A devops platform monitoring cycle intelligent regulation method

Abstract

The application discloses a kind of DevOps platform monitoring cycle intelligent regulation and control methods, first, system multidimensional index data stream is acquired, forms instant multidimensional vector by stream computing, window processing, feature combination etc., to judge system state. Then, the emergency degree is determined by quantifying event information using event level model. Then, the confidence vector is obtained by operating the preset rule base, the corresponding scheduling strategy is triggered, and the resources are allocated according to fuzzy logic and event emergency degree. At the same time, the event priority is determined by combining expert experience and fuzzy logic, and the flexible management plan is executed and the result is evaluated. Finally, the rule weight ratio is adjusted dynamically according to the rule base to optimize the judgment of system multidimensional index. The application realizes dynamic regulation and control of monitoring cycle, accurate evaluation of system and event, intelligent scheduling and disposal, and improves the stability, efficiency and adaptability of DevOps platform.

Description

Technical Field

[0001] This invention belongs to the field of system monitoring, and specifically relates to a method for intelligent control of monitoring cycle of DevOps platform. Background Technology

[0002] Existing monitoring methods are insufficient for comprehensive analysis and in-depth mining of multi-dimensional system indicators. DevOps platforms generate massive amounts of multi-dimensional data during operation, including system load, service response time, and various types of event information. Simply collecting and analyzing this data in isolation is insufficient to comprehensively and accurately grasp the overall operational status of the system. For example, focusing solely on system load or service response time may overlook the potential impact of event types and their quantities on the system's state. Furthermore, the lack of effective data integration and feature extraction methods makes it impossible to extract key information that truly characterizes the system's state from this complex data, hindering accurate assessment of system health and early warning of potential problems.

[0003] Furthermore, traditional monitoring and handling mechanisms lack intelligent scheduling strategies and flexible response plans when facing system anomalies or events. When a system experiences an unhealthy state or an emergency, it often only issues simple alarms and allocates limited resources according to pre-set fixed rules. It cannot comprehensively judge and take more reasonable and efficient scheduling measures based on factors such as the urgency of the event, the actual state of the system, and historical experience. At the same time, there is a lack of effective evaluation and feedback mechanisms for the results of handling operations, making it impossible to optimize and adjust management strategies in a timely manner based on the handling effects. This results in the system being inadequate when dealing with complex and ever-changing system operating conditions.

[0004] In summary, the technical problem that this invention aims to solve is the lack of flexibility in monitoring cycles, insufficient multi-dimensional indicator analysis, and unintelligent scheduling and handling strategies in existing DevOps platform monitoring. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide an intelligent control method for monitoring cycle of DevOps platform. Compared with traditional fixed scheduling strategy, this invention can better adapt to various changes in the system operation process, improve resource utilization, avoid over-allocation or under-allocation of resources, and ensure that the system can operate efficiently under different load conditions.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0007] A method for intelligent control of monitoring cycle in a DevOps platform, comprising the following steps:

[0008] S1. During the operation of the DevOps platform, acquire multi-dimensional indicator data streams of system operation. The multi-dimensional indicator data streams include system load, service response time, and the specific types and quantities of each event that occurs within the collection period. Based on the multi-dimensional indicator data streams, form a multi-dimensional vector representing the system state to obtain an instant multi-dimensional vector.

[0009] S2. Quantify the event information contained in the real-time multidimensional vector of the DevOps platform through the pre-built event level model, match the corresponding level parameters for each event, form a level vector from the level parameters of each event, and determine the urgency value of the corresponding event based on the specific values ​​contained in the level vector.

[0010] S3. Using a preset rule base, perform logical operations on the system status information values ​​in the DevOps platform's real-time multidimensional vector and the event urgency values ​​in the level vector, and infer the confidence vector of the corresponding system status information and the confidence vector of the corresponding event urgency information under the given condition based on the rule base.

[0011] S4. If more than a set number of system status information in the confidence vector of the DevOps platform system status information are in an unhealthy state, then issue a system status alarm command to the system and trigger the fuzzy scheduling strategy mechanism based on the numerical judgment of the urgency of the event.

[0012] S5. Obtain the set of expert experience processing rules corresponding to different types of events that have been set up. Use fuzzy logic judgment technology in combination with the set of expert experience processing rules to comprehensively calculate the real-time indicators and trigger events of the DevOps platform system and determine the priority parameters after comprehensive calculation of the system status and urgency of the events.

[0013] S6. Determine several corresponding flexible management plans through priority parameters, execute specific handling operation instruction sets according to the process in the flexible management plan, obtain the instruction handling results and feed them back to the information processing stage for result evaluation;

[0014] S7. Determine the dynamically adjusted rule weight ratio based on the rule base evaluation results, and re-evaluate the multi-dimensional indicators of the DevOps platform system based on the updated rule base to determine the direction and parameter values ​​for the next execution of the information processing mechanism.

[0015] Preferably, the sub-step of S1 is: S1.1, according to the system load data stream, a streaming computing engine is used to continuously receive system indicator data, and the collected load data is processed by setting a window function to obtain the average load within a set time window;

[0016] Assume the system load data flow is as follows: ,in Representing a time series, the window function is: ,

[0017] Then the average load ,in The number of data points within the window. For the first in the window A point in time;

[0018] S1.2, Combine the average load and service latency information flow of the fusion system to calculate the latency within a fixed time range and obtain statistical data; let the service latency information flow be... The sliding window size is M, and the service latency statistics are as follows:

[0019] ;

[0020] in For the first in the sliding window A point in time;

[0021] S1.3. Integrate the statistical data of average load and service latency to form a two-dimensional information list to represent the comprehensive information of system load and service latency; let the integrated two-dimensional list be: ;

[0022] S1.4 Obtain a two-dimensional information list of all event types and their counts within the window period; let the event type set be: ,in The total number of event types corresponds to the number of events. Then, a two-dimensional list of event types and frequencies. ;

[0023] S1.5. Associate the load and service latency information list and the event type and frequency list, count the number of events occurring within a set time window, and count the number of events of different categories within the set time period to obtain a third list of the frequency and type of each category, and merge them into a third list; let the merged third list be: ;

[0024] S1.6. Using feature combination techniques, the three lists output in the previous step are integrated into a multi-dimensional data stream, transforming them into a data format that comprehensively reflects the multi-dimensional characteristics of the system; let the feature combination function be... The combined data is as follows: ;

[0025] S1.7. Treat the data after feature combination processing as a set of high-dimensional feature vectors; let the high-dimensional feature vectors be... ,in For the feature vector dimension, For combined data Feature values ​​obtained through mapping or transformation;

[0026] S1.8. Use time-series correlation clustering techniques to calculate time-series correlations in order to uncover potential patterns in the data over time; let the time-series clustering algorithm be... The clustering results are ;

[0027] S1.9. The correlation between the data and real-time is calculated by constructing a time decay coefficient. If the time decay coefficient of data at a certain moment is greater than a preset value, the data is discarded to ensure its timeliness. Let the time decay coefficient be: ,in As the attenuation factor, For reference time points,

[0028] like , If the preset time decay coefficient threshold is used, the corresponding data will be discarded;

[0029] S1.10 Obtain a list set containing time information; let this set be:

[0030] ;

[0031] in For the number of data points, For the first The time for each data point For the corresponding feature vector;

[0032] S1.11. Use PCA (Principal Component Analysis) to reduce the dimensionality of the obtained time list set, extract key information, and reduce data dimensionality. Let the PCA transformation matrix be Φ, and the dimensionality-reduced principal component vectors be:

[0033] ;

[0034] in The dimension after dimensionality reduction.

