CN115150255B - Self-adaptive knowledge-graph-based automatic root cause positioning method for application faults - Google Patents
Self-adaptive knowledge-graph-based automatic root cause positioning method for application faults Download PDFInfo
- Publication number
- CN115150255B CN115150255B CN202210822528.1A CN202210822528A CN115150255B CN 115150255 B CN115150255 B CN 115150255B CN 202210822528 A CN202210822528 A CN 202210822528A CN 115150255 B CN115150255 B CN 115150255B
- Authority
- CN
- China
- Prior art keywords
- alarm
- root cause
- fault
- node
- graph
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 46
- 238000012549 training Methods 0.000 claims abstract description 37
- 230000001364 causal effect Effects 0.000 claims abstract description 21
- 238000005065 mining Methods 0.000 claims abstract description 15
- 230000008569 process Effects 0.000 claims abstract description 13
- 230000009467 reduction Effects 0.000 claims abstract description 10
- 238000013145 classification model Methods 0.000 claims description 10
- 238000012163 sequencing technique Methods 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 4
- 238000013139 quantization Methods 0.000 claims description 3
- 230000004807 localization Effects 0.000 claims description 2
- 238000007635 classification algorithm Methods 0.000 abstract description 8
- 230000000694 effects Effects 0.000 abstract description 3
- 238000012423 maintenance Methods 0.000 description 14
- 238000012544 monitoring process Methods 0.000 description 4
- 230000001960 triggered effect Effects 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000009960 carding Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000013515 script Methods 0.000 description 1
- 238000013024 troubleshooting Methods 0.000 description 1
Abstract
The invention discloses a self-adaptive knowledge graph-based automatic root cause positioning method for application faults, which comprises an offline training part and an online root cause positioning part, wherein the offline training part uses CMDB data and an application calling relationship to construct a software and hardware knowledge graph; mining an alarm knowledge graph using historical alarm or fault alarm data; the input of the on-line root cause positioning is alarm data of a fault period and two knowledge maps obtained through training in an off-line training process, namely a software and hardware knowledge map and an alarm knowledge map, and then a fault root cause node and a fault propagation path are obtained through a root cause positioning algorithm. The invention uses the alarm classification algorithm and the causal mining algorithm to automatically mine the causal relation between the abstract alarm classification and the abstract alarm classification from the historical alarm data to construct the alarm knowledge graph, thereby improving the effect and the interpretability of the noise reduction clustering and root cause positioning of the subsequent alarms.
Description
Technical Field
The invention relates to the field of application faults of knowledge graphs, in particular to a self-adaptive automatic root cause positioning method for application faults based on knowledge graphs.
Background
Alarms are data that engineers trigger in order to guarantee IT quality of service through a series of expert rules set in the various monitoring data (metrics, logs, etc.) gathered. When an alarm occurs, the operation and maintenance engineer checks each piece of alarm data to confirm whether or not the system has occurred and what failure has occurred. Due to the volume and complexity of an online system, faults are unavoidable, and due to complex dependency of system components and redundant monitoring among different monitoring systems, once a certain component of the system breaks down, a chain reaction can be triggered to cause a problem of a plurality of components of the system, and the same problem can trigger a plurality of monitoring systems to generate alarms, so that massive alarms can be triggered in a short time, which is called an alarm storm. In this case, it is difficult for the operation and maintenance personnel to efficiently perform the failure analysis and root cause localization.
The existing solution is firstly alarm clustering based on cross entropy, specifically, according to the scene and rule of the alarm, the alarm information is clustered by using the cross entropy, so that the alarm convergence is realized, and the efficiency of fault analysis of operation and maintenance personnel is improved by the mode. Secondly, auxiliary root cause positioning is performed by establishing expert rules.
The prior art has the following disadvantages:
1. convergence rules requiring manual de-carding and maintenance of alarms
2. Rules requiring manual removal of comb and maintenance root cause positioning
3. The determination of the manual rules depends on expert experience, may not be accurate per se, and may not be applicable after a system update.
4. When a new alert rule is introduced by the system, the rule needs to be manually updated to cope with the new alert rule.
5. The method based on alarm convergence can not directly give out the root cause of the fault, and the alarm convergence effect based on cross entropy is limited, so that the method can only reduce about 30% of alarms and can not effectively solve the problem of alarm storm.
