the historical fault data are analyzed in the frequency domain respectively to obtain a first frequency domain feature vector corresponding to the abnormal fault form and a second frequency domain feature vector corresponding to the abnormal network communication data; The first frequency domain feature vector and the second frequency domain feature vector are integrated to obtain a frequency domain feature vector matrix corresponding to the historical fault data.

3. The method of claim 1, wherein, The historical fault data further includes historical fault frame data, and the determination of the repair mode corresponding to the fault frame packet includes: The fault frame packet is parsed to extract a key feature vector, and the key feature vector is compared with a historical feature vector corresponding to the historical fault frame data to obtain a comprehensive similarity, the historical fault frame data including a historical feature vector and a historical repair mode; When there is a comprehensive similarity greater than a first similarity threshold, the comprehensive similarity greater than the first similarity threshold is determined as a first target similarity, and the historical repair mode corresponding to the first target similarity is obtained as the repair mode corresponding to the fault frame packet; When there is no comprehensive similarity greater than the first similarity threshold, and there are multiple comprehensive similarities greater than a second similarity threshold, the comprehensive similarity greater than the second similarity threshold is determined as a second target similarity, and based on a random forest algorithm, the historical repair modes corresponding to the multiple second target similarities are combined to obtain an optimal strategy combination, and the optimal strategy combination is determined as the repair mode corresponding to the fault frame packet, wherein the second similarity threshold is less than the first similarity threshold. When the comprehensive similarity is all not greater than the second similarity threshold, a fault propagation graph model is constructed based on each historical repair method, the key feature vector is input into the fault propagation graph model, and expected success degrees of different repair methods output by the fault propagation graph model are obtained, and a repair method with the highest expected success degree is selected as a repair method corresponding to the fault frame packet.

4. The method of claim 3, wherein, The repairing the fault frame packet based on the repair method comprises: determining whether the repair method meets an automatic repair condition; if the repair method meets the automatic repair condition, generating a repair instruction based on the repair method and sending the repair instruction to a repair end corresponding to the fault frame packet; if the repair method does not meet the automatic repair condition, obtaining a basic repair method, generating a basic repair instruction corresponding to the basic repair method, and sending the basic repair instruction to the repair end corresponding to the fault frame packet.

5. The method of claim 4, wherein, When the repair method meets the automatic repair condition, the determining whether the fault frame packet is successfully repaired comprises: automatically collecting a repair completion signal, obtaining repair monitoring data and a repaired fault frame packet after the repair completion signal is collected; comparing the repair monitoring data with the abnormal packet form and the abnormal network communication data form to determine whether the repair monitoring data has the abnormal packet form or / and the abnormal network communication data form; if it is determined that the repair monitoring data does not have the abnormal packet form or / and the abnormal network communication data form, analyzing the repaired fault frame packet to obtain a repair key feature vector, and determining whether the fault frame packet is successfully repaired based on the historical feature vector.

6. A data center storage backup failure monitoring apparatus characterized by comprising: comprise: a collection module configured to collect device state data and obtain current monitoring data, the device state data being state data of a device corresponding to a data center storage backup system; a state determination module configured to determine whether the data center storage backup system is in a normal state based on the device state data and the current monitoring data; a type determination module configured to determine a fault type of the data center storage backup system if the data center storage backup system is not in the normal state, the fault type being a device fault or a data fault; a record module configured to record a fault frame packet corresponding to the current monitoring data and determine a repair method corresponding to the fault frame packet if the fault type is the data fault; a repair module configured to repair the fault frame packet based on the repair method and determine whether the fault frame packet is successfully repaired; a sending module configured to generate a repair report corresponding to the current monitoring data and send the repair report to a corresponding operation and maintenance terminal if the repair is successful; the device state data comprises device temperature, device rotation speed and device vibration signal, and the current monitoring data comprises running log of backup software and network communication data. The state determination module, when determining whether the data center storage backup system is in a normal state based on the device state data and the current monitoring data, is specifically configured to: The device vibration signal is processed by a CNN subnetwork to obtain a spatial feature vector of the device, and the device temperature and the device rotating speed are processed by an LSTM subnetwork to obtain a time sequence feature vector; The spatial feature vector and the time sequence feature vector are input into an encoder, and inter-modal correlation weights are calculated by a multi-head attention mechanism to generate a fusion feature vector; The fusion feature vector is input into a fully connected classification network, and an output device health score is obtained; Based on the device health score, it is determined whether the data center storage backup system is in a normal state; If it is determined that the data center storage backup system is in a normal state, it is determined whether the data center storage backup system is in a normal state based on the current monitoring data; The state determination module, when determining whether the data center storage backup system is in a normal state based on the current monitoring data, is specifically configured to: Obtain historical fault data, the historical fault data including historical fault frame packets and historical network communication data corresponding to each historical fault frame; Perform time domain analysis and frequency domain analysis on the historical fault data to obtain a first time domain relationship between historical fault frame packets and abnormal fault forms, a second time domain relationship between historical fault frames and abnormal network communication data, and a frequency domain feature vector matrix corresponding to the historical fault data; Based on the first time domain relationship, the second time domain relationship, and the frequency domain feature vector matrix, determine the abnormal packet form and the abnormal network communication data form corresponding to the historical fault frame; Compare the current monitoring data with the abnormal packet form and the abnormal network communication data form respectively; If there is a fault frame in the current monitoring data that matches the abnormal packet form, or the frequency domain features of the network communication data and the abnormal network communication data form have a similarity that exceeds a preset threshold, it is determined that the data center storage backup system is not in a normal state.

7. An electronic device, comprising: The electronic device includes: at least one processor; a memory; at least one application program, wherein the at least one application program is stored in the memory and is configured to be executed by the at least one processor, and the at least one application program is configured to execute the data center storage backup fault monitoring method of any one of claims 1-5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed in the computer, the computer executes the data center storage backup fault monitoring method of any one of claims 1-5.

Images