A fingerprint-based server operation and maintenance fault identification method and positioning method

Abstract

The present invention discloses a fingerprint-based server operation and maintenance fault identification method and positioning method. By collecting, screening, processing and converting operation and maintenance indicators, a "fingerprint" that can reflect the performance status of the server is formed, and a convolutional neural network is used to learn the fault fingerprint rules, thereby realizing real-time detection of power grid dispatching and control system server anomalies, rapid positioning of fault nodes, intelligent identification of fault types, assisting operation and maintenance personnel in isolating fault nodes, and ensuring the safe and stable operation of the power grid dispatching and control system.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grids, and in particular relates to a fingerprint-based server operation and maintenance fault identification method and positioning method. Background Art

[0002] With the advancement of integrated power grid operations, the scope of services in the field of grid dispatching and control continues to expand, and the number of servers deployed for applications is also growing. While modern cloud computing architectures provide fast and flexible deployment solutions for business rollouts, they also pose greater challenges for operations and maintenance personnel. The increase in the number of devices and the introduction of virtualization technology have made operations and maintenance more complex, making rapid fault location and isolation a challenge. Summary of the Invention

[0003] Purpose of the invention: In order to solve the problem that the increase in the number of devices and the introduction of virtualization technology make it difficult to quickly locate and isolate faults, the present invention proposes a fingerprint-based server operation and maintenance fault identification method and positioning method; by collecting, screening, processing and converting operation and maintenance indicators, a "fingerprint" that can reflect the performance status of the server is formed, and a convolutional neural network is used to learn the fault fingerprint rules, thereby realizing real-time detection of power grid dispatching and control system server anomalies, rapid positioning of fault nodes, intelligent identification of fault types, assisting operation and maintenance personnel to isolate fault nodes, and ensuring the safe and stable operation of the power grid dispatching and control system.

[0004] Technical solution: A fingerprint-based server operation and maintenance fault identification method, including the following steps:

[0005] Step 1: Based on various operation and maintenance indicators related to server operation and maintenance, sample each operation and maintenance indicator to obtain a sampled value corresponding to each operation and maintenance indicator, and resample the sampled value corresponding to each operation and maintenance indicator to obtain operation and maintenance indicator data corresponding to each operation and maintenance indicator;

[0006] Step 2: For each operation and maintenance indicator, determine whether its corresponding operation and maintenance indicator data is abnormal at different times. If it is abnormal, the operation and maintenance indicator data corresponding to the operation and maintenance indicator at that time is marked as abnormal; if it is not abnormal, the operation and maintenance indicator data corresponding to the operation and maintenance indicator at that time is marked as normal;

[0007] Step 3: Based on the labels of each operation and maintenance indicator data at different times, calculate the abnormal significance of each operation and maintenance indicator;

[0008] Step 4: Filter out the sensitive indicator set based on the abnormal significance of each operation and maintenance indicator;

[0009] Step 5: Based on the sensitive indicator set obtained in step 4, obtain the server performance status fingerprint dataset;

[0010] Step 6: Use the server performance status fingerprint dataset to train the convolutional neural network model to obtain a convolutional neural network model that can identify fault types;

[0011] Step 7: Based on the sensitive indicator set obtained in step 4, perform real-time sampling for each operation and maintenance indicator in the sensitive indicator set, preprocess the real-time sampled time series data to obtain a sampling matrix, and input the sampling matrix into the convolutional neural network model for fault type identification obtained in step 6 to obtain the fault detection result.

[0012] Furthermore, step 2 specifically includes:

[0013] For the operation and maintenance indicator m, the operation and maintenance indicator data at time t is m t , based on the sampling values ​​within the time range of [tT-1, tT], calculate the mean and standard deviation, recorded as mean t and std t ;

[0014] Judge m t Does it satisfy formula (1)? If it does, it means m t At time t, the abnormality is t At time t, it is marked as abnormal; if it is not satisfied, it means m t Normal at time t, m t At time t, it is marked as normal;

[0015] m t <mean t -3*std t or m t >mean t +3*std t (1).