6. Alarms associated with system faults cannot be aggregated together, i.e., failure diagnosis cannot be well aided.
7. The versatility is low and different systems may require new manual rules to be reconfigured, configuration costs and efficiency are low.
8. Low interpretation
9. The exact root node and fault propagation path cannot be automatically given.
Based on the above reasons, an adaptive knowledge-based automatic root cause positioning method for application faults becomes a technical problem to be solved in the whole society.
Disclosure of Invention
The invention mainly solves the following problems:
(1) When a system fault occurs, an alarm storm is caused, and operation and maintenance personnel do not have an efficient means to automatically perform alarm noise reduction clustering, so that the number of alarms to be checked is excessive.
(2) When an alarm storm is generated, operation and maintenance personnel do not have an effective means to automatically locate root cause and analyze a fault propagation chain.
(3) The existing alarm noise reduction clustering and root cause positioning technology is basically based on expert rules, and has the disadvantages of difficult maintenance, high maintenance cost and weak mobility.
In order to solve the technical problems, the technical scheme provided by the invention is as follows: the self-adaptive knowledge graph-based automatic root cause positioning method for application faults comprises an offline training part and an online root cause positioning part, wherein the offline training part uses CMDB data and application calling relations to construct a software and hardware knowledge graph; mining an alarm knowledge graph using historical alarm or fault alarm data;
the input of the on-line root cause positioning is alarm data of a fault period and two knowledge maps obtained through training in an off-line training process, namely a software and hardware knowledge map and an alarm knowledge map, and then a root cause node and a fault propagation path are obtained through a root cause positioning algorithm.
Further, the input data of the offline training part in the mining of the alarm knowledge graph is historical alarm data or historical fault alarm data with a certain duration, and the historical alarm data or the historical fault alarm data are called training data.
Further, a root cause positioning process is performed for each failure set:
(1) Constructing an initial fault subgraph, namely acquiring graph nodes and edges associated with alarms in a fault set as the initial fault subgraph through the association relation between the alarms and software and hardware knowledge graph nodes; the initial fault subgraph is a subgraph of the software and hardware knowledge graph; mapping the alarms as node attributes on the corresponding fault subgraph nodes, and calculating the initial root scores and the weights of edges of the fault subgraphs through the alarm knowledge graph, the edge relations of the fault subgraphs and the time information of the alarms included in the alarm information on the nodes; specifically, if the alarm time of the alarm associated with a node is earlier in the fault set, the root cause score of the node is higher; two nodes with side relation are exemplified by A and B, the direction of the side is A to point to B, if the alarm type of the A node and the alarm type in the B node have causal relation on the alarm knowledge graph, the root score of the A node is improved, the root score of the B node is reduced, and the weight of the side is improved; finally, the initial root scores and weights of all the nodes are normalized to complete the construction of an initial fault subgraph;
(2) The root cause score reasoning is carried out, the initial fault subgraph is input, the final root cause score of each node of the fault subgraph is calculated by using a root cause score algorithm based on the graph, the fault subgraph can be obtained, and the node with the highest root cause score is taken as the root cause node; the root scoring algorithm based on the graph includes, but is not limited to, pageRank algorithm, personlized PageRank algorithm, etc., as long as the input is a fault subgraph, the output is the root score of each node of the fault subgraph, the so-called root score is the quantization index of the node as the root node, and the root scoring algorithm based on PageRank algorithm uses the pr value of the node output by PageRank algorithm as the root score of the node;
(3) The root cause link is mined, input is the fault subgraph and the root cause node, candidate links are firstly mined, the root cause node is taken as a starting node, and all links starting from the root cause node are mined as candidate root cause links by using a graph algorithm similar to a depth algorithm; and then, calculating the sequencing index of each candidate link, sequencing the candidate root cause links according to the sequencing index, and reserving the TopN as the root cause link which is finally output.
Further, the ranking index in the step (3) refers to the sum of root scores of nodes plus weights of edges divided by the number of edges and nodes on a link, and is used as the ranking index of the link.
Further, the alarm data firstly uses the software and hardware knowledge graph and the alarm knowledge graph obtained by the offline training part and the alarm classification model to perform alarm noise reduction clustering on the alarms.
