Disaster recovery network element node knowledge graph construction method based on graph structure and text fusion

Abstract

The invention relates to the technical field of data processing, provides a disaster recovery network element node knowledge graph construction method based on graph structure and text fusion, solves the problem of complexity of disaster recovery network management and operation and maintenance in a modern cloud environment network, and provides a disaster recovery network element node knowledge graph construction method by collecting network topology information, network element performance indexes and state data from a plurality of network data sources. A graph structure and text data are formed through extraction and processing, then the graph structure and the text data are integrated into a fused graph structure, and disambiguation and redundancy removal processing are carried out. Then, the fusion graph structure is converted into a triple form, a complete knowledge graph is constructed, the knowledge graph comprises multi-dimensional information of network topology performance and network topology state data, and the multi-dimensional information is stored in a graph database and used for query, analysis and network operation and maintenance operation. The efficiency and accuracy of network operation and maintenance are improved, and the management cost and time are reduced.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method for constructing a knowledge graph of disaster recovery network element nodes based on the fusion of graph structure and text. Background Art

[0002] In modern cloud environment networks, as the scale of cloud platforms continues to expand, heterogeneity gradually increases, and the complexity of network management and operation and maintenance also increases significantly, the deployment and management of disaster recovery systems are crucial. Disaster recovery network element nodes refer to node devices that undertake functions such as data backup and business recovery in cloud environments. These devices can quickly restore data and services when cloud service failures or disasters occur, ensuring the normal operation of the business.

[0003] In the current cloud environment, there are many types of devices, different protocol standards, and growing data traffic, and the network structure and operation status are becoming more and more complex. Traditional disaster recovery network management methods often use decentralized tools to process the network topology, performance data, and status information separately. This independent management and display method is difficult to cope with the increasingly complex network environment, resulting in the network management system lacking effective data integration and correlation analysis capabilities when processing massive heterogeneous data.

[0004] Traditional disaster recovery network management usually stores and displays data such as network topology and network performance separately. The lack of data integration and display cannot fully reflect the overall operation of the network, making it difficult for network administrators to directly obtain a complete picture of the network status, reducing network management efficiency and increasing network management costs and time. Summary of the invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a method for constructing a knowledge graph of disaster recovery network element nodes based on the fusion of graph structure and text, so as to solve the problem that in the current cloud environment, with the rapid expansion of network scale and the continuous enhancement of network heterogeneity, the complexity of disaster recovery network management and operation and maintenance has increased. The existing technology is difficult to effectively integrate and associate different types of data, resulting in data fragmentation, which makes it difficult for network administrators to perform troubleshooting, performance optimization and overall network monitoring, and lacks global and comprehensive network status analysis, as well as intelligent analysis and rapid response capabilities.

[0006] In order to solve the above technical problems, the specific technical solutions of the present invention are as follows: The method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion includes: Step S101, collecting network topology information, network element performance indicators and network element status data from multiple network data sources and disaster recovery network element nodes, the network topology information is obtained through the network management system and the device management interface, the network element performance indicators include CPU utilization and memory utilization, the network element performance indicators are obtained through the NetFlow collector, the status data includes fault alarms and equipment operation status, and the status data is collected in real time through SNMP Trap raw data; Step S102, extracting and preliminarily processing the collected raw data to generate a graph structure and text data, the network topology information is processed into a graph structure, the graph structure includes a node set and an edge set, and the network element performance indicators and network element status data are processed into text data form; Step S103, integrating the graph structure and the text data to form a fused graph structure; Step S104, disambiguating the fused graph structure, obtaining the identifier of each node and relationship in the network, measuring the similarity of the nodes according to the cosine similarity, merging similar nodes, and removing duplicate nodes and edges in the graph structure to obtain the fused graph structure after disambiguation; Step S105, converting the fusion graph structure after the disambiguation process into a triple form, where the triple includes a router, a switch, a physical machine, and a virtual machine; Step S106, construct the triples into a complete knowledge graph to generate a knowledge graph, which includes multi-dimensional information of network topology performance and network topology status data. The constructed knowledge graph is stored in a graph database, and the knowledge graph is used for query, analysis and network operation and maintenance operations.

[0007] Furthermore, the method for constructing a disaster recovery network element node knowledge graph based on graph structure and text fusion according to the present invention, step S101 includes: Determine the data collection source, and obtain the network device connection relationship, network device type, and network topology information in the disaster recovery network through the network management system and device management interface based on the data collection source; Use NetFlow collectors to capture data on CPU utilization, memory utilization, and network element performance indicators from the network; The original data is collected in real time through SNMP Trap, which includes fault alarms of network devices, the operating status of network devices, and the status information of network devices.

[0008] Furthermore, the method for constructing a disaster recovery network element node knowledge graph based on graph structure and text fusion according to the present invention, step S102 includes: Classify and identify the raw data collected from multiple network data sources, and distinguish network topology information, network element performance indicators, and network element status data; Process the network topology information into a graph structure, which includes building node sets and edge sets. The node set is the network device, and the edge set is the connection relationship between devices. The network element performance indicators and network element status data are converted into text data format. The network element performance indicators include CPU utilization and memory utilization, and the network element status data includes fault alarms and equipment operation status.

[0009] Furthermore, the method for constructing a disaster recovery network element node knowledge graph based on graph structure and text fusion described in the present invention, step S103 includes: Aligning entities in the graph structure and text data so that the network device entities in the graph structure correspond to the network device entities in the text data based on device names and IP addresses; Associating the performance indicators and status information in the text data with the corresponding nodes in the graph structure, and configuring the attribute values ​​in the text data on the nodes in the graph structure; Based on the aligned data, a fused graph structure is constructed. The fused graph structure includes the connection relationship of network devices and the performance and status information of each device. The connection relationship of network devices is the edge in the graph structure, and the performance and status information of each device is the attribute value in the text data.

[0010] Furthermore, the method for constructing a disaster recovery network element node knowledge graph based on graph structure and text fusion described in the present invention, step S104 includes: Extract all nodes and relationships from the fusion graph structure. Nodes represent devices in the network, and relationships represent connections between devices. Devices in the network include routers and switches. Construct an attribute vector for each node, which includes the node's device type, performance indicators, status information, and connection information; The similarity between any two nodes is calculated according to the cosine similarity formula. The cosine similarity measures the similarity between two nodes by comparing the cosine value of the angle between the attribute vectors of the two nodes. For node pairs whose similarity exceeds a preset threshold, they are merged. During the merging process, one of the nodes is selected as the representative, and the attributes and relationships of the other node are transferred to the representative node. Traverse the graph structure, find and delete duplicate nodes and edges. Duplicate nodes refer to nodes with the same attribute vector, and duplicate edges refer to edges connecting the same node pairs. After merging similar nodes and removing duplicate nodes and edges, a fused graph structure is generated after disambiguation processing.

[0011] Furthermore, the method for constructing a disaster recovery network element node knowledge graph based on graph structure and text fusion according to the present invention, step S105 includes: Parse the fusion graph structure after disambiguation processing to obtain the specific meaning of each node and edge and the relationship between each node and edge; Obtain the requirements for building a knowledge graph, and determine the basic structure of triples based on the requirements for building a knowledge graph. The basic structure of triples includes entities, relationships, and the form of entities or the form of entities, attributes, and values; Entities extracted from the fusion graph structure and the attribute logical relationships between entities are used to obtain the association relationships between entities based on the relationships between nodes and edges in the fusion graph structure. The association relationships between entities include connection relationships and dependency relationships. The extracted entities, attributes and relations are combined according to the basic structure of triples to generate complete triples.