[0016] Furthermore, step 3 specifically includes:

[0017] Assume that the total running time of the server is N and the failure time is N a , the non-failure time is NN a During the fault period, the number of times the operation and maintenance indicator data corresponding to the operation and maintenance indicator m is marked as abnormal is m a Times, during the non-fault period, the number of times the operation and maintenance indicator data corresponding to the operation and maintenance indicator m is marked as abnormal is m b times, then the abnormal significance of the operation and maintenance index m is:

[0018]

[0019] Where S(m) is the abnormal significance of the operation and maintenance indicator m.

[0020] Furthermore, step 4 specifically includes:

[0021] For each server failure type, select k indicators whose abnormal significance exceeds the threshold from various operation and maintenance indicators related to server operation and maintenance, thereby obtaining n×k indicators, where n is the number of server failure types;

[0022] From the n×k indicators, select the indicators that appear more than or equal to 2 times to obtain the sensitive indicator set.

[0023] Furthermore, step 5 specifically includes:

[0024] Let the current sampling time be t0, and take the time series data of each indicator in the sensitive indicator set within the time range of [t0-T, t0] to form a matrix. Each column in the matrix is ​​the time series data of an indicator;

[0025] Two new columns of data are added, which are calculated by calculating the difference between the number of packets received and sent by the network interface between indicators and the proportion of the difference between the number of packets received and sent by the network interface in the number of packets received;

[0026] Perform first-order difference calculations on the data in the matrix and the two new columns of data, and remove the missing values ​​in the first row introduced by the difference calculation to obtain a new matrix;

[0027] Each column in the new matrix is ​​linearly transformed using the deviation normalization method, and the values ​​are mapped to the interval [0, 1] to obtain the server performance status fingerprint dataset.

[0028] Furthermore, step 6 specifically includes:

[0029] In the server performance status fingerprint data set, the server performance status fingerprint data during the fault period is taken to construct a negative sample data set, and the server performance status fingerprint data during the non-fault period is taken to construct a positive sample data set;

[0030] Randomly extract two samples a and b from the negative sample data set, select a random coefficient α, α∈(0,1), and add negative samples to the negative sample data set. The newly added negative samples are x=α*a+(1-α)*b, thereby obtaining a new negative sample data set;

[0031] Extract C samples from the positive sample dataset and C samples from the new negative sample dataset, and divide the 2C samples into training set, validation set and test set;

[0032] The three-layer convolutional neural network based on the Kares framework is trained, verified and tested using training sets, validation sets and test sets. When the accuracy on the test set is greater than 90%, a convolutional neural network model that can identify fault types is obtained.

[0033] Furthermore, the real-time sampled time series data is preprocessed to obtain a sampling matrix, including:

[0034] Take the time series data of each indicator in the sensitive indicator set at the current sampling moment to form a matrix, where each column in the matrix is ​​the time series data of one indicator;

[0035] Two new columns of data are added, which are calculated by calculating the difference between the number of packets received and sent by the network interface and the proportion of the difference between the number of packets received and sent by the network interface in the number of packets received.

[0036] Perform first-order difference calculations on the data in the matrix and the two new columns of data, and remove the missing values ​​in the first row introduced by the difference calculation to obtain a new matrix;

[0037] The deviation normalization method is used to perform a linear transformation on each column in the new matrix, mapping the values ​​to the interval [0, 1] to obtain a sampling matrix.

[0038] Furthermore, step 8 is also included:

[0039] For the same server, determine whether the fault detection result at the current sampling moment and the fault detection result at the previous sampling moment are the same type of fault. If they are the same type of fault, it is determined that the server has indeed experienced the fault; if they are not the same type of fault, it is determined that the server has not experienced a fault.

[0040] The present invention discloses a fingerprint-based server operation and maintenance fault location method, comprising the following steps:

[0041] Step 1: Create a topology diagram based on the deployment of the power grid dispatching control system and the call relationships between applications on each server; and calculate the troubleshooting priority of each server based on the topology diagram;

[0042] Step 2: Use the server operation and maintenance fault identification method to monitor each server for faults and obtain the server where the fault occurs;

[0043] Step 3: Determine whether the number of faulty servers is greater than or equal to 2. If so, arrange the faulty servers in descending order of priority to generate a troubleshooting sequence.

[0044] Step 4: Troubleshoot the problem one by one according to the troubleshooting sequence, and finally locate the root cause server;

[0045] The server operation and maintenance fault identification method is a server operation and maintenance fault identification method based on fingerprint.

[0046] Furthermore, in step 1, the topology diagram is established based on the deployment of the power grid dispatching control system and the calling relationship between the applications on each server, specifically including:

[0047] Each server acts as a node. Based on the calling relationship between the applications on each server, when there is a calling relationship between two servers, a directed edge is established between the two nodes, from the service caller to the service callee, thereby establishing a topological relationship diagram.

[0048] Furthermore, the above-mentioned calculation of the troubleshooting priority of each server based on the topology diagram specifically includes:

[0049] For each server, the troubleshooting priority of the server is calculated based on the out-degree, in-degree, and depth of the server in the topology diagram:

[0050] Priority n =Indegree n +Depth n -Outdegree n (5)

[0051] In the formula, Priority n Indicates the troubleshooting priority of server n; Outdegree n Indicates the out-degree of server n, which refers to how many other servers server n calls; Indegree n Indicates the in-degree of server n, which refers to how many servers server n is called by; Depth n Indicates the depth of server n. The depth refers to the deepest depth from the human-machine interface to server n.

[0052] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0053] (1) The present invention can reduce the cache volume of real-time data of operation and maintenance indicators and speed up anomaly detection by selecting a small number of indicator sets to form a fingerprint that characterizes server performance;

[0054] (2) The present invention uses a convolutional neural network to learn fault fingerprint patterns to automatically and accurately determine the fault category. When multiple server faults are detected simultaneously, a fault location and troubleshooting sequence can be generated based on the application deployment architecture and priority, thereby improving fault location efficiency.

[0055] (3) The present invention can assist operation and maintenance personnel in analyzing fault types, isolating faults in a timely manner, and ensuring the safe and stable operation of the power grid dispatching and control system. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1The figure is a flowchart of a fingerprint-based method for identifying and locating operation and maintenance faults in a power grid dispatching and control system;

[0057] Figure 2 This is a schematic diagram of the server topology relationship. DETAILED DESCRIPTION

[0058] The technical solution of the present invention will now be further described with reference to the accompanying drawings and embodiments.

[0059] Example 1:

[0060] A fingerprint-based server operation and maintenance fault identification method of this embodiment includes the following steps:

[0061] Step 1: Based on various operation and maintenance indicators related to server operation and maintenance, sample each operation and maintenance indicator to obtain a sampled value corresponding to each operation and maintenance indicator, and resample the sampled value corresponding to each operation and maintenance indicator to obtain operation and maintenance indicator data corresponding to each operation and maintenance indicator;

[0062] Step 2: For each operation and maintenance indicator, determine whether its corresponding operation and maintenance indicator data is abnormal at different times. If it is abnormal, the operation and maintenance indicator data corresponding to the operation and maintenance indicator at that time is marked as abnormal; if it is not abnormal, the operation and maintenance indicator data corresponding to the operation and maintenance indicator at that time is marked as normal;

[0063] Step 3: Based on the marks of each operation and maintenance indicator at different times, calculate the abnormal significance of each operation and maintenance indicator;

[0064] Step 4: Filter out the sensitive indicator set based on the abnormal significance of each operation and maintenance indicator;

[0065] Step 5: Based on the sensitive indicator set obtained in step 4, obtain the server performance status fingerprint dataset;

[0066] Step 6: Use the server performance status fingerprint dataset to train the convolutional neural network model to obtain a convolutional neural network model that can identify fault types;

[0067] Step 7: Based on the sensitive indicator set obtained in step 4, perform real-time sampling for each operation and maintenance indicator in the sensitive indicator set, preprocess the real-time sampled time series data to obtain a sampling matrix, and input the sampling matrix into the convolutional neural network model for fault type identification obtained in step 6 to obtain the fault detection result.

[0068] Example 2:

[0069] By collecting, screening, processing and transforming operation and maintenance indicators, a "fingerprint" that can reflect the performance status of the server is formed, and a convolutional neural network is used to learn the fault fingerprint rules, so as to realize real-time detection of server anomalies in the power grid dispatching and control system and intelligently identify the fault type.

[0070] like Figure 1 As shown, a fingerprint-based server operation and maintenance fault identification method of this embodiment mainly includes the following steps:

[0071] Step 1: The scope of operation and maintenance data collection for the power dispatching and control system includes servers, network equipment, storage devices, and power-specific equipment. The corresponding performance indicators include hundreds of indicators such as CPU utilization, disk utilization, disk read and write rates, and network traffic. Due to differences in sampling methods, sampling periods, and sampling statistical windows, the original operation and maintenance indicator data has inconsistent sampling times and frequencies. Through resampling, the sampling period of all operation and maintenance indicators is set to 1 minute, and the operation and maintenance indicator data corresponding to each operation and maintenance indicator is obtained;

[0072] Step 2: The operation and maintenance indicator data is continuous time series data. This embodiment uses the 3-standard deviation method to determine whether a single operation and maintenance indicator time series data is abnormal. After abnormality determination, the continuous time series data of the operation and maintenance indicator is converted into a sequence containing only "abnormal" and "normal".

[0073] Specifically, the server operation and maintenance indicator m, the sampling value at time t is m t , based on the sampling values ​​in the range of [tT-1, tT] before time t, calculate the mean and standard deviation, which are respectively recorded as mean t and std t If m t If formula (1) is satisfied, then m t If it is abnormal, it is recorded as “abnormal”; otherwise, it is considered that m t It is normal, which is recorded as “normal.” In this embodiment, T=30 minutes.

[0074] m t <mean t -3*std t or m t >mean t +3*std t (1)

[0075] Step 3: Calculate the abnormal significance of the operation and maintenance indicator. An operation and maintenance indicator that can significantly reflect a server failure must exhibit abnormal behavior only during the failure period and not during normal periods. Therefore, the abnormal significance of an operation and maintenance indicator is calculated as the ratio of the corresponding operation and maintenance indicator data marked as "abnormal" during the server failure period to the ratio of the corresponding operation and maintenance indicator data marked as "abnormal" during the server's normal period.

[0076] Specifically, let the server's running time be N minutes, of which the failure period is N a Minutes, during the fault period, the operation and maintenance indicator data corresponding to the operation and maintenance indicator m is marked as "abnormal" for m a times, during the non-fault period, the number of times the operation and maintenance indicator data corresponding to the operation and maintenance indicator m is marked as "abnormal" is m b times, then the abnormal significance S(m) of the operation and maintenance indicator m is:

[0077]

[0078] Step 4: In order to improve the efficiency of server status detection and reduce the amount of data cache during real-time detection, some indicators are selected according to the abnormal significance as the sensitive indicator set that constitutes the server performance status fingerprint.

[0079] Specifically, server failure types include: CPU failure, CPU ramp-up failure, memory exhaustion failure, disk space exhaustion failure, disk read I / O failure, disk write I / O failure, and network failure. First, for each failure type, the k most significant indicators are selected from each operation and maintenance indicator. In this embodiment, k = 5. Then, among the k most significant indicators, those that occur two or more times are selected as the sensitive indicator set that constitutes the server performance status fingerprint. In this embodiment, the selected sensitive indicator set includes: system.cpu.pct_usage CPU usage, system.load.1 system load 1-minute statistical value, system.mem.pct_usage memory usage, system.disk.pct_usage disk usage, system.io.r_await disk read response time, system.io.w_await disk write response time, system.io.svctm average service time for each device I / O operation, system.net.packets_in.count number of data packets received by the interface, system.net.packets_out.count number of data packets sent by the interface, system.os.nofile.current number of currently open file handles in the system, a total of 10 operation and maintenance indicators.

[0080] Step 5: Let the current sampling time be t0, select the time range [t0-T, t0], and the time series data of the 10 operation and maintenance indicators form a T*10 matrix, that is, each column is the time series data of an operation and maintenance indicator. In this embodiment, let T=30.

[0081] Step 6: Add a new column by calculating the difference between the number of packets received and sent by the network interface between the operation and maintenance indicators and calculating the proportion of the difference between the number of packets received and sent by the network interface in the network interface received data.

[0082] Calculate the difference between the number of packets received and sent by the network interface system.net.packets_in_out_diff.count:

[0083] system.net.packets_in_out_diff.count=system.net.packets_in.count-system.net.packets_out.count (3)

[0084] In the formula, system.net.packets_in.count indicates that the network interface receives data packets, and system.net.packets_out.count indicates that the network interface sends data packets.

[0085] Calculate the ratio of the difference between the packets received and sent by the network interface to the data received by the network interface system.net.packets_in_out_diff.percent:

[0086]

[0087] Where system.net.packets_in_out_diff.count represents the difference between the number of packets received and sent by the network interface.

[0088] Step 7: Perform first-order difference calculations on the 10 columns of original time series data and the 2 columns of new data, and remove the missing values ​​in the first row introduced by the difference calculation to obtain a (T-1)*12-dimensional matrix.

[0089] Step 8: Use the deviation normalization method to perform a linear transformation on each column in the matrix, map the values ​​to the [0, 1] interval, and obtain the server performance status fingerprint dataset.

[0090] The main purpose of steps 5 to 8 is to construct a server performance status fingerprint dataset by processing and transforming the sensitive indicator set.

[0091] Step 9: Divide the server performance status fingerprint dataset into positive and negative samples. Specifically, the data during the fault period is used to construct a negative sample dataset, and the data during the non-fault period is used to construct a positive sample dataset.

[0092] Step 10: Because the failure period is short, the number of negative samples is far less than the number of positive samples. A new negative sample is created using a sample augmentation method. Specifically, the sample augmentation method involves randomly extracting two samples, a and b, from the original negative sample dataset. The random coefficient α is selected, where α∈(0,1). The new negative sample is x=α*a+(1-α)*b, thus obtaining a new negative sample dataset.

[0093] Step 11: Extract C samples from the positive sample dataset and C samples from the new negative sample dataset to obtain 2C samples. These 2C samples are divided into a training set, a validation set, and a test set according to a 6:2:2 ratio. In this embodiment, a three-layer convolutional neural network based on the Kares framework is used as the classification model. By adjusting parameters such as the number and size of convolution kernels and the model learning rate, the model accuracy on the test set is greater than 90%, ultimately obtaining a convolutional neural network model for fault type identification.

[0094] The main purpose of steps 9 to 11 is to use the convolutional neural network model to learn the fault fingerprint rules.

[0095] Step 12: To avoid misjudgments caused by data jitter, during real-time detection, a server is considered faulty only if its fingerprint is identified as the same fault two or more times in a row. The fingerprint here is a matrix generated by real-time sampling based on indicators in the sensitive indicator set and processing the sampled values ​​through steps 5 through 8. A server's fingerprint being identified as the same fault two or more times in a row means that the server's fingerprint was identified as the same type of fault at two consecutive sampling moments.

[0096] Example 3:

[0097] This embodiment, based on Embodiment 1 or 2, discloses a fingerprint-based server operation and maintenance fault location method, including the following steps:

[0098] Step 1: Create a topology diagram based on the deployment of the power grid dispatching control system and the call relationships between applications on each server; and calculate the troubleshooting priority of each server based on the topology diagram;

[0099] Step 2: Use the server operation and maintenance fault identification method to monitor each server for faults and obtain the server where the fault occurs;

[0100] Step 3: Determine whether the number of faulty servers is greater than or equal to 2. If so, arrange the faulty servers in descending order of priority to generate a troubleshooting sequence.

[0101] Step 4: Troubleshoot the problem one by one according to the troubleshooting sequence, and finally locate the root cause server;

[0102] The server operation and maintenance fault identification method is a fingerprint-based server operation and maintenance fault identification method disclosed in the above embodiment.

[0103] Example 4:

[0104] This embodiment, based on Embodiment 1 or Embodiment 2, discloses a fingerprint-based server operation and maintenance fault location method. When multiple servers are detected to have faults, it is necessary to locate the root cause server, including the following steps:

[0105] Step 1: There is a calling relationship between the applications on the server, so a topology diagram is established based on the deployment of the power grid dispatching control system and the calling relationship between the applications on each server, such as Figure 2 As shown in the figure, each server is a node. If there is a calling relationship between two servers, a directed edge is established between the two nodes, pointing from the service caller to the callee.

[0106] Step 2: Calculate the troubleshooting priority of the server based on the server's out-degree, in-degree, and depth in the topology diagram.

[0107] Among them, outdegree refers to how many other server nodes the server node calls, recorded as Outdegree; indegree refers to how many servers call the server node, recorded as Indegree; depth refers to the deepest depth from the "human-computer interface" node to the server node, that is, when there are multiple paths, the longest path length, recorded as Depth.

[0108] Let the troubleshooting priority be Priority, and the troubleshooting priority of server n be Priority n , calculated according to the following formula:

[0109] Priority n =Indegree n +Depth n -Outdegree n (5)

[0110] In the formula, Priority n Indicates the troubleshooting priority of server n; Outdegree n Indicates the outdegree of server n; Indegree nDepth represents the in-degree of server n; n Indicates the depth of server n.

[0111] Step 3: Use the server operation and maintenance fault identification method to monitor each server for faults and obtain the server where the fault occurs;

[0112] Step 4: Determine whether the number of faulty servers is greater than or equal to 2. If so, arrange the faulty servers in descending order of priority to generate a troubleshooting sequence.

[0113] Step 5: Troubleshoot the problem one by one according to the troubleshooting sequence, and finally locate the root cause server;

[0114] The server operation and maintenance fault identification method is a fingerprint-based server operation and maintenance fault identification method disclosed in Example 1 or Example 2.

[0115] Example 5:

[0116] This embodiment discloses a computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps disclosed in any of the above embodiments are implemented. The computer device may be a server. The computer device comprises a processor, a memory, and a network interface connected via a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store data. The network interface of the computer device is configured to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a fingerprint-based server operation and maintenance fault identification method or a fingerprint-based server operation and maintenance fault location method is implemented.

[0117] Example 6:

[0118] This embodiment discloses a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps disclosed in any one of the above embodiments are implemented.

[0119] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0120] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0121] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A fingerprint-based server operation and maintenance fault identification method, characterized by: The following steps are involved: Step 1: Based on various operation and maintenance indicators related to server operation and maintenance, sample each operation and maintenance indicator to obtain a sampled value corresponding to each operation and maintenance indicator, and resample the sampled value corresponding to each operation and maintenance indicator to obtain operation and maintenance indicator data corresponding to each operation and maintenance indicator; Step 2: For each operation and maintenance indicator, determine whether its corresponding operation and maintenance indicator data is abnormal at different times. If it is abnormal, the operation and maintenance indicator data corresponding to the operation and maintenance indicator at that time is marked as abnormal; if it is not abnormal, the operation and maintenance indicator data corresponding to the operation and maintenance indicator at that time is marked as normal; Step 3: Based on the labels of each operation and maintenance indicator data at different times, calculate the abnormal significance of each operation and maintenance indicator; Step 4: For each server failure type, select k indicators whose abnormal significance exceeds the threshold from various operation and maintenance indicators related to server operation and maintenance, thereby obtaining n × k indicators, where n is the number of server failure types. From these n × k indicators, select those that appear at least twice to obtain the sensitive indicator set. Step 5: Let the current sampling time be t0, and take the time series data of each indicator in the sensitive indicator set within the time range of [t0-T, t0] to form a matrix. Each column in the matrix is ​​the time series data of an indicator; by calculating the difference between the network interface receiving and sending data packets between the indicators and the proportion of the difference between the network interface receiving and sending data packets in the number of received packets, add two new columns of data; perform first-order difference calculations on the data in the matrix and the new two columns of data, and remove the missing values ​​in the first row introduced by the difference calculation to obtain a new matrix; use the deviation normalization method to perform a linear transformation on each column in the new matrix, mapping the values ​​to the interval [0, 1], and obtain the server performance status fingerprint dataset; Step 6: In the server performance status fingerprint dataset, the server performance status fingerprint data during the fault period is taken to construct a negative sample dataset, and the server performance status fingerprint data during the non-fault period is taken to construct a positive sample dataset; two samples a and b are randomly extracted from the negative sample dataset, and a random coefficient α is selected, α∈(0,1), and negative samples in the negative sample dataset are added. The newly added negative samples are x=α*a+(1-α)*b, thereby obtaining a new negative sample dataset; C samples are extracted from the positive sample dataset and C samples are extracted from the new negative sample dataset, and the 2C samples are divided into a training set, a validation set, and a test set; the training set, validation set, and test set are used to train, validate, and test the three-layer convolutional neural network based on the Kares framework. When the accuracy on the test set is greater than 90%, a convolutional neural network model that can identify fault types is obtained; Step 7: Based on the sensitive indicator set obtained in step 4, perform real-time sampling for each operation and maintenance indicator in the sensitive indicator set, preprocess the real-time sampled time series data to obtain a sampling matrix, and input the sampling matrix into the convolutional neural network model for fault type identification obtained in step 6 to obtain the fault detection result.

2. A fingerprint-based server operation and maintenance fault identification method according to claim 1, characterized in that: Step 2 specifically includes: For the operation and maintenance indicator m, the operation and maintenance indicator data at time t is m t , based on the sampling values ​​within the time range of [tT-1, tT], calculate the mean and standard deviation, recorded as mean t and std t ; Judge m t Does it satisfy formula (1)? If it does, it means m t At time t, the abnormality is t At time t, it is marked as abnormal; if it is not satisfied, it means m t Normal at time t, m t At time t, it is marked as normal; m t <mean t -3*std t or m t >mean t +3*std t (1).

3. The fingerprint-based server operation and maintenance fault identification method according to claim 1, characterized in that: Step 3 specifically includes: Assume that the total running time of the server is N and the failure time is N a , the non-failure time is NN a During the fault period, the number of times the operation and maintenance indicator data corresponding to the operation and maintenance indicator m is marked as abnormal is m a Times, during the non-fault period, the number of times the operation and maintenance indicator data corresponding to the operation and maintenance indicator m is marked as abnormal is m b times, then the abnormal significance of the operation and maintenance index m is: Where S(m) is the abnormal significance of the operation and maintenance indicator m.

4. The fingerprint-based server operation and maintenance fault identification method according to claim 1, characterized in that: The preprocessing of the real-time sampled time series data to obtain a sampling matrix includes: Take the time series data of each indicator in the sensitive indicator set at the current sampling moment to form a matrix, where each column in the matrix is ​​the time series data of one indicator; Two new columns of data are added, which are calculated by calculating the difference between the number of packets received and sent by the network interface and the proportion of the difference between the number of packets received and sent by the network interface in the number of packets received. Perform first-order difference calculations on the data in the matrix and the two new columns of data, and remove the missing values ​​in the first row introduced by the difference calculation to obtain a new matrix; The deviation normalization method is used to perform a linear transformation on each column in the new matrix, mapping the values ​​to the interval [0, 1] to obtain a sampling matrix.

5. The fingerprint-based server operation and maintenance fault identification method according to claim 1, characterized in that: Also includes step 8: For the same server, determine whether the fault detection result at the current sampling moment and the fault detection result at the previous sampling moment are the same type of fault. If they are the same type of fault, it is determined that the server has indeed experienced the fault; if they are not the same type of fault, it is determined that the server has not experienced a fault.

6. A fingerprint-based server operation and maintenance fault location method, characterized by: The following steps are involved: Step 1: Establish a topology diagram based on the deployment of the power grid dispatching control system and the calling relationships between applications on each server; And based on the topology diagram, calculate the troubleshooting priority of each server; Step 2: Use the server operation and maintenance fault identification method to monitor each server for faults and obtain the server where the fault occurs; Step 3: Determine whether the number of faulty servers is greater than or equal to 2. If so, arrange the faulty servers in descending order of priority to generate a troubleshooting sequence. Step 4: Troubleshoot the problem one by one according to the troubleshooting sequence, and finally locate the root cause server; The server operation and maintenance fault identification method is a fingerprint-based server operation and maintenance fault identification method as described in any one of claims 1 to 5.

7. The fingerprint-based server operation and maintenance fault location method according to claim 6, characterized in that: In step 1, the topology diagram is established based on the deployment of the power grid dispatching control system and the calling relationship between the applications on each server, specifically including: Each server acts as a node. Based on the calling relationship between the applications on each server, when there is a calling relationship between two servers, a directed edge is established between the two nodes, from the service caller to the service callee, thereby establishing a topological relationship diagram.

8. The fingerprint-based server operation and maintenance fault location method according to claim 6, characterized in that: The above-mentioned calculation of the troubleshooting priority of each server based on the topology diagram specifically includes: For each server, the troubleshooting priority of the server is calculated based on the out-degree, in-degree, and depth of the server in the topology diagram: Priority n =Indegree n +Depth n -Outdegree n (5) In the formula, Priority n Indicates the troubleshooting priority of server n; Outdegree n Indicates the out-degree of server n, which refers to how many other servers server n calls; Indegree n Indicates the in-degree of server n, which refers to how many servers server n is called by; Depth n Indicates the depth of server n. The depth refers to the deepest depth from the human-machine interface to server n.

Images