Compared with the prior art, the invention has the advantages that:
(1) And an alarm knowledge map is constructed by automatically mining causal relations between abstract alarm classifications from historical alarm data by using an alarm classification algorithm and a causal mining algorithm, so that the effect and the interpretability of subsequent alarm noise reduction clustering and root cause positioning are improved.
(2) The CMDB data and application calling relation are used in the alarm noise reduction clustering and root cause positioning process, and the alarm noise reduction clustering and root cause positioning result is strong in effectiveness and interpretability.
(3) The offline training process and the online root cause positioning process only depend on given data (alarm, CMDB data and application calling relation), do not depend on expert knowledge, and have low maintenance cost and strong mobility.
(4) The on-line root cause positioning process can automatically infer the root cause of the fault and the fault propagation link only by the alarm data of the fault period.
(5) The time for troubleshooting and repairing operation and maintenance personnel can be greatly reduced, and the reliability of products is improved.
Drawings
Fig. 1 is a schematic diagram of an adaptive knowledge-based automatic root cause positioning method for application faults.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings.
The present invention will be described in detail with reference to the accompanying drawings.
The invention provides a self-adaptive knowledge-graph-based automatic root cause positioning method for application faults in concrete implementation,
The method comprises an offline training part and an online root cause positioning part, wherein the offline training part uses CMDB data and an application calling relationship to construct a software and hardware knowledge graph; mining an alarm knowledge graph using historical alarm or fault alarm data;
the input of the on-line root cause positioning is alarm data of a fault period and two knowledge maps obtained through training in an off-line training process, namely a software and hardware knowledge map and an alarm knowledge map, and then a root cause node and a fault propagation path are obtained through a root cause positioning algorithm.
The specific implementation principle process is as follows:
Off-line training (atlas mining) and on-line root cause positioning. An offline training part, which uses CMDB data and application calling relations to construct a software and hardware knowledge graph; the historical alarm or fault alarm data is used to mine the alarm knowledge graph. The input of the on-line root cause positioning is alarm data of a fault period and two knowledge maps obtained through training in an off-line training process, namely a software and hardware knowledge map and an alarm knowledge map, and then a fault root cause node and a fault propagation path are obtained through a root cause positioning algorithm.
In the offline training part, the software and hardware knowledge graph is constructed, and the input of the block is CMDB data and application calling relations (possibly stored in the CMDB), wherein the CMDB data refers to resource deployment relations of an application system and comprises node-node side relations. The node types include systems, modules, components, DUs (deployment units), groups (host instance groups), software, virtual machines, physical machines, access switches, core switches, aggregation switches, routers, and the like; the edge relationship includes constitute (constitution), call, logical connection, cluster, bearer, host, connection, etc. The application call relationship refers to a call relationship between application class nodes (e.g. systems, modules, components, micro-services, etc.) in the CMDB. And then inserting the edge relations (application calling relations are also variable relations) between the nodes into a graph database by using scripts to complete the construction of the software and hardware knowledge graph.