[0012] Furthermore, the method for constructing a disaster recovery network element node knowledge graph based on graph structure and text fusion described in the present invention, step S106 includes: Obtain the characteristics and business requirements of the disaster recovery network element nodes, and build the basic framework of the knowledge graph based on the characteristics and business requirements of the disaster recovery network element nodes. The basic framework of the knowledge graph includes the hierarchical structure and relationship types of the graph; Import the sorted triple data into the knowledge graph framework to form a preliminary knowledge graph, and optimize the preliminary knowledge graph. The optimization of the preliminary knowledge graph includes adjusting the relationship type, merging duplicate nodes, and optimizing the graph structure; The optimized knowledge graph is verified and stored in a dedicated graph database. The graph is stored in the graph database for query and analysis operations.

[0013] Beneficial effects of the present invention: The present invention achieves comprehensive integration of network topology, network element performance indicators and status data by introducing technical means of fusing graph structure with text data. This solves the problem of insufficient data separation and integration in traditional methods, allowing network administrators to obtain more comprehensive and accurate network status information. The construction of the fusion graph structure not only retains the connection relationship of network devices, but also incorporates the performance indicators and status information of each device. This deep fusion enables the system to perform more complex correlation analysis, which helps to quickly locate network problems and optimize network performance.

[0014] The intelligent operation and maintenance system based on the knowledge graph can use the topology and status data in the graph for intelligent reasoning, quickly locate abnormal devices and provide possible fault cause analysis. This significantly improves the efficiency of fault handling and reduces the time of network interruption. By real-time monitoring of network element performance indicators, the system can promptly identify performance bottlenecks and provide optimization suggestions. This helps administrators to implement precise policies and improve the overall performance and stability of the network.

[0015] The construction and application of knowledge graphs make network management more automatic and intelligent. The system can automatically collect, process and analyze network data, reducing the need for manual intervention, thereby reducing operation and maintenance costs and time. By providing intuitive graph views and query interfaces, the system enables network administrators to more conveniently view network status and perform management operations. This further simplifies the operation and maintenance process and improves work efficiency.

[0016] The present invention is not only applicable to disaster recovery management in cloud computing environments, but can also be extended to virtualized networks, hybrid clouds, and multi-cloud scenarios. This wide applicability enables the system to better meet network management requirements in different scenarios. By independently processing various types of data and performing standardized operations based on their characteristics, the present invention can flexibly process multi-source, heterogeneous disaster recovery network metadata. This makes the system more adaptable and flexible when facing complex network environments.

[0017] The implementation of the present invention lays a foundation for the intelligent development of network operation and maintenance. With the continuous advancement of technology and the expansion of application scenarios, the intelligent operation and maintenance system based on knowledge graph will play a more important role in the future. By providing a new technical solution to meet the challenges of modern disaster recovery management, the present invention helps to stimulate technological innovation and development vitality in related fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative labor.

[0019] Figure 1 A schematic flow chart of a method for constructing a knowledge graph of disaster recovery network element nodes based on the fusion of graph structure and text provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described in conjunction with the specific embodiments of the present invention and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. The technical solutions provided by the embodiments of the present invention are described in detail below in conjunction with the drawings.

[0021] In order to better understand the purpose of the present invention, the present invention is described in further detail below.

[0022] The method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion includes: Step S101, collecting network topology information, network element performance indicators and network element status data from multiple network data sources and disaster recovery network element nodes, the network topology information is obtained through the network management system and the device management interface, the network element performance indicators include CPU utilization and memory utilization, the network element performance indicators are obtained through the NetFlow collector, the status data includes fault alarms and equipment operation status, and the status data is collected in real time through SNMP Trap raw data; In step S101, the present invention adopts a comprehensive data collection strategy, aiming to comprehensively collect information related to disaster recovery network element nodes from multiple network data sources, including network topology information, network element performance indicators and network element status data.

[0023] Network topology information is mainly collected through the network management system (NMS) and device management interface. The information describes in detail the connection relationship of devices in the network, device types, and the overall structure of the network.

[0024] The collection of network element performance indicators uses NetFlow collectors to capture performance indicator data from the network. Mainly including CPU utilization and memory utilization, these data reflect the current load and performance status of network element devices.

[0025] Real-time collection of network element status data: The original status data is collected in real time through the SNMP Trap mechanism. Status data includes but is not limited to fault alarm information, equipment operation status, etc. This information is crucial for timely discovery and problem solving.

[0026] Determine the source of data collection. First, clarify the source of data collection and ensure that all network data sources are covered.

[0027] Utilize the network management system and device management interface to obtain network topology information through these interfaces, including device connection relationships, device types, etc.

[0028] Deploy NetFlow collectors. Deploy NetFlow collectors in the network to capture performance indicator data such as CPU utilization and memory utilization.

[0029] Configure SNMP Trap and the SNMP Trap mechanism to receive status information and alarms sent by network devices in real time.

[0030] Step S101 provides a solid foundation for the subsequent construction of the knowledge graph by comprehensively collecting network topology information, network element performance indicators and status data. The accuracy and completeness of the data directly affect the effectiveness of the knowledge graph, and thus affect the efficiency of network management and operation and maintenance. By comprehensively utilizing multiple data sources and collection methods, the present invention improves the comprehensiveness and real-time nature of the data, and provides strong support for the intelligent management of disaster recovery network element nodes.

[0031] Step S102, extracting and preliminarily processing the collected raw data to generate a graph structure and text data, the network topology information is processed into a graph structure, the graph structure includes a node set and an edge set, and the network element performance indicators and network element status data are processed into text data form; In step S102, the present invention extracts and preliminarily processes the raw data collected from multiple network data sources to prepare for the subsequent knowledge graph construction. This process mainly includes two aspects: one is to process the network topology information into a graph structure, and the other is to process the network element performance indicators and status data into text data form.

[0032] Processing of network topology information: Convert network topology information from raw data format into a graph structure. The graph structure consists of a node set (V) and an edge set (E). The node set (V) represents each device in the network, such as routers, switches, physical machines, etc. The edge set (E) represents the connection relationship between devices, that is, the link or channel in the network. After processing, the network topology information is clearly represented as a graph structure composed of nodes and edges, which is convenient for subsequent analysis and display.

[0033] Processing of network element performance indicators and status data: converting these data from raw data format to text data format.

[0034] Text data mainly includes network element performance indicators (such as CPU utilization, memory utilization) and status information (such as fault alarms, equipment operating status, etc.). During the conversion process, the data needs to be standardized to ensure data consistency and comparability.

[0035] Results: After processing, the network element performance indicators and status data were organized into structured text data, which is convenient for subsequent query and analysis.

[0036] Implementation details: Data extraction: Accurately extract network topology information, performance indicators and status data from raw data.

[0037] Data cleaning: Clean the extracted data to remove invalid, duplicate or erroneous data.

[0038] Data conversion: Convert the cleaned data into graph structure and text data forms according to the predetermined format and requirements.

[0039] Data storage: Store the processed data in appropriate data structures for subsequent use.

[0040] Step S102 provides high-quality data input for the subsequent knowledge graph construction by extracting and preliminarily processing the original data. It processes the network topology information into a graph structure, which helps to clearly display the structure and connection relationship of the network. It processes the network element performance indicators and status data into text data format, which facilitates subsequent query, analysis and display. This process is one of the steps to build an accurate and effective disaster recovery network element node knowledge graph.

[0041] Step S103, integrating the graph structure and the text data to form a fused graph structure; In step S103, the present invention integrates the graph structure and text data generated in step S102 to form a fused graph structure. This step is a link in constructing a disaster recovery network element node knowledge graph based on the fusion of graph structure and text, which realizes the deep integration of network topology information with network element performance indicators and status data, and provides a more comprehensive and accurate data foundation for subsequent data analysis and knowledge graph construction.

[0042] The specific integration process is as follows: Entity alignment: align the network device entities (nodes) in the graph structure with the network device entities in the text data. This is usually done based on unique identifiers such as the device name and IP address. Entity alignment ensures that the same device entity in the graph structure and text data can correspond, laying the foundation for subsequent data fusion.

[0043] Data association associates the performance indicators (such as CPU utilization, memory utilization) and status information (such as fault alarms, equipment operation status) in the text data to the corresponding nodes in the graph structure, which means that each network device node not only contains its connection relationship (edges), but also contains detailed information such as its performance indicators and status data.

[0044] The fusion graph structure is constructed based on the aligned data and association relationships. The fusion graph structure not only retains the graph structure representation of the network topology information, but also integrates the performance indicators and status data of the network elements. The nodes in the fusion graph structure not only represent network devices, but also contain the performance indicators and status information of the device, and the edges represent the connection relationship between devices.

[0045] Data consistency check,Before performing entity alignment and data association, it is necessary to ensure the data consistency,in the graph structure and text data to avoid data conflicts or,errors.

[0046] Data standardization: In order to facilitate data fusion, it may be necessary to standardize the performance indicators and status data in the text data to ensure that they can be seamlessly connected with the data in the graph structure.

[0047] Fusion algorithm selection,According to specific needs and data characteristics, select a suitable fusion algorithm to build the fusion graph structure.,The selection of the algorithm should take into account factors such as data accuracy,,completeness and efficiency.

[0048] Step S103 integrates the graph structure and text data to form a fusion graph structure containing network topology information, network element performance indicators and status data. This structure not only reflects the actual situation of the network more comprehensively, but also provides a richer data source for subsequent data analysis and knowledge graph construction. Through the fusion graph structure, network administrators can more intuitively understand the connection relationship, performance indicators and status information of network devices, so as to perform network management and operation and maintenance more efficiently.

[0049] Step S104, disambiguating the fused graph structure, obtaining the identifier of each node and relationship in the network, measuring the similarity of the nodes according to the cosine similarity, merging similar nodes, and removing duplicate nodes and edges in the graph structure to obtain the fused graph structure after disambiguation; In step S104, the present invention performs disambiguation processing on the fusion graph structure so that each node and relationship in the graph structure has a unique and clear identifier, and removes redundant information to improve the consistency and accuracy of the data. The following is a detailed explanation of this step: Extract all nodes and relationships from the fusion graph structure. Nodes represent devices in the network (such as routers, switches, etc.), and relationships represent connections between devices.

[0050] Construct an attribute vector for each node. The attribute vector contains multiple attributes of the node, such as device type, performance indicators (CPU utilization, memory utilization, etc.), status information (fault alarm, device operation status, etc.) and connection information.

[0051] The cosine similarity formula is used to calculate the similarity between any two nodes. Cosine similarity measures the similarity between nodes by comparing the cosine value of the angle between the attribute vectors of two nodes.

[0052] The cosine similarity formula is: cosine_similarity(A,B)=∥A∥∥B∥A⋅B, where A and B are the attribute vectors of node A and node B respectively.

[0053] For node pairs whose similarity exceeds a preset threshold (such as 0.8), they are considered similar and merged.

[0054] During the merging process, one of the nodes is selected as the representative node, and the attributes and relationships of the other node are transferred to the representative node.

[0055] Traverse the graph structure, find and delete duplicate nodes and edges. Duplicate nodes refer to nodes with the same attribute vector, and duplicate edges refer to edges connecting the same node pairs. After merging similar nodes and removing duplicate nodes and edges, a fused graph structure is generated after disambiguation processing. Each node and relationship in this structure has a unique and clear identifier, and there is no redundant information.

[0056] Attribute vector encoding: When constructing an attribute vector, non-numeric data needs to be encoded (such as using one-hot encoding to encode the device type) for calculation.

[0057] Threshold setting: The preset similarity threshold needs to be set according to the actual situation to ensure that too many errors are not introduced when merging similar nodes.

[0058] Data consistency check: After merging nodes and removing duplicate information, a data consistency check is required to ensure that the data in the graph structure is still accurate and complete.

[0059] Step S104 solves the problems of unclear node and relationship identification and data redundancy that may exist in the fusion graph structure through disambiguation processing. The fusion graph structure after disambiguation processing is more accurate, concise and consistent, providing a more reliable data foundation for subsequent data analysis and knowledge graph construction. At the same time, disambiguation processing also helps to improve the efficiency and accuracy of network management and reduce the cost and time of network operation and maintenance.

[0060] Step S105, converting the fusion graph structure after the disambiguation process into a triple form, where the triple includes a router, a switch, a physical machine, and a virtual machine; In step S105, the present invention further converts the fusion graph structure after disambiguation processing into a triple form. This step is an important part of building a knowledge graph, which simplifies complex graph structure data into a triple form that is easy to process and understand, and facilitates subsequent knowledge graph construction and query analysis.

[0061] Parse the fusion graph structure after disambiguation processing to clarify the specific meaning of each node and edge, as well as the relationship between them. Determine the basic structure of triples according to the construction requirements of the knowledge graph. Triples usually include entities, relationships, and entities (or entities, attributes, and values).

[0062] Entities (such as routers, switches, physical machines, virtual machines, etc.) and the attribute logical relationships between entities are extracted from the fusion graph structure. The association relationships between entities include connection relationships (such as routers connected to switches) and dependency relationships (such as virtual machines relying on physical machines to provide resources).

[0063] The extracted entities, attributes and relations are combined according to the basic structure of triples to generate complete triples.

[0064] For example, the following triples can be generated: (router, connected to,switch), (physical machine, running,virtual machine), (CPU utilization, is,75%), and so on.

[0065] Example of a triple: Router-related triples: (Router, hasIP, IP address) (Router, hasSubnetMask, subnet mask) (Router, RouterToSwitch, SwitchID) Switch-related triples: (Switch, hasIP, IP address) (Switch, SwitchToPhysicalMachine, PhysicalMachineID) (Switch, SwitchToSwitch, AnotherSwitchID) Physical machine related triples: (PhysicalMachine, hasIP, IP address) (PhysicalMachine, CPUUsage, Utilization) (PhysicalMachine, arrange, VirtucalMachineID) Virtual machine related triples: (VirtucalMachine, hasIP, IP address) (VirtucalMachine, CPUUsage, utilization) (VirtucalMachine, runsOn, PhysicalMachineID) During the conversion process, data integrity check is required to ensure that each triple contains complete information without omissions or errors.

[0066] Step S105 simplifies the data representation and improves the readability and operability of the data by converting the fusion graph structure into a triple form. As the basic building block of the knowledge graph, the triple facilitates the subsequent knowledge graph construction, query and analysis. At the same time, the triple form also helps to realize data sharing and exchange and promote interoperability between different systems.

[0067] Step S106, construct the triples into a complete knowledge graph to generate a knowledge graph, which includes multi-dimensional information of network topology performance and network topology status data. The constructed knowledge graph is stored in a graph database, and the knowledge graph is used for query, analysis and network operation and maintenance operations.

[0068] In step S106, the present invention further constructs the triple data generated in the previous step into a complete knowledge graph, and stores this knowledge graph in a dedicated graph database for subsequent query, analysis and network operation and maintenance operations. The following is a detailed description of this step: According to the characteristics of the disaster recovery network element nodes and business needs, the basic framework of the knowledge graph is built. This framework includes core elements such as the hierarchical structure and relationship types of the graph, providing guidance for subsequent data import and graph construction. The sorted triple data is imported into the knowledge graph framework to form a preliminary knowledge graph. At this stage, it is necessary to ensure the accuracy and completeness of the triple data to avoid the impact of data errors or omissions on the quality of the graph.

[0069] Optimize the initially constructed knowledge graph, including adjusting relationship types, merging duplicate nodes, optimizing graph structure, etc. These optimization measures are aimed at improving the accuracy and readability of the graph, making it more in line with actual needs and business scenarios.

[0070] Verify the optimized knowledge graph to ensure that it meets the expected functional and performance requirements. The verification process may include data consistency checking, query performance testing, graph visualization, and other aspects.

[0071] The verified knowledge graph is stored in a dedicated graph database, such as Neo4j. Graph databases can efficiently store and query graph structure data, providing strong support for subsequent query, analysis, and network operation and maintenance operations.

[0072] The knowledge graph contains multi-dimensional information of network topology performance and network topology status data, which can fully reflect the connection relationship, performance indicators and status information of network devices. The graph database supports efficient graph structure data query and analysis operations, and can quickly locate network devices and analyze network performance bottlenecks. The knowledge graph can be displayed through visualization tools to help network administrators intuitively understand the network structure and device status and improve operation and maintenance efficiency. The intelligent operation and maintenance system based on the knowledge graph can realize advanced functions such as fault prediction and performance optimization, and provide intelligent support for network operation and maintenance.

[0073] In the process of building a knowledge graph, it is necessary to ensure the quality and accuracy of the input data. Any data errors or omissions may have an adverse effect on the quality of the graph and subsequent analysis. As the network environment changes and new devices are added, the knowledge graph needs to be updated and maintained regularly. This includes operations such as adding new nodes and relationships and deleting outdated data. When storing and querying the knowledge graph, attention should be paid to data security and privacy protection. In particular, for sensitive network equipment information and performance indicator data, appropriate security measures need to be taken to protect them.

[0074] Through the above steps and measures, the present invention successfully constructs the triple data into a complete knowledge graph and stores it in the graph database, providing strong support for subsequent query, analysis and network operation and maintenance operations.

[0075] Specifically, the method for constructing a knowledge graph of a disaster recovery network element node based on graph structure and text fusion according to the present invention, step S101 includes: Determine the data collection source, and obtain the network device connection relationship, network device type, and network topology information in the disaster recovery network through the network management system and device management interface based on the data collection source; Use NetFlow collectors to capture data on CPU utilization, memory utilization, and network element performance indicators from the network; The original data is collected in real time through SNMP Trap, which includes fault alarms of network devices, the operating status of network devices, and the status information of network devices.

[0076] In step S101, the present invention describes in detail the process of collecting data related to the disaster recovery network element node from multiple network data sources, including network topology information, network element performance indicators and network element status data. The following is the specific content of this step: Network Management System (NMS) is used to obtain the connection relationship of network devices, network device types and network topology information in the disaster recovery network. NMS is a software system specially used to monitor, configure and manage network devices, which can provide a comprehensive network view and real-time data.

[0077] Device Management Interface (DMI), in addition to NMS, can also directly access network devices through the device management interface to obtain their detailed configuration information and status data. DMI is usually a set of APIs or command line interfaces that allow administrators to remotely manage devices.

[0078] The connection relationship of network devices, collecting the physical or logical connection relationship between various devices in the network, including the connection between routers, switches, physical machines, virtual machines, etc., is crucial for building a network topology diagram.

[0079] Network device type, identifying different types of devices in the network, such as routers, switches, firewalls, etc. Device type information helps to classify and identify data in subsequent processing.

[0080] Network topology information, combining the above information, constructs a complete network topology map. The topology map shows the connection relationship and layout between various devices in the network, which is the basis for understanding the network structure and conducting network management.

[0081] Using NetFlow collector: NetFlow is a network traffic monitoring protocol. By enabling the NetFlow function on a network device, network traffic data passing through the device can be captured. The present invention uses a NetFlow collector to capture data on network element performance indicators such as CPU utilization and memory utilization from the network. These data reflect the operating efficiency and performance status of the network device.

[0082] Through SNMP Trap: Simple Network Management Protocol (SNMP) is a standard protocol for network device management. SNMP Trap is a mechanism in the SNMP protocol that allows network devices to actively send notifications to management stations when specific events (such as fault alarms and status changes) occur. The present invention collects the original status data of network devices such as fault alarms and operating status in real time through SNMP Trap. These data are of great significance for timely discovery and handling of network problems.

[0083] In summary, step S101 comprehensively collects data related to network element nodes in the disaster recovery network through various channels and means, providing a solid foundation for the subsequent knowledge graph construction.

[0084] Specifically, the method for constructing a knowledge graph of a disaster recovery network element node based on graph structure and text fusion according to the present invention, step S102 includes: Classify and identify the raw data collected from multiple network data sources, and distinguish network topology information, network element performance indicators, and network element status data; Process the network topology information into a graph structure, which includes building node sets and edge sets. The node set is the network device, and the edge set is the connection relationship between devices. The network element performance indicators and network element status data are converted into text data format. The network element performance indicators include CPU utilization and memory utilization, and the network element status data includes fault alarms and equipment operation status.

[0085] In step S102, the present invention performs detailed classification, identification and processing on the raw data collected from multiple network data sources to provide structured data input for the subsequent knowledge graph construction. The specific content of this step is as follows: Network topology information distinguishes data describing the connection relationship of network devices and network structure from the original data. These data are the basis for building a network topology diagram. Network topology information includes the physical or logical connection of devices, device type, etc.

[0086] Network element performance indicators identify and extract data that reflects the operating efficiency and performance status of network equipment, such as CPU utilization, memory utilization, etc. The data is expressed in numerical form and is an important basis for evaluating equipment performance.

[0087] Network element status data, extracts data describing the current status of network devices and abnormal events, such as fault alarms, device operating status, etc. Status data helps to promptly discover and handle problems in the network.

[0088] Processing network topology information into a graph structure is one of the steps in building a knowledge graph. The graph structure consists of a node set and an edge set, where the node set represents the network device and the edge set represents the connection relationship between the devices. By building a graph structure, the connection status of network devices and the network layout can be intuitively displayed.

[0089] The node set contains the identification information of all devices in the network, such as device name, IP address, etc. The edge set records the connection relationship between devices, including the type of connection (such as physical connection, logical connection), bandwidth and other attributes, which together constitute the skeleton of the network topology.

[0090] The network element performance indicators and status data are converted into text data for integration with the graph structure. Performance indicator data (such as CPU utilization, memory utilization) and status data (such as fault alarms, equipment operation status) are converted into text data with a specific format, which will be used as the attribute information of nodes or edges in the graph structure.

[0091] During the conversion process, the data needs to be standardized to ensure consistency and comparability. For example, the performance indicator data is normalized to be in the same numerical range; the status data is encoded to have a unified representation.

[0092] In summary, step S102 converts the original data into a structured graph structure and text data form through processing methods such as data classification recognition and graph structure construction, providing clear and accurate data input for subsequent knowledge graph construction.

[0093] Specifically, the method for constructing a disaster recovery network element node knowledge graph based on graph structure and text fusion according to the present invention, step S103 includes: Aligning entities in the graph structure and text data so that the network device entities in the graph structure correspond to the network device entities in the text data based on device names and IP addresses; Associating the performance indicators and status information in the text data with the corresponding nodes in the graph structure, and configuring the attribute values ​​in the text data on the nodes in the graph structure; Based on the aligned data, a fused graph structure is constructed. The fused graph structure includes the connection relationship of network devices and the performance and status information of each device. The connection relationship of network devices is the edge in the graph structure, and the performance and status information of each device is the attribute value in the text data.

[0094] In step S103, the present invention introduces in detail how to integrate the graph structure and text data to form a fusion graph structure. This process is one of the steps to build a knowledge graph of disaster recovery network element nodes, which improves the close integration of network topology information with network element performance and status data, and provides comprehensive data support for subsequent analysis and operation and maintenance. The following is the specific content of this step: The purpose of entity alignment is to make the network device entities in the graph structure and text data correspond one to one to avoid information confusion and incorrect association. Alignment is mainly based on device names and IP addresses. In a network environment, each device usually has a unique device name and IP address, which can be used as a unique identifier for the device. The system finds and matches the corresponding network device entities by comparing the device names and IP addresses in the graph structure and text data.

[0095] On the basis of entity alignment, the system further associates the performance indicators (such as CPU utilization, memory utilization) and status information (such as fault alarms, equipment operation status) in the text data with the corresponding nodes in the graph structure. After the association is completed, these performance indicators and status information in the text data will be configured as the attribute values ​​of the nodes in the graph structure. In this way, each node not only represents the network device itself, but also contains the real-time performance and status information of the device.

[0096] The fusion graph structure is the result of the integration of graph structure and text data. It contains the connection relationship of network devices and the performance and status information of each device. In the fusion graph structure, the connection relationship of network devices is represented by the edges in the graph structure, while the performance and status information of each device is represented by the attribute values ​​configured on the nodes.

[0097] The construction of the fusion graph structure realizes the deep integration of network topology information and network element performance and status data. The fusion graph structure not only retains the connection relationship of network devices, but also provides detailed performance and status information of each device, providing comprehensive data support for subsequent fault detection, performance optimization and other operation and maintenance operations.

[0098] Specifically, the method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion according to the present invention, step S104 includes: Extract all nodes and relationships from the fusion graph structure. Nodes represent devices in the network, and relationships represent connections between devices. Devices in the network include routers and switches. Construct an attribute vector for each node, which includes the node's device type, performance indicators, status information, and connection information; The similarity between any two nodes is calculated according to the cosine similarity formula. The cosine similarity measures the similarity between two nodes by comparing the cosine value of the angle between the attribute vectors of the two nodes. For node pairs whose similarity exceeds a preset threshold, they are merged. During the merging process, one of the nodes is selected as the representative, and the attributes and relationships of the other node are transferred to the representative node. Traverse the graph structure, find and delete duplicate nodes and edges. Duplicate nodes refer to nodes with the same attribute vector, and duplicate edges refer to edges connecting the same node pairs. After merging similar nodes and removing duplicate nodes and edges, a fused graph structure is generated after disambiguation processing.

[0099] In step S104, the present invention elaborates on how to perform disambiguation processing on the fusion graph structure to improve the consistency and accuracy of the data, thereby providing a high-quality data foundation for the subsequent knowledge graph construction. The following is the specific content of this step: First, all nodes and relationships are extracted from the fusion graph structure. In this context, nodes represent devices in the network (such as routers, switches, etc.), while relationships represent connections between devices. This step is to fully understand the devices and their connections in the graph structure in preparation for subsequent processing.

[0100] Construct an attribute vector for each node. The attribute vector contains multiple attributes of the node, such as device type, performance indicators (such as CPU utilization, memory utilization), status information (such as fault alarms, device operating status), and connection information (such as connection status with other devices). The attribute vector can comprehensively describe the characteristics of the node and provide basic data for the subsequent calculation of the similarity between nodes.

[0101] Calculate node similarity and calculate the similarity between any two nodes based on the cosine similarity formula. Cosine similarity measures the similarity between two nodes by comparing the cosine value of the angle between the attribute vectors of the two nodes. By calculating the similarity, similar or duplicate nodes that may exist in the graph can be identified, providing a basis for subsequent deduplication and merging operations.

[0102] Merge similar nodes. For node pairs whose similarity exceeds the preset threshold, they are considered similar and need to be merged. During the merging process, one of the nodes is selected as the representative node, and all the attributes and relationships of the other node are transferred to the representative node. By merging similar nodes, redundant information in the graph can be eliminated, and the consistency and accuracy of the data can be improved.

[0103] Remove duplicate nodes and edges, traverse the entire graph structure, find and delete duplicate nodes and edges. Duplicate nodes refer to nodes with the same attribute vector, while duplicate edges refer to edges connecting the same node pairs. Removing duplicate nodes and edges can further simplify the graph structure, reduce data redundancy, and improve the efficiency of subsequent processing.

[0104] Generate a fusion graph structure after disambiguation. After similar nodes are merged and duplicate nodes and edges are removed in the above steps, a fusion graph structure after disambiguation is generated. Each node in the graph structure is unique, and the connection relationship between nodes is clear, providing a high-quality data foundation for the subsequent knowledge graph construction.

[0105] In summary, step S104 achieves disambiguation of the fusion graph structure and data consistency and accuracy by extracting nodes and relationships, constructing attribute vectors, calculating node similarity, merging similar nodes, and removing duplicate nodes and edges.

[0106] Specifically, the method for constructing a disaster recovery network element node knowledge graph based on graph structure and text fusion according to the present invention, step S105 includes: Parse the fusion graph structure after disambiguation processing to obtain the specific meaning of each node and edge and the relationship between each node and edge; Obtain the requirements for building a knowledge graph, and determine the basic structure of triples based on the requirements for building a knowledge graph. The basic structure of triples includes entities, relationships, and the form of entities or the form of entities, attributes, and values; Entities extracted from the fusion graph structure and the attribute logical relationships between entities are used to obtain the association relationships between entities based on the relationships between nodes and edges in the fusion graph structure. The association relationships between entities include connection relationships and dependency relationships. The extracted entities, attributes and relations are combined according to the basic structure of triples to generate complete triples.

[0107] In step S105, the present invention describes in detail how to convert the fusion graph structure after the disambiguation process into a triple form. The specific content of this step is as follows: Parse the fusion graph structure and parse the disambiguated fusion graph structure to gain a deeper understanding of the specific meaning of each node and edge, as well as the relationship between them. This step is to extract valuable information from the fusion graph structure and provide data support for subsequent triple construction.

[0108] Obtain the knowledge graph construction requirements and clarify the knowledge graph construction requirements, which usually includes determining the application scenarios of the graph, user query requirements, and system performance requirements. Understanding the knowledge graph construction requirements helps determine the basic structure and content of triples so that the constructed knowledge graph can meet the needs of practical applications.

[0109] According to the requirements of knowledge graph construction, the basic structure of triples is determined. The basic structure of triples usually includes entities, relationships, and entity forms (such as entity A-relationship-entity B), or entities, attributes, and value forms (such as entity-attribute-value). Clarifying the basic structure of triples provides a guiding framework for the subsequent extraction of entities, attributes, and relationships from the fusion graph structure.

[0110] Extract entities and attribute logical relationships. Extract entities and attribute logical relationships between entities from the fusion graph structure. Entities may represent devices in the network (such as routers, switches, etc.), while attribute logical relationships describe the connection relationship, performance differences, and state changes between these devices. The extracted entities and attribute logical relationships are the data for constructing triples, which will be directly reflected in the knowledge graph and form the basic content of the graph.

[0111] Based on the relationship between nodes and edges in the fusion graph structure, the associations between entities are further determined. These associations may include connection relationships (such as the physical connection between two devices), dependency relationships (such as the impact of performance changes of one device on another device), etc. Clarifying the associations between entities helps to accurately express the actual operation of network devices and their interaction relationships in the knowledge graph.

[0112] Generate triples, combine the extracted entities, attributes, and relationships according to the basic structure of triples, and generate complete triples. For example, you can generate triples in the form of "router A-connection-switch B" or "physical machine C-CPU utilization-80%".

[0113] The generated triples are the basic building blocks of the knowledge graph. Together, they will constitute the content framework of the knowledge graph and provide strong support for subsequent query, analysis, and network operation and maintenance.

[0114] To summarize, step S105 successfully converts the fusion graph structure into triple form by parsing the fusion graph structure, obtaining the knowledge graph construction requirements, determining the basic structure of triples, extracting the logical relationship between entities and attributes, determining the entity association relationship, and generating triples, laying a solid foundation for building a complete knowledge graph.

[0115] Specifically, the method for constructing a disaster recovery network element node knowledge graph based on graph structure and text fusion according to the present invention, step S106 includes: Obtain the characteristics and business requirements of the disaster recovery network element nodes, and build the basic framework of the knowledge graph based on the characteristics and business requirements of the disaster recovery network element nodes. The basic framework of the knowledge graph includes the hierarchical structure and relationship types of the graph; Import the sorted triple data into the knowledge graph framework to form a preliminary knowledge graph, and optimize the preliminary knowledge graph. The optimization of the preliminary knowledge graph includes adjusting the relationship type, merging duplicate nodes, and optimizing the graph structure; The optimized knowledge graph is verified and stored in a dedicated graph database. The graph is stored in the graph database for query and analysis operations.

[0116] In step S106, the present invention describes in detail how to construct the triple data into a complete knowledge graph, and optimize, verify and store it. The following is the specific content of this step: Build the basic framework of the knowledge graph and obtain the characteristics and business requirements of the disaster recovery network element nodes. These characteristics and requirements are important bases for building the knowledge graph, and they determine the content that the knowledge graph needs to cover. According to the characteristics and business requirements of the disaster recovery network element nodes, build the basic framework of the knowledge graph. This framework includes the hierarchical structure of the graph (such as device layer, network layer, business layer, etc.) and relationship types (such as connection relationship, dependency relationship, performance relationship, etc.).

[0117] Import triple data and import the sorted triple data into the knowledge graph framework. These triple data are the core content of the knowledge graph, which describes the properties, status and relationship between network devices. By importing triple data, the content of the knowledge graph is initially formed. At this point, the knowledge graph already has the basic structure and data support, but it may need further optimization and improvement.

[0118] Optimize the initially constructed knowledge graph. Optimization operations may include adjusting relationship types (such as correcting incorrect relationship definitions, adding new relationship types, etc.), merging duplicate nodes (such as merging multiple nodes representing the same device into one node), and optimizing the graph structure (such as adjusting the hierarchical relationship of nodes, simplifying complex paths, etc.). Through optimization operations, the accuracy and usability of the knowledge graph are improved. The optimized knowledge graph will be more in line with actual needs and can better support subsequent query and analysis operations.

[0119] Verify the optimized knowledge graph. The verification operation may include checking the logical errors in the graph, testing the query performance of the graph, and comparing the consistency of the graph with the actual network, etc., to improve the accuracy of the knowledge graph. Through the verification operation, possible problems in the graph can be discovered and corrected to ensure that the quality of the graph meets the requirements of practical applications.

[0120] The verified knowledge graph is stored in a dedicated graph database, which is a database system specifically used to store graph structure data. It can efficiently store and query a large number of nodes and relationships in the knowledge graph. By storing the knowledge graph, it can be used in subsequent query and analysis operations. At the same time, storing the graph in the graph database is also conducive to the long-term preservation and maintenance of the graph.

[0121] To summarize, step S106 successfully constructs the triple data into a complete knowledge graph by building the basic framework of the knowledge graph, importing triple data, optimizing the knowledge graph, verifying the knowledge graph, and storing the knowledge graph, and stores it in a dedicated graph database, providing strong support for subsequent network management and operation and maintenance.

[0122] The present invention provides a method for constructing a knowledge graph about disaster recovery network element nodes, aiming to achieve efficient integration and intelligent analysis of multi-source heterogeneous data in the disaster recovery network. By systematically collecting, processing, integrating and expressing network topology, network element performance data and status information, a comprehensive, accurate and semantically rich knowledge graph is constructed to provide strong support for network management, operation and maintenance, and optimization. The process is as follows: Figure 1 shown.

[0123] Data collection is to collect data related to disaster recovery network element nodes from multiple network data sources, including network topology information from network management systems and device management interfaces, obtain network element performance indicators such as CPU utilization and memory utilization through SNMP requests or NetFlow collectors, and collect network element status data reflecting the operating status and fault information of network devices in real time through Syslog logs and SNMP Traps. After data collection is completed, these raw data are passed to subsequent processing links for analysis.

[0124] After data collection, the system extracts and preliminarily processes the collected raw data to generate graph structure and text data. Network topology information is processed into graph structure ;in is a set of nodes, representing network devices; It is an edge set, which represents the connection relationship between nodes. The graph structure not only retains the physical or logical connection of the network, but also records the communication links and transmission paths between nodes, which enables the topology structure to be displayed intuitively and facilitates subsequent processing.

[0125] The performance and status data are processed into text data form T. The text data mainly records the performance indicators (such as CPU and memory usage) and status (such as fault alarms and device operating status) of the device. This information is represented in the form of logs or time series. The formula for generating the graph structure is: ,in , The processing of text data can be described as: ,in, Indicates the content of each text.

[0126] After generating the graph structure and text data, the present invention integrates the two together through data fusion to form a fused graph structure. The fusion graph structure retains the network topology information and combines the performance and status data. The formula for the fusion process is: ,in It is the set of nodes in the fusion graph, including topological nodes and their associated performance and status attributes; is the edge set, representing the connections in the network topology; It is a set of attributes of nodes and edges, recording the performance and status data of network elements extracted from text data. Through this fusion process, the physical structure, performance and status information of the network are uniformly expressed, laying the foundation for subsequent disambiguation, redundancy removal and knowledge extraction.

[0127] After the fusion graph structure is generated, it is disambiguated and redundancy is removed.

[0128] Disambiguation mainly deals with ambiguous nodes and relationships in the graph structure, ensuring that each node and relationship has a unique identification and definition in the network, using cosine similarity To measure the similarity of nodes, the calculation formula is , when the similarity exceeds the threshold (the threshold is set to 0.8), the two nodes are considered similar and the system will merge similar nodes. , Respectively represent nodes , The attribute vector includes all the attributes of the node (type of equipment, performance indicators, status, etc.). The specific calculation steps are as follows: For each network element node, you first need to determine the attribute set it contains. The attribute set is defined based on the data characteristics of the network element node, usually including device type, performance indicators, status information, connection information, etc.

[0129] Encode the attributes of each node and convert non-numeric data into numeric data to facilitate subsequent calculations. The encoding method is as follows: Device Type : Use one-hot encoding to convert the device type into a sparse vector. If there are n types of devices, the code has n bits. The code for the i-th device is that the i-th bit is 1 and the rest are 0. For example, there are three types of devices: routers, switches, and virtual machines. Their one-hot codes are [1, 0, 0], [0, 1, 0], and [0, 0, 1] respectively.

[0130] Performance Indicators : Directly use the performance value as the attribute value. For example, if the CPU utilization is 75% and the memory utilization is 60%, the encoding is [75, 60].

[0131] Status Information : Use 0 and 1 to indicate the status of the device. 0 indicates warning status and 1 indicates active status.

[0132] Connection Information :Using one-hot encoding, if there are n nodes, the encoding has n bits, and the i-th bit is 1, it means there is a connection with the i-th node. For example, if it is connected with node A and node B, the encoding is [1, 1, 0, 0, ...].

[0133] Combine the attribute vectors and concatenate the above-mentioned vectors to obtain the comprehensive attribute vector of node v .

[0134] Perform Min-Max normalization on the attribute vector. The normalized attribute vector is: , .

[0135] Redundancy removal is to remove duplicate nodes and edges in the graph structure, ensure the simplicity and accuracy of the data, and avoid information redundancy. The formula is: .

[0136] After the fusion graph structure is processed by disambiguation and redundancy removal, it is converted into triples. Each triple describes the connection relationship and attributes of network elements in the form of entity-relationship-entity or entity-attribute-value. The formula for generating triples is: .

[0137] When triples are used for knowledge representation, the left and right sides represent objects, and the middle is the data attribute or object attribute. Common triples are as follows.

[0138] The attributes and relationship triples of the router entity are represented as follows: (Router, hasIP, IP) (Router, hasSubnetMask, SubnetMask) (Router, hasRouterInterface, RouterPort) (Router, RouterToSwitch, Switch) (Router, RouterToPhysicalMachine, PhysicalMachine) (Router, RouterToRouter, Router) The attributes and relationship triples of the router port entity are represented as follows: (RouterPort, hasIP, IP) (RouterPort, hasMAC, MAC) (RouterPort, state, on / off) (RouterPort, bandwidth, xMbps) The attributes and relationship triples of the switch entity are represented as follows: (Switch, hasIP, IP) (Switch, hasSwitchInterface, SwitchPort) (Switch, SwitchToPhysicalMachine,PhysicalMachine) (Switch, SwitchToSwitch, Switch) The attributes and relationship triples of the switch port entity are represented as follows: (SwitchPort, hasMAC, MAC) (SwitchPort, state, on / off) (SwitchPort, bandwidth, xMbps) The attributes and relationship triples of the physical machine entity are represented as follows: (PhysicalMachine, hasIP, IP) (PhysicalMachine, hasMAC, MAC) (PhysicalMachine, CPUUsage, 10%) (PhysicalMachine, hasDisc, Disc) (PhysicalMachine, arrange, VirtucalMachine) The attributes and relationship triples of the virtual machine entity are represented as follows: (VirtucalMachine, hasIP, IP) (VirtucalMachine, hasMAC, MAC) (VirtucalMachine, CPUUsage, 10%) (VirtucalMachine, hasDisc, Disc) The attributes and relationship triples of the disk entity are represented as follows: (Disc, storageCapacity, xGB) (DIsc, useCapacity, xGB) After generating the triples, further operations such as completion are performed to ensure the completeness and accuracy of the triples. The completion formula is: .

[0139] Finally, the processed triples are constructed into a complete knowledge graph. The generated knowledge graph not only includes a detailed representation of the network topology, but also combines the multi-dimensional information of performance and status data to form a highly expressive and query-capable network knowledge graph. The constructed knowledge graph is stored in a dedicated graph database, such as Neo4j, for subsequent query, analysis, and network operation and maintenance operations.

[0140] The innovation of the present invention in data processing and knowledge graph construction is mainly reflected in processing the collected disaster recovery network metadata into graph structure and text data, so that the system can flexibly process multi-source, heterogeneous disaster recovery network metadata, thereby improving the data integration and expression capabilities, avoiding the limitation of directly forcibly unifying all data into triples, but by independently processing various types of data, more efficient extraction and standardization operations are performed according to their respective characteristics. In addition, by introducing the concept of fusion graph structure, heterogeneous network data is cleverly processed in a unified manner, realizing the deep integration of network topology, performance and status information in the cloud environment. This processing method not only improves the expressiveness and relevance of the data, but also ensures the accuracy and consistency of the data through technical means such as disambiguation and de-redundancy. Compared with the traditional method of managing network topology and performance data separately, the present invention provides a more efficient and flexible cloud disaster recovery network knowledge graph construction solution, which significantly improves the intelligence level of network management and operation and maintenance.

[0141] In a cloud environment, network administrators collect information about multiple disaster recovery network element nodes through the system of the present invention, including network topology, performance data of network element devices (such as CPU utilization, memory usage) and status information (such as fault logs, network traffic anomalies). These data are extracted from different data sources, and after preprocessing, they are stored in the system in the form of graph structures and text data.

[0142] In the process of building the knowledge graph, the topology data is first converted into a graph structure consisting of nodes (network element devices) and edges (connection relationships between devices); then the performance data and status information are converted into attribute values ​​and fused with the topology graph. Through disambiguation and redundancy removal operations, a consistent and accurate fusion graph structure is generated. Finally, this information is converted into triples and a knowledge graph is constructed and stored in the graph database.

[0143] Through this knowledge graph system, network administrators can view the connection relationships, performance indicators, and status information of all disaster recovery network element nodes in the cloud environment in real time. When an abnormality occurs in the network, the system can use the topology and status data contained in the knowledge graph for intelligent reasoning, quickly locate abnormal devices, and provide possible fault cause analysis. Compared with traditional decentralized management methods, this system significantly improves the efficiency of fault handling.

[0144] The technical solution of the present invention effectively solves the problems of data separation, insufficient integration and incomplete display in traditional disaster recovery network management by introducing a disaster recovery network element node knowledge graph construction method based on graph structure and text fusion. The specific solutions are as follows: The present invention comprehensively collects network topology information, network element performance indicators (such as CPU utilization, memory utilization) and network element status data (such as fault alarms, equipment operating status) related to disaster recovery network element nodes from multiple network data sources.

[0145] The collected raw data is extracted and preliminarily processed and converted into two forms: graph structure and text data. The network topology information is processed into a graph structure containing node sets and edge sets, while the network element performance indicators and status data are processed into structured text data.

[0146] Integrate the graph structure and text data to form a fused graph structure. This step ensures the organic combination of network topology information and network element performance and status data, avoiding the isolated existence of data. The fused graph structure not only includes the connection relationship of network devices, but also integrates the performance indicators and status information of each device, thereby achieving comprehensive data integration. Disambiguation is performed on the fused graph structure, measuring the similarity of nodes by calculating the cosine similarity between nodes, merging similar nodes, and removing duplicate nodes and edges.

[0147] The fusion graph structure after disambiguation is converted into triples, each of which describes the connection relationship and attributes of network elements in the form of entity-relationship-entity or entity-attribute-value. A complete knowledge graph is constructed based on triples, which contains multi-dimensional information of network topology, network element performance indicators and status data, and can fully reflect the overall operation of the network.

[0148] The constructed knowledge graph is stored in a dedicated graph database, such as Neo4j, for subsequent query, analysis, and network operation and maintenance. Network administrators can quickly obtain a full picture of the network status by querying the knowledge graph, including the connection relationship, performance indicators, and status information of network devices, thereby significantly improving the efficiency and accuracy of network management and reducing management costs and time.

[0149] To summarize, the technical solution of the present invention effectively solves the problems of data separation, insufficient integration and incomplete display in traditional disaster recovery network management through the steps of comprehensive data collection and processing, data fusion and graph structure generation, disambiguation and de-redundancy processing, triple conversion and knowledge graph construction, and knowledge graph storage and application, and provides network administrators with a comprehensive, accurate and efficient network management method.

Claims

1. A method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion, characterized in that: include: Step S101, collecting network topology information, network element performance indicators and network element status data from multiple network data sources and disaster recovery network element nodes, the network topology information is obtained through the network management system and the device management interface, the network element performance indicators include CPU utilization and memory utilization, the network element performance indicators are obtained through the NetFlow collector, the status data includes fault alarms and equipment operation status, and the status data is collected in real time through SNMP Trap raw data; Step S102, extracting and preliminarily processing the collected raw data to generate a graph structure and text data, the network topology information is processed into a graph structure, the graph structure includes a node set and an edge set, and the network element performance indicators and network element status data are processed into text data form; Step S103, integrating the graph structure and the text data to form a fused graph structure; Step S104, disambiguating the fused graph structure, obtaining the identifier of each node and relationship in the network, measuring the similarity of the nodes according to the cosine similarity, merging similar nodes, and removing duplicate nodes and edges in the graph structure to obtain the fused graph structure after disambiguation; Step S105, converting the fusion graph structure after the disambiguation process into a triple form, where the triple includes a router, a switch, a physical machine, and a virtual machine; Step S106, construct the triples into a complete knowledge graph to generate a knowledge graph, which includes multi-dimensional information of network topology performance and network topology status data. The constructed knowledge graph is stored in a graph database, and the knowledge graph is used for query, analysis and network operation and maintenance operations.

2. The method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion according to claim 1, characterized in that: The step S101 includes: Determine the data collection source, and obtain the network device connection relationship, network device type, and network topology information in the disaster recovery network through the network management system and device management interface based on the data collection source; Use NetFlow collectors to capture data on CPU utilization, memory utilization, and network element performance indicators from the network; The original data is collected in real time through SNMP Trap, which includes fault alarms of network devices, the operating status of network devices, and the status information of network devices.

3. The method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion according to claim 1, characterized in that: The step S102 includes: Classify and identify the raw data collected from multiple network data sources, and distinguish network topology information, network element performance indicators, and network element status data; Process the network topology information into a graph structure, which includes building node sets and edge sets. The node set is the network device, and the edge set is the connection relationship between devices. The network element performance indicators and network element status data are converted into text data format. The network element performance indicators include CPU utilization and memory utilization, and the network element status data includes fault alarms and equipment operation status.

4. The method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion according to claim 1, characterized in that: The step S103 includes: Aligning entities in the graph structure and text data so that the network device entities in the graph structure correspond to the network device entities in the text data based on device names and IP addresses; Associating the performance indicators and status information in the text data with the corresponding nodes in the graph structure, and configuring the attribute values ​​in the text data on the nodes in the graph structure; Based on the aligned data, a fused graph structure is constructed. The fused graph structure includes the connection relationship of network devices and the performance and status information of each device. The connection relationship of network devices is the edge in the graph structure, and the performance and status information of each device is the attribute value in the text data.

5. The method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion according to claim 1, characterized in that: The step S104 includes: Extract all nodes and relationships from the fusion graph structure. Nodes represent devices in the network, and relationships represent connections between devices. Devices in the network include routers and switches. Construct an attribute vector for each node, which includes the node's device type, performance indicators, status information, and connection information; The similarity between any two nodes is calculated according to the cosine similarity formula. The cosine similarity measures the similarity between two nodes by comparing the cosine value of the angle between the attribute vectors of the two nodes. For node pairs whose similarity exceeds a preset threshold, they are merged. During the merging process, one of the nodes is selected as the representative, and the attributes and relationships of the other node are transferred to the representative node. Traverse the graph structure, find and delete duplicate nodes and edges. Duplicate nodes refer to nodes with the same attribute vector, and duplicate edges refer to edges connecting the same node pairs. After merging similar nodes and removing duplicate nodes and edges, a fused graph structure is generated after disambiguation processing.

6. The method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion according to claim 1, characterized in that: The step S105 includes: Parse the fusion graph structure after disambiguation processing to obtain the specific meaning of each node and edge and the relationship between each node and edge; Obtain the requirements for building a knowledge graph, and determine the basic structure of triples based on the requirements for building a knowledge graph. The basic structure of triples includes entities, relationships, and the form of entities or entities, attributes, and values. Entities extracted from the fusion graph structure and the attribute logical relationships between entities are used to obtain the association relationships between entities based on the relationships between nodes and edges in the fusion graph structure. The association relationships between entities include connection relationships and dependency relationships. The extracted entities, attributes and relations are combined according to the basic structure of triples to generate complete triples.

7. The method for constructing a knowledge graph of disaster recovery network element nodes based on graph structure and text fusion according to claim 1, characterized in that: The step S106 includes: Obtain the characteristics and business requirements of the disaster recovery network element nodes, and build the basic framework of the knowledge graph based on the characteristics and business requirements of the disaster recovery network element nodes. The basic framework of the knowledge graph includes the hierarchical structure and relationship types of the graph; Import the sorted triple data into the knowledge graph framework to form a preliminary knowledge graph, and optimize the preliminary knowledge graph. The optimization of the preliminary knowledge graph includes adjusting the relationship type, merging duplicate nodes, and optimizing the graph structure; The optimized knowledge graph is verified and stored in a dedicated graph database. The graph is stored in the graph database for query and analysis operations.

Images