[0035] The PCA algorithm specifically calculates the covariance matrix:

[0036] ;

[0037] in The mean of the eigenvectors is obtained, and then the eigenvalues ​​and eigenvectors of the covariance matrix are solved. The eigenvectors corresponding to the larger eigenvalues ​​are selected to form the transformation matrix Φ.

[0038] S1.12. Extract the vector group of the principal components in the set, and obtain the principal component with the largest proportion at each time step as the feature data to obtain the instantaneous multidimensional vector; let the instantaneous multidimensional vector be:

[0039] ;

[0040] in The final number of principal components selected. From The principal component values ​​are selected based on their proportions.

[0041] S1.13. Calculate the instantaneous multidimensional vector using an unsupervised method, calculate the Euclidean distance of the vector to the known system state category center, and determine the minimum distance to the center point of all known vectors; let the set of known system state category centers be:

[0042] ;

[0043] in The number of system state categories, a vector. Go to Category Center Euclidean distance:

[0044] ;

[0045] in Center of category The One eigenvalue;

[0046] S1.14. Determine which system state is most similar; if the distance between the instantaneous vector and the center of each known system state category is less than the state distance threshold, then the instantaneous vector is assigned to the most similar system state category. Let the state distance threshold be... ,like ,but Category ;

[0047] S1.15. The state similarity results of multiple time periods are processed by an adaptive deep neural network. Each system state is output to the deep network. If the network has a new output, the parameters of a certain system state are updated. The trained deep neural network is then used as the basis for decision-making in real-time state calculation.

[0048] Preferably, the S2 method is as follows: Let the event level model be... The event information is Then the event level parameter Let the level vector be ,in The number of event types and the numerical value of event urgency. according to The elements in the array are obtained through mapping relationships.

[0049] Preferably, sub-step S3 is as follows:

[0050] S3.1 Collect multi-dimensional vectors and level vectors generated during the operation of the DevOps platform system. The obtained information includes system status information values ​​and event urgency values, and construct an information database; let the system status information values ​​be... The urgency level of the event is [value missing]. The information database is ;

[0051] S3.2. Perform normalization processing on the system status information values ​​in the database. Based on the distribution of event urgency levels, divide the event urgency values ​​into multiple intervals to obtain the normalized database and the level value information of the divided intervals; let the normalization function be... The normalized system state information values ​​are:

[0052] ,

[0053] Let the urgency of the event be divided into intervals as follows: The corresponding level value is ;

[0054] S3.3. Divide the information into two parts: a state information group and a hierarchy information group. Perform clustering on the state information group. For each cluster center generated by the clustering, calculate the distance between other state information values ​​and the cluster center to obtain the clustering data. Let the state information group clustering algorithm be... The clustering results are as follows: ,

[0055] Let the set of cluster centers be... ,

[0056] Status information value To the cluster center Distance:

[0057] ;

[0058] in for The 1 eigenvalue, for The 1 eigenvalue, The dimension of the feature vector;

[0059] S3.4. Preset a threshold and compare the distance between the system status information value in the database and the threshold. If the value is higher than the threshold, determine that the current value is in a high-risk state, the corresponding status information group is in a high-risk state, and output an alarm command for the high-risk state to obtain the alarm command; let the threshold be... ,like If the condition is deemed high-risk, an alarm command will be output. ;

[0060] S3.5 Use alarm commands to query the numerical information of the level information group. According to the rules defined in the rule base, if the queried level information group information matches the threshold range of the high urgency level of the level numerical information, generate an alarm signal for the event, output the event alarm signal data, and obtain the information data of the emergency event.

[0061] S3.6. Train a convolutional neural network classification model using the training set data, set the network loss function and optimize the network to obtain a mature network. Input emergency event information data into the network, determine the matching relationship between the confidence levels of the output emergency event information and the system state information, and generate a confidence level association list; Let the training set be:

[0062] ;

[0063] in For training sample index, To provide the true confidence level label, the network loss function uses the mean squared error loss function:

[0064] ;

[0065] in The number of training samples. For the first The network output confidence level for each sample is used to adjust the network parameters using the gradient descent algorithm to minimize the loss function, resulting in a mature network. Then, the input emergency event information data is processed to obtain a confidence level association list. .

[0066] Preferably, the sub-step of S4 is as follows:

[0067] S4.1 Collect various state parameters of the DevOps platform system during runtime, construct a multi-dimensional system state information set, and obtain the system state vector after data cleaning;

[0068] S=(s1, s2; s i ), where s i The value representing the i-th state information;

[0069] S4.2, Based on the numerical value s of each state information in the system state vector S. i The system compares the data with a pre-defined healthy range, and calculates the number of unhealthy states, N. It then determines whether the number of unhealthy states N is greater than a set threshold T. If N > T, it outputs an alarm information set and generates an alarm command A. If N ≤ T, it returns to step S4.1 to continuously collect system status information. The healthy range is defined as follows:

[0070] ;

[0071] in For the first The health range interval of each status information, if If the status information is unhealthy, then the number of unhealthy statuses is counted: ;

[0072] in For indicator functions, the following conditions must be met: Otherwise ,like Output alarm information set and alarm commands ;

[0073] S4.3 Obtain the urgency values ​​of multiple events corresponding to the alert information set of the DevOps platform system, normalize the urgency values, and obtain the normalized urgency value E=(e1, e2; e i ), where e i Let the urgency value of the i-th event be represented by the normalization function. ,but ;

[0074] S4.4. Based on the normalized urgency value E, a fuzzy logic-based algorithm is used to determine the triggering scheduling level L. The core rule of the fuzzy logic algorithm is to determine the final scheduling level by considering the correspondence between different urgency value ranges and pre-set scheduling levels, while also taking into account the combined impact of the urgency levels of multiple events. The corresponding scheduling level is then obtained. Assume the fuzzy logic system has an input fuzzy set: , corresponding to fuzzy sets of different urgency levels

[0075] Output fuzzy set , corresponding to fuzzy sets of different scheduling levels;

[0076] Fuzzy Inference Rule Table This represents the correspondence between different input fuzzy sets and output fuzzy sets, achieved by fuzzifying the input... ;

[0077] S4.5 After determining the scheduling level L, select multiple scheduling strategies that meet the level from the pre-established scheduling strategy library, and obtain the fused scheduling strategy G through scheduling strategy fusion; execute the fused scheduling strategy G through the DevOps platform scheduling engine, and send the resource allocation information data stream through the communication module to determine the resource reallocation scheme.

[0078] Let the scheduling strategy library be ,in For the number of strategies, select the scheduling level. A subset of matching strategies: ,

[0079] Fusion scheduling strategy: ,in For the strategy fusion function, the resource allocation information data stream is set as follows: According to the fusion scheduling strategy Sure The content involves allocating resources and assigning them to specific objects to achieve resource reallocation.

[0080] Preferably, the sub-step of S5 is as follows:

[0081] S5.1 Collect historical operation and maintenance data and expert handling experience of the DevOps platform, classify, organize and analyze the historical operation and maintenance data and expert handling experience, and construct a set of event handling rules and a corresponding set of historical resource status data.

[0082] S5.2 When an event is triggered, the DevOps platform system's real-time operating status information is synchronously acquired from multiple monitoring devices. The operating status information is matched with the type and urgency of the triggered event, and the status information is assigned according to a preset fuzzy set interval. For each event type, the fuzzy set and membership function parameters are queried to calculate the membership degree of the assigned status information, resulting in several operating status membership degree data. Based on the expert experience rules set in the expert knowledge base, a set of rules corresponding to different states in the records of the same type of emergency event is selected to obtain a targeted solution set. The membership degree is calculated together with the expert rule set to obtain a membership degree normalization value. The normalization value is weighted to obtain the event priority. After the fuzzy set is defined according to the status information, the fuzzy set is matched with the membership degree of the indicators in the generated operating data to obtain the calculated priority data. The priority data is then used to update the event records. Based on the event urgency and event type, the calculated priority data is compared with the range of the priority data to obtain the priority ranking of the processing rules. This ranking result is input into the existing equipment expert rule base to guide the generation of operation and maintenance tasks.

[0083] Preferably, in S5.2, the fuzzy set is defined as follows: A fuzzy set corresponding to different state information or event types;

[0084] The membership function is: ,in This refers to status information or event-related data.

[0085] The running status membership data is set as follows: , where H is the number of status information or event types;

[0086] The expert experience rule set is as follows: , where h is the number of rules;

[0087] Event priority is set as follows: ,in These are the weighting coefficients;

[0088] The priority order of processing rules is set as follows: ,in This is a sorting function.

[0089] Preferably, the sub-step of S6 is as follows:

[0090] S6.1. Based on the content and priority information of the contingency plan group, the Support Vector Machine (SVM) algorithm is used to generate the initial instruction set for each disposal group. The SVM algorithm, based on the mapping relationship between priority parameters and contingency plan group features, learns and trains on a large amount of historical disposal data to find the optimal classification hyperplane, thereby generating the initial instruction set corresponding to the current priority. S6.2. The initial instruction set of the disposal group generated in S6.1 is obtained and used as input to the process execution module to obtain an output result set, which is then sent to the information flow module. S6.3. Through the data information flow of the information flow module, a corresponding data relationship topology graph G'=(V, E) is constructed for the result set; where V is the set of nodes, representing each data result in the result set, and E is the set of edges, representing the logical or influence relationships between results; e is used as the basis for the topology graph. i，j Represents node v i and v j The edge between them; S6.4, obtain the correlation R between the two results based on the feedback value analysis. i，j =F(v i v j The function F is a function of the correlation between two data points, and R is a function of the correlation between them. i，j The calculation will be determined in conjunction with the feedback value; a correlation threshold δ will be set, if R i，j If the value is greater than δ, then add an edge e to the graph. i，j Otherwise, do not add;

[0091] S6.5. Obtain the data relationship topology graph, construct a prediction model using a graph neural network (GNN), and calculate the representation vector Z of each node in the graph. The GNN will train and update the model parameters θ based on the local information and feedback values ​​of each node in the topology graph. The node representation vector Zv is used to describe the node state, Z = (Zv1, Zv2; Zv3, Zv4, Zv5). |v| ); Graph neural networks here learn the relationships and feedback information between nodes in the topological graph, which can uncover the deep structural features inside the result set data, thereby providing more comprehensive and accurate node representation vectors for subsequent evaluation.

[0092] S6.6 Based on the representation vector Z of the nodes in the graph calculated in the previous step, the information is transmitted to the evaluation group, and the conditional decision model is generated by training with the collected data.

[0093] S6.7. The Random Forest algorithm is used to process the categorical variables in the data. The Random Forest algorithm constructs multiple decision trees, analyzes and processes the categorical variables in the node representation vectors, and combines the results of multiple decision trees to improve the accuracy and stability of the evaluation. S6.8. Path analysis in the decision trees is used to generate the result evaluation probability P'=DT(Zv), where DT... S6.9. A decision tree trained on training samples is used to predict the probability of an optimal outcome. When the probability of an unoptimized outcome is greater than 60%, the flexibility value in the original management plan is updated, and the updated information is synchronized to the execution entity in the first step. After adjusting the flexibility of the management plan, a new initial instruction set for the disposal group is generated. S6.10. Based on the outcome evaluation probability generated by the decision tree and the disposal group adjustment information, a pre-built dynamic flow prediction model is used to determine the new disposal result set prediction data for the disposal group. S6.11. This new disposal group result set is sent to the disposal result update module through a data interface. S6.12. Result verification and updates are performed according to the logic defined in this module. S6.13. Based on the verified results, the disposal result update module data is input to the pre-designed classification model to generate a priority update model. S6.14. The classification model is called to label the data by type, updating the priority of each plan in the management plan.

[0094] Preferably, the sub-step of S7 is as follows:

[0095] S7.1 Obtain evaluation result information based on the initial rule base;

[0096] S7.2 Obtain the rule base evaluation result information and calculate the new weight value for each rule based on its original weight;

[0097] S7.3 Update the original rule base according to the new weight values ​​of each rule to form an updated rule base;

[0098] S7.4 Utilize the update library to judge the multi-dimensional indicators of the DevOps platform system and obtain the new indicators corresponding to the original indicators;

[0099] S7.5 Calculate the adjustment value for each dimension's new and old indicators, and clarify the adjustment direction under the execution interval;

[0100] S7.6 If the indicator adjustment value exceeds the preset range, the difference comparison algorithm is used to obtain the corresponding change gradient;

[0101] S7.7 Determine the operating parameter values ​​of each dimension in the information processing mechanism by using the gradient of change in each dimension and the adjustment values ​​of each dimension index.

[0102] A DevOps platform monitoring cycle intelligent control system adopts the aforementioned DevOps platform monitoring cycle intelligent control method.

[0103] The present invention can achieve the following beneficial effects:

[0104] 1. This invention acquires multi-dimensional indicator data streams during system operation and employs complex data processing procedures, such as stream computing, window functions, clustering techniques, and principal component analysis, to accurately form multi-dimensional vectors representing the system state, thereby accurately determining the system status. Compared with traditional monitoring methods, this invention can more sensitively capture the comprehensive impact of system load, service response time, and various events on the system's operating status, detect potential anomalies in advance, significantly advance the warning time for system failures, and effectively reduce the risk of business interruption due to system failures. Based on unsupervised learning to calculate the distance between the real-time multi-dimensional vector and the known system state category centers, it can automatically identify the similarity between the current system state and various typical states, achieving intelligent classification and precise positioning of the system state, providing a reliable basis for subsequent targeted processing measures.

[0105] 2. This invention utilizes a specially constructed event level model, combined with event information from multi-dimensional indicator data, to quantify the urgency of events. This data-driven and model-based approach, compared to traditional empirical judgments or simple threshold settings, can more objectively and accurately assess the impact of events on system operation, avoiding misjudgments of event urgency due to human error or unreasonable rule settings. Based on the event urgency value, corresponding alarms and processing mechanisms can be triggered promptly and effectively, ensuring that maintenance personnel can obtain key information and take action immediately in the face of emergencies, greatly improving the system's response speed and handling capabilities in dealing with emergencies.

[0106] 3. Compared with traditional fixed scheduling strategies, this invention can better adapt to various changes in the system operation process, improve resource utilization, avoid over-allocation or under-allocation of resources, and ensure that the system can operate efficiently under different load conditions. Attached Figure Description

[0107] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0108] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0109] Preferred solutions include Figure 1 As shown, a method for intelligent control of the monitoring cycle of a DevOps platform includes the following steps:

[0110] S1. During the operation of the DevOps platform, acquire multi-dimensional indicator data streams of the system operation. The data streams include system load, service response time, and the specific types and quantities of events occurring within the collection period. Based on the multi-dimensional indicator data streams, form a multi-dimensional vector representing the system state to obtain an instantaneous multi-dimensional vector; specifically:

[0111] S1.1. Based on the system load data stream, a streaming computing engine continuously receives system indicator data. The collected load data is processed using a defined window function to obtain the average load within a defined time window. Let the system load data stream be... (in (representing a time series), window function is Then the average load (in The number of data points within the window. For the first in the window (a specific time point).

[0112] S1.2 The system load average and service latency information flow are integrated. The service latency information flow uses a sliding window method to calculate the latency within a fixed time range, obtaining statistical data. Let the service latency information flow be... The sliding window size is M, and the service latency statistics are as follows. (in For the first in the sliding window (a specific time point).

[0113] S1.3. Integrate the above statistical data on average system load and service latency to form a two-dimensional information list to represent the comprehensive information on system load and service latency. Let the integrated two-dimensional list be... .

[0114] S1.4 Obtain a two-dimensional information list of all event types and their counts within the window period. Let the set of event types be... (in (Total number of event types), corresponding to the number of events. Then, a two-dimensional list of event types and frequencies. .

[0115] S1.5. Associate the load and service latency information list with the event type and frequency list, count the number of events occurring within a set time window, and count the number of events of different categories within the set time period to obtain a third list of the frequency and type of each category, and merge them into a third list. Let the merged third list be... .

[0116] S1.6. Using feature combination techniques, the three lists output in the previous step are processed to integrate multi-dimensional data streams, transforming them into a data format that comprehensively reflects the multi-dimensional characteristics of the system. Let the feature combination function be... The combined data .

[0117] S1.7. The data after feature combination processing is treated as a set of high-dimensional feature vectors. Let the high-dimensional feature vectors be... (in (the dimension of the feature vector). For combined data The feature values ​​obtained through some kind of mapping or transformation.

[0118] For example, Euclidean distance:

[0119] ,

[0120] Where (x, y) are two data points. Each of its first Each feature value.

[0121] S1.8. Use time-series correlation clustering techniques to calculate time-series correlations in order to uncover potential patterns in the data over time. Let the time-series clustering algorithm be... The clustering results are Specific temporal clustering algorithms can be, for example, density-based temporal clustering algorithms. The calculation process involves complex operations such as distance measurement between data points and density calculation, which will not be elaborated on here.

[0122] S1.9. By constructing a time decay coefficient, the correlation with real-time is calculated. If the time decay coefficient of data at a certain moment is greater than a preset value, it is discarded to ensure the timeliness of the data. Let the time decay coefficient be... (in As the attenuation factor, (For reference time point), if ( If the preset time decay coefficient threshold is used, the corresponding data will be discarded.

[0123] S1.10. Obtain a list set containing time information. Let this set be:

[0124] (in For the number of data points, For the first The time for each data point (For the corresponding feature vector).

[0125] S1.11. Use Principal Component Analysis (PCA) to reduce the dimensionality of the obtained time list set, extract key information, and reduce data dimensionality. Let the PCA transformation matrix be Φ, and the dimensionality-reduced principal component vectors be... (in (The dimension after dimensionality reduction) The PCA algorithm specifically calculates the covariance matrix.

[0126] (in (The eigenvectors are the mean of the eigenvectors), then the eigenvalues ​​and eigenvectors of the covariance matrix are solved, and the eigenvectors corresponding to the larger eigenvalues ​​are selected to form the transformation matrix Φ.

[0127] S1.12. Extract the vector groups of the principal components in the set, and obtain the principal component with the largest proportion at each time step as the feature data to obtain the instantaneous multidimensional vector. Let the instantaneous multidimensional vector be:

[0128] (in (The final number of principal components selected). From The principal component values ​​are selected based on their proportions.

[0129] S1.13. Calculate the instantaneous multidimensional vector using an unsupervised method, calculate the Euclidean distance of the vector to the known system state category centers, and determine the minimum distance to all known vector center points. Let the set of known system state category centers be... (in (number of system state categories), vector Go to Category Center Euclidean distance:

[0130] (in Center of category The (Eigenvalues).

[0131] S1.14. Determine which system state is most similar; if this distance is less than the state distance threshold, then the calculated instantaneous vector is assigned to the most similar system state. Let the state distance threshold be... ,like ,but Category .

[0132] S1.15. An adaptive deep neural network is used to process the state similarity results from multiple time periods. Each system state is output to the deep network, and if the network has a new output, the parameters of a certain system state are updated. The trained deep neural network is then used as the decision basis for immediate state calculation. Let the deep neural network be NN, and the input be the state similarity results from multiple time periods. (in (Number of time periods), the output is the system state category prediction. The network training process uses the backpropagation algorithm, which minimizes the loss function (such as the cross-entropy loss function).

[0133] ,in For real labels, The network parameters are updated using the network predictions.

[0134] S2. The event information contained in the real-time multi-dimensional vector of the DevOps platform is quantified using the pre-constructed event grading model. Each event is matched with corresponding grading parameters, and these parameters are combined to form a grading vector. The urgency level of the event is determined based on the specific values ​​contained in the grading vector. Additional explanation: The event grading model is constructed based on historical event data, the scope of event impact, business criticality, and other factors. Each grading parameter is determined through analysis of a large amount of sample data combined with expert experience to ensure the accuracy and rationality of the event urgency assessment. Let the event grading model be... The event information is Then the event level parameter Let the level vector be (in (Number of event types), event urgency value. According to The elements in the array are obtained through some mapping relationship (such as weighted summation), for example: (in For the first (Weight of each event type).

[0135] S3. Using a pre-defined rule base, perform logical operations on the system status information values ​​in the DevOps platform's real-time multidimensional vector and the event urgency values ​​in the level vector. Based on the rule base, deduce the confidence vector of the corresponding system status information and the confidence vector of the corresponding event urgency information under that condition; specifically:

[0136] S3.1. Collect multi-dimensional vectors and level vectors generated during the operation of the DevOps platform system. The obtained information includes system status information values ​​and event urgency values, and construct an information database. Let the system status information values ​​be... The urgency level of the event is [value missing]. The information database is .

[0137] S3.2. Perform normalization processing on the system status information values ​​in the database. Based on the distribution of event urgency levels, divide the event urgency values ​​into multiple intervals to obtain the normalized database and the level value information for each interval. Let the normalization function be... The normalized system state information value is Let the urgency of the event be divided into intervals as follows: The corresponding level value is .

[0138] S3.3 Divide the information into two parts: a state information group and a level information group. Perform clustering on the state information group. For each cluster center point generated by the clustering, calculate the values ​​of other state information and its distance to obtain the clustering data.

[0139] Let the state information group clustering algorithm be: The clustering results are ,

[0140] Let the set of cluster centers be... Status information value To the cluster center Distance:

[0141] (in for The 1 eigenvalue, for The 1 eigenvalue, (The dimension of the feature vector).

[0142] S3.4. A preset threshold is set, and the distance between the system status information value in the database and the threshold is compared. If the value is higher than the threshold, the current value is determined to be in a high-risk state, the corresponding status information group is identified as high-risk, and an alarm command for the high-risk state is output. The alarm command is then obtained. Let the threshold be... ,like If the condition is deemed high-risk, an alarm command will be output. .

[0143] S3.5. Use alarm commands to query the numerical information of the level information group. According to the rules defined in the rule base, if the queried level information group information matches the threshold range of the high urgency level of the level numerical information, then generate an alarm signal for the event, output the event alarm signal data, and obtain the information data of the emergency event. Transitional explanation: After determining the high-risk state, a convolutional neural network classification model is used because it can learn the complex nonlinear relationship between the system state information value and the event urgency value. Through training with a large amount of sample data, the confidence matching relationship between the two can be determined more accurately, thus providing a more reliable basis for subsequent decision-making. Let the convolutional neural network classification model be CNN, and the input be the combination of the system state information value and the event urgency value under the high-risk state. The output is the confidence level matching relationship. A convolutional neural network structure contains convolutional layers (such as convolutional kernels). Perform convolution operation on the input data ,in (representing convolution operations), pooling layers (such as max pooling operations) ), fully connected layers, etc., and the parameters of each layer are adjusted through training to optimize model performance.

[0144] S3.6. Train a convolutional neural network classification model using the training set data. Set the network loss function and optimize the network to obtain a mature network. Input emergency event information data into the network, determine the matching relationship between the confidence level of the output emergency event information and the system state information, and generate a confidence level association list. Let the training set be... (in For training sample index, (For the true confidence level label), the network loss function can be the mean squared error loss function:

[0145] (in The number of training samples. For the first (The confidence level of the network output for each sample) After adjusting the network parameters using optimization methods such as gradient descent to minimize the loss function and obtaining a mature network, the input emergency event information data is processed to obtain a confidence level association list. .

[0146] S4. If the confidence vector of the DevOps platform's system status information indicates an unhealthy state exceeding a set number, a system status alarm command is issued to the system, and the fuzzy scheduling strategy mechanism is triggered based on the event urgency value; specifically:

[0147] S4.1 Collect various state parameters of the DevOps platform system during runtime, construct a multi-dimensional system state information set, and obtain the system state vector after data cleaning;

[0148] S=(s1, s2; s i ), where s i The value representing the i-th state information;

[0149] S4.2, Based on the numerical value s of each state information in the system state vector S. i The system compares the data with a pre-defined healthy range, counting the number of unhealthy states to obtain the number N. It then determines whether the number N is greater than a set threshold T. If N > T, an alarm message set is output and an alarm command A is generated. If N ≤ T, the system returns to step S4.1 to continuously collect system status information. Let the healthy range be... (in For the first (health range range of each status information), if If the status information is unhealthy, then the number of unhealthy statuses is counted:

[0150] (in For indicator functions, the following conditions must be met: Otherwise ),like Output alarm information set and alarm commands .

[0151] S4.3 Obtain the urgency values ​​of multiple events corresponding to the alert information set of the DevOps platform system, normalize the urgency values, and obtain the normalized urgency value E=(e1, e2; e i ), where e i Let represent the urgency value of the i-th event. Let the normalization function be... ,but .

[0152] S4.4. Based on the normalized urgency value E, a fuzzy logic-based algorithm is used to determine the triggering scheduling level L. The core rule of the fuzzy logic algorithm is to determine the final scheduling level by considering the correspondence between different urgency value ranges and pre-set scheduling levels, while also taking into account the combined influence of the urgency levels of multiple events, through fuzzy reasoning. The corresponding scheduling level is then obtained. Assume the fuzzy logic system has an input fuzzy set... (Fuzzy sets corresponding to different urgency ranges), output fuzzy sets (Fuzzy sets corresponding to different scheduling levels), Fuzzy inference rule table (This represents the correspondence between different input fuzzy sets and output fuzzy sets), through fuzzification of the input... If a triangular membership function is used:

[0153] ,

[0154] in For fuzzy sets The parameters are then used to perform inference calculations based on the inference rule table, and finally the fuzzification is achieved.

[0155] For example, the center of gravity method:

[0156] ,

[0157] in To output fuzzy sets The representative value is used to obtain the scheduling level. .

[0158] S4.5 After determining the scheduling level L, select multiple scheduling policies that meet this level from the pre-established scheduling policy library. After policy fusion, obtain the fused scheduling policy G. Execute the fused scheduling policy G through the DevOps platform scheduling engine, and send resource allocation information data streams through the communication module to determine the resource reallocation scheme. Let the scheduling policy library be... (in (Number of strategies), selection and scheduling level Matching strategy subset Integrated scheduling strategy (in (For strategy fusion functions, such as weighted summation or rule-based combination, etc.), the resource allocation information data stream is set as follows: According to the fusion scheduling strategy Sure The specific details, such as the amount of resources allocated and the recipients, are used to achieve resource reallocation.

[0159] S5. Obtain the pre-configured expert experience processing rule sets corresponding to different types of events. Use fuzzy logic judgment technology combined with the expert experience processing rule sets to comprehensively calculate the real-time operation indicators and trigger events of the DevOps platform system, and determine the priority parameters after comprehensive calculation of the system status and urgency of events. Specifically: Build the expert knowledge base logic diagram. Explanation: First, collect historical operation and maintenance data and expert processing experience of the DevOps platform. Classify, organize and analyze this information to build an event processing rule set and a corresponding historical resource status data set. Then, upon event triggering, the DevOps platform system's real-time operational status information is synchronously acquired from multiple monitoring devices. This operational status information is matched against the trigger event type and urgency level, and the status information is assigned to a pre-defined fuzzy set interval. For each event type, the fuzzy set and membership function parameters are queried, and membership degree calculations are performed on the assigned status information to obtain several operational status membership degree data. Based on expert experience rules already set in the expert knowledge base, a set of rules corresponding to different states in the same type of emergency event records is selected to obtain a targeted solution set. The membership degree is jointly calculated with the expert rule set to obtain a membership degree normalization value. The normalization value is then weighted to determine the event priority. After defining the fuzzy set based on the status information, the fuzzy set is matched against the membership degree of indicators in the generated operational data to obtain the calculated priority data. This priority data is then used to update the event records. Based on the event urgency and event type, the calculated priority data is compared within a range to obtain a priority ranking of the processing rules. This ranking result is input into the existing equipment expert rule base to guide the generation of maintenance tasks. Let the fuzzy set be... (A fuzzy set corresponding to different state information or event types), the membership function is: (in (For status information or event-related data), the running status membership data is set to... (Where H represents the number of status information or event types), the expert experience rule set is: (where h is the number of rules), and the event priority is set to... (in (As weighting coefficients), processing rule priority sorting is set to (in (This is a sorting function).

[0160] S6. Determine several corresponding flexible management plans based on priority parameters, execute specific handling operation instruction sets according to the processes in the flexible management plans, and obtain the instruction handling results to the information processing stage for result evaluation; specifically:

[0161] S6.1. Based on the content and priority information of the contingency plan groups, the Support Vector Machine (SVM) algorithm is used to generate initial instruction sets for each response group. Here, the SVM algorithm, based on the mapping relationship between priority parameters and contingency plan group features, learns and trains from a large amount of historical response data to find the optimal classification hyperplane, thereby generating the initial instruction set corresponding to the current priority, providing the starting operation basis for subsequent response procedures. The specific method is as follows:

[0162] Let the support vector machine model be SVM, and the input be the priority parameters. and characteristics of the contingency plan group ,in Given the number of features, the output is the initial instruction set for the treatment group: ;

[0163] Support Vector Machines construct decision functions by:

[0164] ,in The number of training samples. For Lagrange multipliers, For training sample labels, For kernel functions, such as radial basis function kernels , For kernel parameters, The bias term is determined by solving an optimization problem. and To build the model.

[0165] S6.2. Obtain the initial instruction set of the processing group generated in the previous step, use it as input to the process execution module, obtain an output result set, and send the output result set to the information flow module; let the process execution module be... The output result set is .

[0166] S6.3. Through the data information flow of the information flow module, construct a corresponding data relationship topology diagram for the result set. ;in, It is a set of nodes, representing each data result in the result set. It is a set of edges, representing the logical or influence relationships between results; using Represents a node and The edge between;

[0167] S6.4 Analyze the correlation between the two results based on the feedback values. ,function This is a function representing the correlation between two sets of data. The calculation will be determined in conjunction with feedback values; a correlation threshold will be set. ,like Greater than Then add edges to the graph. Otherwise, do not add; set the correlation function. For some kind of similarity metric function, such as cosine similarity:

[0168] ;

[0169] in Represents the vector dot product. Represents the vector norm.

[0170] S6.5. Obtain the data relationship topology graph, construct a prediction model using a graph neural network (GNN), and calculate the representation vector Z of each node in the graph. The GNN will train and update the model parameters based on the local information and feedback values ​​of each node in the topology graph. ; Node representation vector Used to describe the node status Graph neural networks utilize message passing mechanisms, such as in each layer of nodes... Update formula:

[0171] ;

[0172] in For nodes The set of neighboring nodes, and For the first Layer weight matrix and bias vector, For activation functions, such as the ReLU function The node representation vector is obtained after multiple updates. .

[0173] S6.6. Based on the representation vector of the graph node calculated in the previous step. The information is transmitted to the evaluation group, which uses the collected data to train and generate a conditional decision model; let the evaluation group model be... The training data is ,in The decision result labels for the corresponding nodes are obtained through training using a certain machine learning algorithm. .

[0174] S6.7. The Random Forest algorithm is used to process the categorical variables in the data. The Random Forest algorithm constructs multiple decision trees, analyzes and processes the categorical variables in the node representation vectors, and combines the results of multiple decision trees to improve the accuracy and stability of the evaluation. The specific method is as follows:

[0175] Let random forest be composed of It consists of several decision trees. For the input node representation vector Each decision tree outputs a decision result. The final decision is obtained through a combination of voting or averaging.

[0176] S6.8. Using path analysis in the decision tree, generate the result evaluation probability P'=DT (Zv), where DT represents a decision tree trained on the training samples. The function is used to predict the probability of the result being evaluated as excellent. The decision tree is constructed by recursively partitioning the training data, such as selecting the best partitioning attribute based on indicators such as information gain or Gini index, constructing the tree structure, and then obtaining the evaluation probability based on the path of the input data in the tree.

[0177] S6.9 When the probability of the result being unfavorable is greater than 60%, update the flexibility value in the original management plan and synchronize the updated information to the executing entity in the first step. After adjusting the flexibility of the management plan, generate a new initial instruction set for the disposal group; let the flexibility value be... The update rule is as follows: ,in The new initial instruction set for the disposal team is based on the adjustment amount determined by the assessment results. Regenerated based on the updated flexibility value and other relevant information.

[0178] S6.10. Based on the outcome evaluation probability generated by the decision tree and the treatment group adjustment information, use a pre-built dynamic flow prediction model to determine the new treatment outcome set prediction data for the treatment group; let the dynamic flow prediction model be DFM, and the input be the outcome evaluation probability p' and the treatment group adjustment information. The output is a new set of predicted disposal results. Dynamic flow prediction models can employ time series prediction models or deep learning-based sequence models, such as the LSTM model, which predicts future outcomes by learning from historical disposal result sequences.

[0179] S6.11. Send this new set of disposal results to the disposal result update module through a data interface;

[0180] S6.12. Verify and update the results based on the logic already defined in this module;

[0181] S6.13. After verification, update the data in the disposal result module and input it into the pre-designed classification model to generate a priority update model; let the classification model be CM, and the input be the verification result data. The output is the priority update model. Classification models can employ methods such as support vector machine classifiers and neural network classifiers, which build models by learning from the classification of historical result data.

[0182] S6.14. By calling the classification model, the data is labeled with its type, and the priority of each contingency plan in the management plan is updated.

[0183] S7. Based on the rule base evaluation results, determine the dynamically adjusted rule weight allocation. Based on the updated rule base, re-evaluate the multi-dimensional indicators of the DevOps platform system to determine the direction and parameter values ​​for the next execution of the information processing mechanism; specifically:

[0184] S7.1 Obtain evaluation result information based on the initial rule base; let the evaluation result information be... .

[0185] S7.2 Obtain the rule base evaluation result information and calculate the new weight value for each rule based on its original weight; assuming the rule base is... ,in The number of rules is [number], and the original weight is [weight]. The new weight calculation function is Then the new weight ,For example It could be a weighted function that adjusts the importance of different rules based on the evaluation results, such as... ,in To adjust the coefficient, For the evaluation results information and rules The relevant parts.

[0186] S7.3. Update the original rule base according to the new weight values ​​of each rule to form an updated rule base; the updated rule base .

[0187] S7.4. Use the update library to judge the multi-dimensional indicators of the DevOps platform system and obtain the new indicators corresponding to the original indicators; let the system multi-dimensional indicators be... The judgment function is Then the new indicator ,For example It can be a reasoning or calculation function based on a rule base, which processes the indicator according to different rules and weights to obtain a new indicator value.

[0188] S7.5. For both new and old indicators across all dimensions, calculate the adjustment value for each indicator and clarify the adjustment direction within the execution interval; let the indicator adjustment value be... ,in ,according to The positive or negative sign determines the direction of adjustment.

[0189] S7.6 If the indicator adjustment value exceeds the preset range, the difference comparison algorithm is used to obtain the corresponding change gradient. The difference comparison algorithm calculates the difference between the indicator adjustment value and the preset range boundary value, and then combines this with the historical trend of indicator changes and data distribution characteristics to determine the corresponding change gradient, thereby quantifying the speed and direction of indicator changes. Let the preset range be... ,in For the first The preset range of each indicator, if Exceeding Gradient change ,in The time interval can be adjusted by methods such as moving average or statistical distribution fitting, taking into account historical trends and data distribution characteristics.

[0190] S7.7. Determine the operating parameter values ​​for each dimension of the information processing mechanism by using the gradient of change in each dimension and the adjustment values ​​of the indicators for each dimension; let the operating parameters of the information processing mechanism be... ,according to and Sure The adjustment amount, such as (in and (This is an adjustment factor), resulting in the adjusted operating parameter values. .

[0191] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for intelligent control of monitoring cycle in a DevOps platform, characterized in that... Includes the following steps: S1. During the operation of the DevOps platform, acquire multi-dimensional indicator data streams of system operation. The multi-dimensional indicator data streams include system load, service response time, and the specific types and quantities of each event that occurs within the collection period. Based on the multi-dimensional indicator data streams, form a multi-dimensional vector representing the system state to obtain an instant multi-dimensional vector. S2. Quantify the event information contained in the real-time multidimensional vector of the DevOps platform through the pre-built event level model, match the corresponding level parameters for each event, form a level vector from the level parameters of each event, and determine the urgency value of the corresponding event based on the specific values ​​contained in the level vector. S3. Using a preset rule base, perform logical operations on the system status information values ​​in the DevOps platform's real-time multidimensional vector and the event urgency values ​​in the level vector, and infer the confidence vector of the corresponding system status information and the confidence vector of the corresponding event urgency information under the given condition based on the rule base. S4. If more than a set number of system status information in the confidence vector of the DevOps platform system status information are in an unhealthy state, then issue a system status alarm command to the system and trigger the fuzzy scheduling strategy mechanism based on the numerical judgment of the urgency of the event. S5. Obtain the set of expert experience processing rules corresponding to different types of events that have been set up. Use fuzzy logic judgment technology in combination with the set of expert experience processing rules to comprehensively calculate the real-time indicators and trigger events of the DevOps platform system and determine the priority parameters after comprehensive calculation of the system status and urgency of the events. S6. Determine several corresponding flexible management plans through priority parameters, execute specific handling operation instruction sets according to the process in the flexible management plan, obtain the instruction handling results and feed them back to the information processing stage for result evaluation; S7. Determine the dynamically adjusted rule weight ratio based on the rule base evaluation results, and re-evaluate the multi-dimensional indicators of the DevOps platform system based on the updated rule base to determine the direction and parameter values ​​for the next execution of the information processing mechanism.

2. The intelligent control method for monitoring cycle of a DevOps platform according to claim 1, characterized in that: The sub-steps of S1 are: S1.1 Based on the system load data stream, a streaming computing engine is used to continuously receive system indicator data. The collected load data is processed by setting a window function to obtain the average load within a set time window. Assume the system load data flow is as follows: ,in Representing a time series, the window function is: , Then the average load ,in The number of data points within the window. For the first in the window A point in time; S1.2, Combine the average load and service latency information flow of the fusion system to calculate the latency within a fixed time range and obtain statistical data; let the service latency information flow be... The sliding window size is M, and the service latency statistics are as follows: ； in For the first in the sliding window A point in time; S1.

3. Integrate the statistical data of average load and service latency to form a two-dimensional information list to represent the comprehensive information of system load and service latency; let the integrated two-dimensional list be: ; S1.4 Obtain a two-dimensional information list of all event types and their counts within the window period; let the event type set be: ,in The total number of event types corresponds to the number of events. Then, a two-dimensional list of event types and frequencies. ; S1.

5. Associate the load and service latency information list and the event type and frequency list, count the number of events that occur within the set time window, and count the number of different types of events within the set time to obtain a third list of the frequency and type of each category, and merge them into a third list; Let the merged third list be: ; S1.

6. Use feature combination technology to integrate the three lists output in the previous step to transform them into a data form that can comprehensively reflect the multi-dimensional characteristics of the system. Let the characteristic combination function be The combined data is as follows: ; S1.

7. Treat the data after feature combination processing as a set of high-dimensional feature vectors; let the high-dimensional feature vectors be... ,in For the feature vector dimension, For combined data Feature values ​​obtained through mapping or transformation; S1.

8. Use time-series correlation clustering techniques to calculate time-series correlations in order to uncover potential patterns in the data over time; let the time-series clustering algorithm be... The clustering results are ; S1.

9. The correlation between the data and real-time is calculated by constructing a time decay coefficient. If the time decay coefficient of data at a certain moment is greater than a preset value, the data is discarded to ensure its timeliness. Let the time decay coefficient be: ,in As the attenuation factor, For reference time points, like , If the preset time decay coefficient threshold is used, the corresponding data will be discarded; S1.10 Obtain a list set containing time information; let this set be: ； in For the number of data points, For the first The time for each data point For the corresponding feature vector; S1.

11. Use PCA (Principal Component Analysis) to reduce the dimensionality of the obtained time list set, extract key information, and reduce data dimensionality. Let the PCA transformation matrix be Φ, and the dimensionality-reduced principal component vectors be: ； in The dimension after dimensionality reduction. The PCA algorithm specifically calculates the covariance matrix: ； in The mean of the eigenvectors is obtained, and then the eigenvalues ​​and eigenvectors of the covariance matrix are solved. The eigenvectors corresponding to the larger eigenvalues ​​are selected to form the transformation matrix Φ. S1.

12. Extract the vector group of the principal components in the set, and obtain the principal component with the largest proportion at each time step as the feature data to obtain the instantaneous multidimensional vector; let the instantaneous multidimensional vector be: ； in The final number of principal components selected. From The principal component values ​​are selected based on their proportions. S1.

13. Calculate the instantaneous multidimensional vector using an unsupervised method, calculate the Euclidean distance of the vector to the known system state category center, and determine the minimum distance to the center point of all known vectors; let the set of known system state category centers be: ； in The number of system state categories, a vector. Go to Category Center Euclidean distance: ； in Center of category The One eigenvalue; S1.

14. Determine which system state is most similar; if the distance between the instantaneous vector and the center of each known system state category is less than the state distance threshold, then the instantaneous vector is assigned to the most similar system state category. Let the state distance threshold be... ,like ,but Category ; S1.

15. The state similarity results of multiple time periods are processed by an adaptive deep neural network. Each system state is output to the deep network. If the network has a new output, the parameters of a certain system state are updated. The trained deep neural network is then used as the basis for decision-making in real-time state calculation.

3. The intelligent control method for monitoring cycle of a DevOps platform according to claim 1, characterized in that: The S2 method is as follows: Let the event level model be... The event information is Then the event level parameter Let the level vector be ,in The number of event types and the numerical value of event urgency. according to The elements in the array are obtained through mapping relationships.

4. The intelligent control method for monitoring cycle of a DevOps platform according to claim 1, characterized in that: The S3 sub-step is as follows: S3.1 Collect multi-dimensional vectors and level vectors generated during the operation of the DevOps platform system. The information obtained includes system status information values ​​and event urgency values, and build an information database. Let the system status information value be The urgency level of the event is [value missing]. The information database is ; S3.

2. Perform normalization processing on the system status information values ​​in the database. Based on the distribution of event urgency levels, divide the event urgency values ​​into multiple intervals to obtain the normalized database and the level value information of the divided intervals; let the normalization function be... The normalized system state information values ​​are: ， Let the urgency of the event be divided into intervals as follows: The corresponding level value is ; S3.

3. Divide the information into two parts: a state information group and a hierarchy information group. Perform clustering on the state information group. For each cluster center generated by the clustering, calculate the distance between other state information values ​​and the cluster center to obtain the clustering data. Let the state information group clustering algorithm be... The clustering results are as follows: , Let the set of cluster centers be... , Status information value To the cluster center Distance: ； in for The 1 eigenvalue, for The 1 eigenvalue, The dimension of the feature vector; S3.

4. Preset a threshold and compare the distance between the system status information value in the database and the threshold. If the value is higher than the threshold, determine that the current value is in a high-risk state, the corresponding status information group is in a high-risk state, and output an alarm command for the high-risk state to obtain the alarm command; let the threshold be... ,like If the condition is deemed high-risk, an alarm command will be output. ; S3.5 Use alarm commands to query the numerical information of the level information group. According to the rules defined in the rule base, if the queried level information group information matches the threshold range of the high urgency level of the level numerical information, generate an alarm signal for the event, output the event alarm signal data, and obtain the information data of the emergency event. S3.

6. Train a convolutional neural network classification model using the training set data, set the network loss function and optimize the network to obtain a mature network. Input emergency event information data into the network, determine the matching relationship between the confidence levels of the output emergency event information and the system state information, and generate a confidence level association list; Let the training set be: ； in For training sample index, To provide the true confidence level label, the network loss function uses the mean squared error loss function: ； in The number of training samples. For the first The network output confidence level for each sample is used to adjust the network parameters using the gradient descent algorithm to minimize the loss function, resulting in a mature network. Then, the input emergency event information data is processed to obtain a confidence level association list. .

5. The intelligent control method for monitoring cycle of a DevOps platform according to claim 2, characterized in that: The sub-steps of S4 are: S4.1 Collect various state parameters of the DevOps platform system during runtime, construct a multi-dimensional system state information set, and obtain the system state vector after data cleaning; S = (s1, s2; s i i), where s i represents the numerical value of the i-th state information; S4.2, the value s of each state information in the system state vector S i comparing with the preset health range, obtaining the number N of non-healthy state information by counting the number of non-healthy state information; judging whether the number N of non-healthy state is greater than the set threshold T, if N>T, outputting the alarm information set and generating the alarm instruction A; if N≤T, returning to step S4.1 to continuously collect system state information; the health range is: ； in For the first The health range interval of each status information, if If the status information is unhealthy, then the number of unhealthy states is counted: ; in For indicator functions, the following conditions must be met: Otherwise ,like Output alarm information set and alarm commands ; S4.3 Obtain the urgency values ​​of multiple events corresponding to the alert information set of the DevOps platform system, normalize the urgency values, and obtain the normalized urgency value E=(e1, e2; e i ), where e i Let the urgency value of the i-th event be represented by the normalization function. ,but ; S4.

4. Based on the normalized urgency value E, the scheduling level L triggered is determined using a fuzzy logic-based algorithm. The core rule of the fuzzy logic algorithm is to determine the final scheduling level by considering the correspondence between different urgency value ranges and pre-set scheduling levels, while also taking into account the combined impact of the urgency of multiple events. Obtain the corresponding scheduling level; set up Fuzzy logic systems have input fuzzy sets: , corresponding to fuzzy sets of different urgency levels Output fuzzy set , corresponding to fuzzy sets of different scheduling levels; Fuzzy Inference Rule Table This represents the correspondence between different input fuzzy sets and output fuzzy sets, achieved by fuzzifying the input... ; S4.5 After determining the scheduling level L, select multiple scheduling strategies that meet the level from the pre-established scheduling strategy library, and obtain the fused scheduling strategy G through scheduling strategy fusion; execute the fused scheduling strategy G through the DevOps platform scheduling engine, and send the resource allocation information data stream through the communication module to determine the resource reallocation scheme. Let the scheduling strategy library be ,in For the number of strategies, select the scheduling level. A subset of matching strategies: , Fusion scheduling strategy: ,in For the strategy fusion function, the resource allocation information data stream is set as follows: According to the fusion scheduling strategy Sure The content involves allocating resources and assigning them to specific objects to achieve resource reallocation.

6. The intelligent control method for monitoring cycle of a DevOps platform according to claim 1, characterized in that: The sub-steps of S5 are: S5.1 Collect historical operation and maintenance data and expert handling experience of the DevOps platform, classify, organize and analyze the historical operation and maintenance data and expert handling experience, and construct a set of event handling rules and a corresponding set of historical resource status data. S5.2 When an event is triggered, the DevOps platform system's real-time operating status information is synchronously acquired from multiple monitoring devices. The operating status information is matched with the type and urgency of the triggered event, and the status information is assigned according to a preset fuzzy set interval. For each event type, the fuzzy set and membership function parameters are queried to calculate the membership degree of the assigned status information, resulting in several operating status membership degree data. Based on the expert experience rules set in the expert knowledge base, a set of rules corresponding to different states in the records of the same type of emergency event is selected to obtain a targeted solution set. The membership degree is calculated together with the expert rule set to obtain a membership degree normalization value. The normalization value is weighted to obtain the event priority. After the fuzzy set is defined according to the status information, the fuzzy set is matched with the membership degree of the indicators in the generated operating data to obtain the calculated priority data. The priority data is then used to update the event records. Based on the event urgency and event type, the calculated priority data is compared with the range of the priority data to obtain the priority ranking of the processing rules. This ranking result is input into the existing equipment expert rule base to guide the generation of operation and maintenance tasks.

7. The intelligent control method for monitoring cycle of a DevOps platform according to claim 6, characterized in that: In S5.2, let the fuzzy set be: A fuzzy set corresponding to different state information or event types; The membership function is: ,in This refers to status information or event-related data; The running status membership data is set as follows: , where H is the number of status information or event types; The expert experience rule set is as follows: , where h is the number of rules; Event priority is set as follows: ,in These are the weighting coefficients; The priority order of processing rules is set as follows: ,in This is a sorting function.

8. The intelligent control method for monitoring cycle of a DevOps platform according to claim 1, characterized in that: The sub-steps of S6 are: S6.1 Based on the content and priority information of the contingency plan group, the support vector machine algorithm is used to generate the initial instruction set for each disposal group. The support vector machine algorithm finds the optimal classification hyperplane by learning and training on a large amount of historical disposal data based on the mapping relationship between priority parameters and contingency plan group features, thereby generating the initial instruction set corresponding to the current priority. S6.2 Obtain the initial instruction set of the handling group generated in S6.1, use the initial instruction set of the handling group as the input of the process execution module, obtain an output result set, and send the output result set to the information flow module; S6.

3. Through the data information flow of the information flow module, construct a corresponding data relationship topology graph G'=(V, E) for the result set; where V is the set of nodes, representing each data result in the result set, and E is the set of edges, representing the logical or influence relationships between results; using e i，j Represents node v i and v j The edge between; S6.

4. Analyze the feedback values ​​to obtain the correlation R between the two results. i，j =F (v i v j The function F is a function of the correlation between two data points, and R is a function of the correlation between them. i，j The calculation will be determined in conjunction with the feedback value; a correlation threshold δ will be set, if R i，j If the value is greater than δ, then add an edge e to the graph. i，j Otherwise, do not add; S6.

5. Obtain the data relationship topology graph, construct a prediction model using a graph neural network (GNN), and calculate the representation vector Z of each node in the graph. The GNN will train and update the model parameters θ based on the local information and feedback values ​​of each node in the topology graph. The node representation vector Zv is used to describe the node state, Z = (Zv1, Zv2; Zv3, Zv4, Zv5). |v| ); Graph neural networks here learn the relationships and feedback information between nodes in the topological graph, which can uncover the deep structural features inside the result set data, thereby providing more comprehensive and accurate node representation vectors for subsequent evaluation. S6.6 Based on the representation vector Z of the nodes in the graph calculated in the previous step, the information is transmitted to the evaluation group, and the conditional decision model is generated by training with the collected data. S6.

7. The random forest algorithm is used to process the categorical variables in the data. The random forest algorithm constructs multiple decision trees, analyzes and processes the categorical variables in the node representation vectors, and combines the results of multiple decision trees to improve the accuracy and stability of the evaluation. S6.

8. Using path analysis in the decision tree, generate the result evaluation probability P'=DT (Zv), where DT represents a decision tree trained on the training samples, and the function is used to predict the probability of the result being evaluated as excellent. S6.9 When the probability of the result being unfavorable is greater than 60%, the flexibility value in the original management plan is updated, and the updated information is synchronized to the executing entity in the first step. After adjusting the flexibility of the management plan, a new initial instruction set for the disposal group is generated. S6.

10. Based on the results generated by the decision tree, evaluate the probability and treatment group adjustment information, and use the pre-built dynamic flow prediction model to determine the new treatment outcome set prediction data for the treatment group; S6.

11. Send this new set of disposal results to the disposal result update module through a data interface; S6.

12. Verify and update the results based on the logic already defined in this module; S6.13 After verification, update the data of the disposal result module and input it into the pre-designed classification model to generate the priority update model; S6.

14. By calling the classification model, the data is labeled with its type, and the priority of each contingency plan in the management plan is updated.

9. The intelligent control method for monitoring cycle of a DevOps platform according to claim 1, characterized in that: S7 sub-steps: S7.1 Obtain evaluation result information based on the initial rule base; S7.2 Obtain the rule base evaluation result information and calculate the new weight value for each rule based on its original weight; S7.3 Update the original rule base according to the new weight values ​​of each rule to form an updated rule base; S7.4 Utilize the update library to judge the multi-dimensional indicators of the DevOps platform system and obtain the new indicators corresponding to the original indicators; S7.5 Calculate the adjustment value for each dimension's new and old indicators, and clarify the adjustment direction under the execution interval; S7.6 If the indicator adjustment value exceeds the preset range, the difference comparison algorithm is used to obtain the corresponding change gradient; S7.7 Determine the operating parameter values ​​of each dimension in the information processing mechanism by using the gradient of change in each dimension and the adjustment values ​​of each dimension index.

10. A DevOps platform monitoring cycle intelligent control system, characterized in that: The method for intelligent control of the monitoring cycle of a DevOps platform, as described in any one of claims 1-9, is adopted.

Images