And the offline training part is used for mining the alarm knowledge graph, the input data of the block is historical alarm data (or historical fault alarm data) with a certain time length, and the subsequent input data is called training data. Firstly, training an alarm classification algorithm by using training data to obtain an alarm classification model, wherein the alarm classification model is used for mapping alarm instances to abstract alarm types (for example, all CPU alarm types are mapped to abstract alarm types of CPU alarm), the abstract alarm types should meet two conditions, and an operation and maintenance engineer can know the alarm meaning of the alarm types through the description of the abstract alarm types; secondly, the number is quite a few, namely the abstract alarm types are not too much, otherwise the abstract classification function is lost. The specific alarm classification algorithm is not limited, and the above purposes can be achieved, for example, a rule-based alarm classification algorithm can be used, namely if alarm data comprises alarm rules configured by an operation and maintenance engineer, the alarm classification algorithm can be realized by directly taking the alarm rules as abstract alarm types; for example, an unsupervised clustering-based alarm classification algorithm, namely clustering by using a text clustering algorithm of the message content of the alarm, and using a class cluster center as abstract alarm classification; for example, a supervised alarm classification algorithm, i.e., an operation and maintenance engineer first performs alarm classification labeling on training data, and then trains a supervised classification model (including but not limited to a neural network model). After training to obtain the alarm classification model, causal relationships between the abstract alarm classifications are mined using a causal discovery algorithm. First, a causal discovery sample M is constructed by using training data and an alarm classification model, wherein the causal discovery sample is a two-dimensional matrix, each column represents an alarm classification, each row represents a time window, and M [ i, j ] represents the number of alarm instances corresponding to the j-th abstract alarm type in the i-th time window. The alarm data is time sequence data, so the causal discovery sample can be directly constructed through training data and an alarm classification model. And then, using a causal discovery algorithm to mine causal relations among alarm types by using a causal mining algorithm, wherein the output causal relations are a plurality of edge triples, the data structures of the edge triples are (A, B, F), wherein A, B are abstract alarm types, F is the weight of an edge, the value is between 0 and 1, the strength of the edge (namely the strength of the causal relations) is represented, and the direction of the edge is that A points to B. The specific algorithm of the causal mining algorithm mentioned above is not limited, and the causal relationship mentioned above can be effectively mined, and representative algorithms include a PC algorithm, a PCMCI algorithm, a clustering algorithm, and the like, and integration of these algorithms. And then, all alarm classifications obtained by the alarm classification model and all causal relations obtained by causal mining are inserted into a graph database to obtain an alarm knowledge graph.
The online root cause positioning part can automatically trigger the root cause positioning process when the system fault occurs, and the specific automatic triggering mechanism is not in the scope of the discussion herein. The method is characterized in that the software and hardware knowledge graphs and alarm knowledge graphs obtained by an offline training part and an alarm classification model are used for carrying out alarm noise reduction clustering on alarms according to the principle that the weighted distances of the alarms in the time dimension, the topology dimension (calculated by the software and hardware knowledge graphs) and the description text dimension (calculated by the alarm knowledge graphs) are calculated (the smaller the distance is, the more likely that the two alarms belong to the same fault), then a clustering algorithm (s.g. DBSCAN algorithm) is used for carrying out noise reduction clustering on the alarms through the weighted distances so as to aggregate the alarms which belong to the same fault into the same class of clusters and remove noise alarms irrelevant to the fault, and one class of clusters is called as a fault set and consists of a plurality of alarms.
Then, a root cause positioning process is carried out on each fault set:
(1) And constructing an initial fault subgraph, namely acquiring graph nodes and edges associated with alarms in the fault set as the initial fault subgraph through the association relation between the alarms and the software and hardware knowledge graph nodes. The initial fault subgraph is a subgraph of the software and hardware knowledge graph. And then mapping the alarms as node attributes on the corresponding fault subgraph nodes. And then calculating the initial root score and the weight of the edge of the fault subgraph through the alarm knowledge graph, the edge relation of the fault subgraph and alarm information (including time information of alarm) on the nodes. Specifically, if the alarm time of the alarm associated with a node is earlier in the fault set, the root cause score of the node is higher; two nodes with edge relations exist (in an example of A and B, the direction of the edge is A to B), if the alarm type of the node A and the alarm type in the node B have causal relations on an alarm knowledge graph, the root score of the node A is improved, the root score of the node B is reduced, and the weight of the edge is improved. And finally, normalizing the initial root scores and weights of all the nodes to complete the construction of the initial fault subgraph.
(2) And (3) root cause score reasoning, namely inputting the initial fault subgraph, calculating the final root cause score of each node of the fault subgraph by using a root cause score algorithm based on the graph, and obtaining the fault subgraph, wherein the node with the highest root cause score is taken as the root cause node. The root scoring algorithm based on the graph includes, but is not limited to, pageRank algorithm, personlized PageRank algorithm, etc., as long as the input is a fault subgraph, the output is the root score of each node of the fault subgraph, and the root score is the quantization index of the node being the root node. The root cause scoring algorithm based on the PageRank algorithm takes the pr value of the node output by the PageRank algorithm as the root cause score of the node.
(3) Root cause link mining. The inputs are the above-described fault subgraph and root node, first the candidate links are mined, the root node is taken as the starting node, and all links starting from the root node are mined as candidate root links using a graph algorithm similar to the depth algorithm. And then, calculating the ranking index of each candidate link (the sum of the root cause score of the node on one link plus the weight of the edge divided by the number of the edge and the node is used as the ranking index of the link.) and ranking the candidate root cause links according to the ranking index, and reserving TopN as the root cause link which is finally output.
The output of the whole root cause positioning part is a fault subgraph, a root cause node and a root cause link.
The invention and its embodiments have been described above with no limitation, and the actual construction is not limited to the embodiments of the invention as shown in the drawings. In summary, if one of ordinary skill in the art is informed by this disclosure, a structural manner and an embodiment similar to the technical solution should not be creatively devised without departing from the gist of the present invention.
Claims (3)
1. An adaptive knowledge-graph-based automatic root cause positioning method for application faults is characterized by comprising the following steps of: the method comprises an offline training part and an online root cause positioning part, wherein the offline training part uses CMDB data and an application calling relationship to construct a software and hardware knowledge graph; mining an alarm knowledge graph by using historical alarm or fault alarm data, wherein the CMDB data refers to a resource deployment relationship of an application system;
the input of the on-line root cause positioning is alarm data of a fault period and a software and hardware knowledge graph and an alarm knowledge graph which are obtained through training in an off-line training process, and then a fault root cause node and a fault propagation path are obtained through a root cause positioning algorithm;
The method comprises the steps that firstly, a software and hardware knowledge graph, an alarm knowledge graph and an alarm classification model which are obtained by an offline training part are used for carrying out alarm noise reduction clustering on alarms, so that alarms which are caused by the same fault are clustered into the same class of cluster, noise alarms which are irrelevant to the fault are removed, and one class of cluster is called a fault set;
The following root cause localization procedure is performed for each failure set:
(1) Constructing an initial fault subgraph, and acquiring graph nodes and edges associated with alarms in a fault set as the initial fault subgraph through the association relation between the alarms and software and hardware knowledge graph nodes; the initial fault subgraph is a subgraph of the software and hardware knowledge graph; then mapping the alarm as node attribute on the corresponding fault sub-graph node, and calculating the initial root score and the weight of the edge of the fault sub-graph through the alarm knowledge graph, the edge relation of the fault sub-graph and the alarm information on the node, wherein the alarm information comprises the time information of the alarm; if the alarm time of the alarm associated with a node is earlier in the fault set, the root cause score for the node is higher; the method comprises the steps that two nodes A and B with edge relations exist, the direction of the edge is that A points to B, if the alarm type of the node A and the alarm type in the node B have causal relations on an alarm knowledge graph, the root cause score of the node A is improved, the root cause score of the node B is reduced, and the weight of the edge is improved; finally, the initial root scores and weights of all the nodes are normalized to complete the construction of an initial fault subgraph;
(2) The root cause score reasoning is carried out, the initial fault subgraph is input, the final root cause score of each node of the fault subgraph is calculated by using a root cause score algorithm based on the graph, the fault subgraph can be obtained, and the node with the highest root cause score is taken as the root cause node; the root scoring algorithm based on the graph comprises a PageRank algorithm and a Personlized PageRank algorithm, wherein the root scoring is a quantization index of a root node, and the root scoring algorithm based on the PageRank algorithm takes the pr value of the node output by the PageRank algorithm as the root scoring of the node;
(3) The root cause link is mined, input is the fault subgraph and the root cause node, candidate links are firstly mined, the root cause node is taken as a starting node, and all links starting from the root cause node are mined as candidate root cause links by using a graph algorithm; and then, calculating the sequencing index of each candidate link, sequencing the candidate root cause links according to the sequencing index, and reserving the TopN as the root cause link which is finally output.
2. The adaptive knowledge-based automatic root cause positioning method for application faults according to claim 1, which is characterized in that: the input data of the offline training part in the alarm knowledge graph mining is historical alarm data or historical fault alarm data with a certain duration, and the historical alarm data or the historical fault alarm data are called training data.
3. The adaptive knowledge-based automatic root cause positioning method for application faults according to claim 1, which is characterized in that: the ranking index in the step (3) refers to the sum of the root cause score of a node on a link plus the weight of an edge divided by the number of edges and nodes, and is used as the ranking index of the link.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210822528.1A CN115150255B (en) | 2022-07-12 | Self-adaptive knowledge-graph-based automatic root cause positioning method for application faults |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210822528.1A CN115150255B (en) | 2022-07-12 | Self-adaptive knowledge-graph-based automatic root cause positioning method for application faults |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115150255A CN115150255A (en) | 2022-10-04 |
| CN115150255B true CN115150255B (en) | 2024-05-31 |
Family
ID=
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113268370A (en) * | 2021-05-11 | 2021-08-17 | 西安交通大学 | Root cause alarm analysis method, system, equipment and storage medium |
| CN113377567A (en) * | 2021-06-28 | 2021-09-10 | 东南大学 | Distributed system fault root cause tracing method based on knowledge graph technology |
| CN114430365A (en) * | 2022-04-06 | 2022-05-03 | 北京宝兰德软件股份有限公司 | Fault root cause analysis method and device, electronic equipment and storage medium |
| CN114564580A (en) * | 2022-02-15 | 2022-05-31 | 北京云集智造科技有限公司 | Self-adaptive alarm aggregation method based on knowledge graph |
| CN114598539A (en) * | 2022-03-16 | 2022-06-07 | 京东科技信息技术有限公司 | Root cause positioning method and device, storage medium and electronic equipment |
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113268370A (en) * | 2021-05-11 | 2021-08-17 | 西安交通大学 | Root cause alarm analysis method, system, equipment and storage medium |
| CN113377567A (en) * | 2021-06-28 | 2021-09-10 | 东南大学 | Distributed system fault root cause tracing method based on knowledge graph technology |
| CN114564580A (en) * | 2022-02-15 | 2022-05-31 | 北京云集智造科技有限公司 | Self-adaptive alarm aggregation method based on knowledge graph |
| CN114598539A (en) * | 2022-03-16 | 2022-06-07 | 京东科技信息技术有限公司 | Root cause positioning method and device, storage medium and electronic equipment |
| CN114430365A (en) * | 2022-04-06 | 2022-05-03 | 北京宝兰德软件股份有限公司 | Fault root cause analysis method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111209131B (en) | Method and system for determining faults of heterogeneous system based on machine learning | |
| CN111047082B (en) | Early warning method and device of equipment, storage medium and electronic device | |
| KR102118670B1 (en) | System and method for management of ict infra | |
| WO2022083576A1 (en) | Analysis method and apparatus for operating data of network function virtualization device | |
| CN109343995A (en) | Intelligent O&M analysis system based on multi-source heterogeneous data fusion, machine learning and customer service robot | |
| CN106209405B (en) | Method for diagnosing faults and device | |
| CN101783749B (en) | Network fault positioning method and device | |
| CN104639368A (en) | Method and device for processing faults of communications network equipment | |
| CN110493025A (en) | It is a kind of based on the failure root of multilayer digraph because of the method and device of diagnosis | |
| CN113590451B (en) | Root cause positioning method, operation and maintenance server and storage medium | |
| CN112769605B (en) | Heterogeneous multi-cloud operation and maintenance management method and hybrid cloud platform | |
| CN110032463A (en) | A kind of system fault locating method and system based on Bayesian network | |
| CN112217674A (en) | Alarm root cause identification method based on causal network mining and graph attention network | |
| CN113987001A (en) | Rail transit signal system fault analysis method and device and electronic equipment | |
| CN112559237A (en) | Operation and maintenance system troubleshooting method and device, server and storage medium | |
| CN115514627A (en) | Fault root cause positioning method and device, electronic equipment and readable storage medium | |
| CN113484693B (en) | Transformer substation secondary circuit fault positioning method and system based on graph neural network | |
| CN115033450A (en) | Bayesian cluster monitoring early warning analysis method based on distribution | |
| Yao et al. | A fault diagnosis expert system based on aircraft parameters | |
| CN113395182A (en) | Intelligent network equipment management system and method with fault prediction | |
| CN115150255B (en) | Self-adaptive knowledge-graph-based automatic root cause positioning method for application faults | |
| JP2003032253A (en) | In-advance countermeasure on-line diagnosis in manageable network | |
| CN107454089A (en) | A kind of network safety situation diagnostic method based on multinode relevance | |
| CN115470025A (en) | Intelligent root cause analysis method, device, medium and equipment in distributed cloud scene | |
| CN116011813A (en) | Urban rail transit emergency monitoring method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |