Community-detection-based reasoning method and device for IP network routing simulation

Abstract

Since IP network topology often has a clear hierarchical structure, using community detection algorithm in graph theory can divide the topology into several regions with high cohesion and low coupling, and selectively spread the routing update in the region to other regions, which can effectively avoid the calculation amount of full routing simulation, and realize fast and accurate IP network routing simulation reasoning. The application proposes an IP network routing simulation reasoning method based on community detection, which first divides the whole network into multiple cluster structures based on the community detection algorithm, and the shortest path between the nodes in the cluster does not pass through other clusters, and uses this division result as the basis for calculating the routing update. Then, the change operation of the network topology is accurately identified, and according to the network elements involved in the change point and the cluster information to which they belong, the network element node pairs that need to be recalculated are further arranged, and the SPT tree of the node pairs is recalculated and the routing update is performed. At the same time, an IP network routing simulation reasoning device based on community detection is proposed, which includes a topology management module, a network configuration module, a network segmentation module, a change analysis module and a routing calculation module, to realize the topology segmentation and routing simulation reasoning process of the IP network.

Description

Technical Field

[0001] The present invention relates to an IP network routing simulation reasoning method and device based on community detection, belonging to the technical field of communications. Background Art

[0002] When simulating IP networks, there are usually two mainstream technical methods: numerical simulation and analog simulation.

[0003] (1) Simulation: Run the actual network operating system through a virtual machine or container and build connections between simulated network elements based on the virtualization platform. In the simulation method, each virtual machine contains a complete operating system and network protocol stack, ensuring high consistency with the behavior of physical devices. However, this method is limited by the performance of the server and requires the simultaneous use of a large number of server computing resources to support the simulation of large-scale networks.

[0004] (2) Numerical simulation is mainly based on discrete event simulation technology, which constructs a mathematical model of the virtual network to simulate the physical network. The current mainstream numerical simulation software includes OPNET, OMNeT++ and NS-3. It can define the logical behavior of each network device through program code, such as the forwarding mechanism and queue management of routers and switches, and use message queues and event scheduling to simulate protocol operations such as message transmission, packet loss, and retransmission. This method can simulate network behavior and data analysis in a short time and at a low cost. However, due to the complexity of the network system, the calculation time is long when simulating large-scale networks, which makes it difficult to meet the real-time requirements of digital twin networks.

[0005] Network routing simulation is the core of network simulation and is also a necessary prerequisite for key applications such as simulation network planning analysis, fault analysis, and scheduling analysis. In current mainstream numerical simulation schemes, it is often implemented by writing protocol stacks for each route, which is computationally intensive and has low simulation efficiency. Since network routing protocols often use the Dijkstra algorithm for routing calculations, some schemes abstract this problem into the shortest path algorithm in graph theory for simulation (e.g., CN108418757), and use i-SPF and PRC methods to narrow the convergence range of routing (e.g., CN102916879), thereby effectively improving the computational efficiency of routing.

[0006] Community detection refers to the discovery of densely connected subnetworks in a graph structure and is one of the important techniques in graph analysis. Typical algorithms include the Girvan-Newman (GN) algorithm, which discovers communities in a graph by iteratively removing edges based on the graph's edge betweenness centrality (EBC) value. Summary of the Invention

[0007] When simulating the routing protocol of an IP network, Dijkstra algorithm is usually used for routing calculation, which has a large amount of calculation and leads to low routing calculation efficiency. Although some schemes simulate the shortest path algorithm in graph theory and use i-SPF and PRC methods to reduce the convergence range of routing, it is still difficult to meet the real-time requirements of the deduction calculation in a large-scale network, especially in a network environment that requires fast response, and there is still room for improvement in the calculation efficiency.

[0008] Although there are some methods for improving the efficiency of network routing simulation at present, there is no scheme for reducing the calculation range by dividing the topology into regions. Since the IP network topology often has a clear hierarchical structure, the community detection algorithm in graph theory can be used to divide the topology into several regions with high cohesion and low coupling. Selectively spreading the routing updates within the region to other regions can effectively avoid the calculation amount of full routing simulation and realize fast and accurate IP network routing simulation reasoning.

[0009] In order to achieve the above purpose, the technical scheme of the present application is as follows: a community detection-based IP network routing simulation reasoning method, the method comprising the following steps:

[0010] Step 1: according to the network simulation scene, an IP network simulation environment is built through a visual or programming interface, a basic model of network elements and links is extracted, simulation network elements are instantiated, links between network elements are added, and an initial network topology is formed,

[0011] Step 2: the simulation network elements are configured with IP simulation protocols through a configuration channel, including the configuration of interconnection interfaces between network elements, the configuration of ISIS protocol neighbor relationships, and the configuration of network routing prefix import. After the initial configuration is completed, the simulation task is started,

[0012] Step 3: the routing calculation module performs a full calculation on the whole network according to the initial network topology and network element configuration information through Dijkstra algorithm, obtains an SPT tree with each node as the root node, and constructs a node routing library based on the SPT tree,

[0013] Step 4: the simulation network topology and the SPT tree with each node as the root node are collected as the input of the GN algorithm, and the network topology structure is cut by the GN algorithm to obtain several network clusters. Each network cluster is a sub-topology of the entire simulation network topology. The final effect of the cutting is that the SPT tree of any node in each cluster to other nodes in the cluster does not exist a path passing through another cluster,

[0014] Step 5: based on the stable IP simulation network, a network topology structure change operation is performed, including adding or deleting links, adding or deleting nodes, and configuration changes of network elements,

[0015] Step 6: After the change is completed, the change analysis module accurately identifies the change operations of the network topology, including adding intra-cluster links, deleting intra-cluster links, adding inter-cluster links, deleting inter-cluster links, adding single-link nodes within a cluster, deleting single-link nodes within a cluster, adding multi-link nodes within a cluster, and deleting multi-link nodes within a cluster. Based on the network elements involved in the above change points and the network cluster information segmented by the GN algorithm, the network element node pairs that need to be recalculated are further arranged, and the routing calculation module recalculates the SPT tree of the node pairs and updates the routes.

[0016] Step 4 is as follows:

[0017] Topology such as Figure 5 It shows that the topology is divided into clusters using a community detection algorithm. Since network routing protocols often use shortest path algorithms such as Dijsktra for path calculation, community detection algorithms based on edge betweenness (the number of shortest paths passing through an edge in the network) are more suitable for regional division. For example, the Girvan-Newman algorithm (GN algorithm), the Newman fast algorithm (NF algorithm) and related improved algorithms measure the degree of community division based on modularity. The modularity calculation method is as follows:

[0018]

[0019] In the formula, e is a k×k matrix, e ij represents the ratio of the number of edges connecting vertices between community i and community j to the total number of edges, a i represents the ratio of the number of edges connected to the nodes in community i to the total number of edges, Tr(e) represents the sum of the ratios of the number of edges within the same community to the total number of edges, ||E 2 || represents the sum of the matrix elements. The modularity function Q measures the difference between the proportion of edges within a community and the proportion of edges within a community in the case of random connections. When Q is close to 1, the network has a better community structure.

[0020] However, in the IP network simulation partitioning scenario, in order to decompose the network structure into highly cohesive and low-coupling regional structures and support the simulation goals of local routing calculation and global diffusion, modularity evaluation cannot be used as the only evaluation parameter. At the same time, it is necessary to add the restriction that the shortest path between nodes in the region does not cross other regions, that is, the routing independence judgment. In summary, the steps for regional division are as follows: Figure 4 As shown, the details are as follows:

[0021] Step 4.1: Convert the network topology G = (V, E) into the adjacency matrix A for easy calculation.

[0022]

[0023] Step 4.2: Calculate the shortest paths between all nodes and obtain the edge betweenness of each edge in turn. The following is the edge betweenness adjacency matrix B of topology G.

[0024]

[0025] Step 4.3: Delete the edge with the largest betweenness. If a deleted edge is generated in step 5, exclude the deleted edge and determine the largest edge to be deleted. At the same time, set the value of the corresponding edge in the adjacency matrix A to zero.

[0026] Step 4.4: Determine whether there is a new independent cluster structure in the topology after deleting the edge, R n Represents the n independent regions divided out, and If no independent cluster structure is obtained, repeat steps 4.2 and 4.3.

[0027] Step 4.5: Determine whether the shortest path between nodes in each cluster after deleting the edge passes through a node outside the cluster. Use P(u,v) to represent the shortest path between nodes u and v. For all edges e∈P(u,v) (where e=(x,y) is an edge in the path P(u,v)), satisfy x,y∈R i , then the cluster R i Meet the routing independence judgment. If it does not meet the routing independence, restore the deleted edge e re , stored in set E re and repeat steps 4.3 and 4.4.

[0028] Step 4.6: Calculate the modularity Q(n) of the split topology and store it in the set Q total middle,

[0029] Step 4.7: Determine whether all nodes have degenerated into an independent cluster (excluding the recovered edges), that is, remove E from the adjacency matrix. re The values ​​of all other elements are zero, and the formula is expressed as If the target is not achieved, repeat steps 4.2-4.6.

[0030] Step 4.8: Q total Get the value of modularity Q closest to 1 The cluster corresponding to this value is the optimal partitioning scheme for the current topology.

[0031] A community detection-based IP network routing simulation and reasoning device comprises a topology management module, a network configuration module, a network segmentation module, a change analysis module and a routing calculation module, so as to realize the topology segmentation and routing simulation and reasoning process of the IP network.

[0032] Among them, the topology management module is the functional basis of this simulation equipment. It includes the management of basic network models such as network elements and links, supports the arrangement of network models through graphical interfaces or programming interfaces, defines network topology structures, generates specific simulation network element instances such as routers and switches, and the connection methods between them, and can build a network environment according to specific network simulation scenarios.

[0033] Network configuration module. The network configuration module configures and manages IP network protocols for the simulated network element instances created by the topology management module, including the configuration of link interconnection interface IP addresses, the configuration of ISIS protocol neighbor relationships, and the configuration of network routing prefix import, providing network control plane configuration capabilities for network simulation scenarios.

[0034] The network segmentation module is based on the improved GN algorithm. It divides the cluster structure of the simulation network topology and provides the division results to the change analysis module to support the acceleration of routing reasoning calculation.

[0035] Change analysis module: The change analysis module analyzes and judges the adjustment actions of the topology management module on the network topology during the simulation process, and identifies which cluster and network elements are specifically changed and adjusted in the case of multiple network clusters, including adding and deleting links, adding and deleting nodes, etc. Combined with the division of network clusters, the module analyzes and identifies the specific computing tasks required for routing convergence, so as to achieve the purpose of narrowing the calculation scope and increasing the convergence speed. Its operation process is as follows: Figure 2 shown.

[0036] The routing calculation module includes two aspects of calculation. The first is the initial full calculation when the IP simulation task is started. The SPT tree with each node as the root node is calculated through the Dijkstra algorithm, and the node routing library is constructed based on this SPT tree. The second is that after the simulation task is started, based on the topology changes, the calculation content obtained by the change analysis module is used as input, and the node SPT tree calculation and routing library construction are executed on demand.

[0037] This solution is generally divided into two phases. Phase 1, based on the constructed simulation network and after configuring the ISIS protocol control plane, first uses the traditional Dijkstra algorithm to perform a full SPT tree calculation for all nodes in the topology, forming an SPT tree set with the node as the root. The SPT tree then generates a node routing library. Then, using the SPT tree set as input, the GN algorithm is used to segment the network topology to form several network clusters, ensuring that the SPT path between any two nodes in a cluster does not pass through another cluster. Phase 2, building on Phase 1, minimizes protocol convergence calculations by accurately identifying topology changes during network simulation, combining node SPT trees with network cluster information, and improving network simulation efficiency.

[0038] Compared with the existing technology, the advantages of the present invention are as follows: the present invention aggregates and divides the network nodes in the network, and uses the improved community detection algorithm to decompose the topology into several highly cohesive and low-coupling areas. Compared with the traditional community detection algorithm, the present invention adds the judgment of the routing independence of the cluster, thereby ensuring the accuracy of the cluster division and providing a local calculation basis for the simulation of network routing behavior; in the network topology change scenario, by identifying specific change types such as adding and deleting nodes, adding and deleting links, on the basis of the Dijkstra algorithm, the network cluster information divided by the improved community detection algorithm is superimposed, and the minimized SPT tree convergence calculation amount is combined with evaluation and analysis. Through the incremental SPT tree construction results, routing convergence calculation is further performed, and finally a steady-state network routing library is formed. This routing calculation method based on network cluster division can effectively improve the overall efficiency of large-scale network simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic diagram of an IP network routing simulation and reasoning device based on community detection provided in an embodiment of the present application;

[0040] Figure 2 A schematic diagram of a cluster-based topology change routing calculation process provided in an embodiment of the present application;

[0041] Figure 3 Interaction flow between modules provided in the embodiments of this application;

[0042] Figure 4 A network segmentation process based on community detection provided in an embodiment of the present application;

[0043] Figure 5 An example of a topology to be divided provided in an embodiment of the present application;

[0044] Figure 6 The independent cluster structure topology after division provided by the embodiment of the present application;

[0045] Figure 7 A schematic diagram of newly added intra-cluster links in one of the topology change scenarios provided in an embodiment of the present application;

[0046] Figure 8 A schematic diagram of deleting a link within a cluster, one of the topology change scenarios provided in an embodiment of the present application;

[0047] Figure 9 A schematic diagram of a newly added inter-cluster link in one of the topology change scenarios provided in an embodiment of the present application;

[0048] Figure 10A schematic diagram of deleting an inter-cluster link in one of the topology change scenarios provided in an embodiment of the present application;

[0049] Figure 11 A schematic diagram of a newly added single-link node in a cluster, one of the topology change scenarios provided in an embodiment of the present application;

[0050] Figure 12 A schematic diagram of deleting a single-link node in a cluster, one of the topology change scenarios provided in an embodiment of the present application;

[0051] Figure 13 A schematic diagram of a newly added multi-link node in a cluster in one of the topology change scenarios provided in an embodiment of the present application;

[0052] Figure 14 This is a schematic diagram of deleting a multi-link node in a cluster, one of the topology change scenarios provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] In order to deepen the knowledge and understanding of the present invention, the present invention is further described below with reference to the accompanying drawings and implementation methods.

[0054] Example 1: The present invention covers the following modules, such as Figure 1 As shown: An IP network routing simulation and reasoning device based on community detection, the device includes a topology management module, a network configuration module, a network segmentation module, a change analysis module and a routing calculation module to realize the topology segmentation and routing simulation and reasoning process of the IP network.

[0055] Module 1: Topology Management Module. The topology management module is the functional foundation of this simulation equipment. It includes the management of basic network models such as network elements and links. It supports the orchestration of network models through a graphical interface or programming interface, defines the network topology structure, generates specific simulation network element instances such as routers and switches, and the connection between them, and can build a network environment based on specific network simulation scenarios.

[0056] Module 2: Network Configuration Module. The network configuration module configures and manages IP network protocols for the simulated network element instances created by the topology management module. This includes configuring link interconnection interface IP addresses, configuring ISIS protocol neighbor relationships, and configuring network routing prefix import. This provides network control plane configuration capabilities for network simulation scenarios.

[0057] Module 3: Network Segmentation Module. Based on the improved GN algorithm, the network segmentation module divides the simulated network topology into cluster structures and provides the segmentation results to the change analysis module, supporting the acceleration of routing inference calculations.

[0058] Module 4: Change Analysis Module. The change analysis module analyzes and judges the network topology adjustment actions of the topology management module during the simulation process, and identifies which clusters and network elements are specifically targeted for changes and adjustments in the case of multiple network clusters, including adding and deleting links, adding and deleting nodes, etc. In combination with the division of network clusters, it analyzes and identifies the specific computing tasks required for routing convergence, so as to achieve the purpose of narrowing the calculation scope and increasing the convergence speed. Its operation process is as follows: Figure 2 shown.

[0059] Module 5: Routing Calculation Module. The routing calculation module includes two aspects of calculation. The first is the initial full calculation when the IP simulation task starts. It uses the Dijkstra algorithm to calculate the SPT tree with each node as the root node and construct the node routing library based on this SPT tree. The second is to perform node SPT tree calculation and routing library construction on demand after the simulation task starts, based on topology changes, using the calculation results obtained by the change analysis module as input.

[0060] The interaction process between modules is as follows Figure 3 The specific process is as follows:

[0061] Step 1: Based on the network simulation scenario, build an IP network simulation environment through a visualization or programming interface, extract the basic model of network elements and links, instantiate the simulation network elements, add links between network elements, and form the initial network topology.

[0062] Step 2: Configure the IP emulation protocol for the simulated NEs through the configuration channel, including the configuration of the interconnection interface between NEs, the configuration of the ISIS protocol neighbor relationship, and the configuration of the network routing prefix import. After the initial configuration is completed, start the simulation task.

[0063] Step 3: The routing calculation module performs a full calculation of the entire network using the Dijkstra algorithm based on the initial network topology and network element configuration information, obtains an SPT tree with each node as the root node, and uses this SPT tree to build a node routing library.

[0064] Step 4: The simulated network topology and the SPT tree set rooted at each node are used as inputs to the GN algorithm. The GN algorithm then slices the network topology to obtain several network clusters. Each cluster is a subtopology of the entire simulated network topology. The final result of this slicing is that no path from any node in a cluster to any other node in the same cluster through the SPT tree passes through another cluster. This example uses the slicing into two clusters as an example.

[0065] The following is a detailed description of step 4:

[0066] Topology such as Figure 5It shows that the topology is divided into clusters using a community detection algorithm. Since network routing protocols often use shortest path algorithms such as Dijsktra for path calculation, community detection algorithms based on edge betweenness (the number of shortest paths passing through an edge in the network) are more suitable for regional division. For example, the Girvan-Newman algorithm (GN algorithm), the Newman fast algorithm (NF algorithm) and related improved algorithms measure the degree of community division based on modularity. The modularity calculation method is as follows:

[0067]

[0068] In the formula, e is a k×k matrix, e ij represents the ratio of the number of edges connecting vertices between community i and community j to the total number of edges, a i represents the ratio of the number of edges connected to the nodes in community i to the total number of edges, Tr(e) represents the sum of the ratios of the number of edges within the same community to the total number of edges, ||E 2 || represents the sum of the matrix elements. The modularity function Q measures the difference between the proportion of edges within a community and the proportion of edges within a community in the case of random connections. When Q is close to 1, the network has a better community structure.

[0069] However, in the IP network simulation partitioning scenario, in order to decompose the network structure into highly cohesive and low-coupling regional structures and support the simulation goals of local routing calculation and global diffusion, modularity evaluation cannot be used as the only evaluation parameter. At the same time, it is necessary to add the restriction that the shortest path between nodes in the region does not cross other regions, that is, the routing independence judgment. In summary, the steps for regional division are as follows: Figure 4 As shown, the details are as follows:

[0070] Step 4.1: Convert the network topology G = (V, E) into the adjacency matrix A for ease of calculation. The following is an example of converting the topology G into the adjacency matrix A.

[0071]

[0072] Step 4.2: Compute the shortest paths between all nodes and calculate the betweenness of each edge. Below is an example of the betweenness adjacency matrix B for topology G.

[0073]

[0074] Step 4.3: Delete the edge with the largest edge betweenness. If a deleted edge is generated in step 5, exclude the deleted edge and determine the largest edge to be deleted. At the same time, set the value of the corresponding edge in the adjacency matrix A to zero.

[0075] Step 4.4: Determine whether there is a new independent cluster structure in the topology after deleting the edge, R ndenotes the n partitioned independent regions, and If the independent cluster structure is not obtained, repeat steps 4.2 and 4.3. Figure 6 is an example graph after the first independent cluster partitioning of the topology G.

[0076] Step 4.5: Determine whether the shortest path between the nodes in each cluster after deleting the edges passes through the nodes outside the cluster. Use P(u, v) to represent the shortest path between nodes u and v. For all edges e E P(u, v) (where e = (x, y) is an edge in the path P(u, v)), it must satisfy x, y E R i , the cluster R i meets the routing independence criterion. If it does not meet the routing independence criterion, restore the deleted edge e re and store it in the set E re , and repeat steps 4.3 and 4.4.

[0077] Step 4.6: Calculate the modularity Q(n) of the partitioned topology and store it in the set Q total .

[0078] Step 4.7: Determine whether all nodes have degenerated into an independent cluster (excluding the restored edges), i.e., the values of all elements in the adjacency matrix except E re are zero, which can be expressed as If the target is not reached, repeat steps 4.2-4.6.

[0079] Step 4.8: Obtain the value of Q total that is closest to 1 The cluster corresponding to this value is the optimal partitioning scheme for the current topology.

[0080] Through the above steps 1-4, the steady state of the IP simulation network is formed, including the SPT tree information and routing information of each node, as well as the network cluster information partitioned based on the GN algorithm. Subsequent changes in network topology will be based on this basic information to achieve the purpose of network rapid convergence and simplified calculation under several topology change conditions.

[0081] Step 5: On the basis of the steady-state IP simulation network, perform topology structure change operations, including adding and deleting links, adding and deleting nodes, and configuration changes of network elements.

[0082] Step 6: After the change is completed, the change analysis module accurately identifies the change operations of the network topology, which may include adding intra-cluster links, deleting intra-cluster links, adding inter-cluster links, deleting inter-cluster links, adding single-link nodes in the cluster, deleting single-link nodes in the cluster, adding multi-link nodes in the cluster, and deleting multi-link nodes in the cluster. According to the network elements involved in the above change points and the network cluster information segmented by the GN algorithm, the network element node pairs that need to be recalculated are further arranged. The routing calculation module recalculates the SPT tree of the node pairs and updates the routes. The specific calculation situation is determined by different change situations. Examples of the change scenarios involved are as follows:

[0083] Scenario 1: If Figure 7 As shown in the figure, if the change is to add a link within the cluster, the nodes within the cluster need to be fully calculated in both directions. The nodes within the cluster need to be fully calculated in both directions from the nodes in another cluster. No calculation is required between the nodes in the other cluster.

[0084] Scenario 2: If Figure 8 As shown in Figure 1, if the change is to delete a link within a cluster, the SPT trees within the cluster, between the cluster and another cluster, and between the other cluster and the cluster need to be recalculated in both directions through the link, but not between nodes in the other cluster.

[0085] Scenario 3: If Figure 9 As shown, if the change is the addition of a new inter-cluster link, this may cause the network cluster structure to change. Jump to step 4 to recalculate and divide the network cluster.

[0086] Scenario 4: If Figure 10 As shown in Figure 2, if the change is to delete an inter-cluster link, only the SPT tree between the nodes of this cluster and the other cluster that passes through the link needs to be recalculated, and no calculation is required within this cluster or the other cluster.

[0087] Scenario 5: Figure 11 As shown in Figure 2, if the change is the addition of a single-link node within a cluster, the new node within the cluster needs to be calculated pairwise to other nodes in the cluster and another cluster, but no calculation is required between nodes in the other cluster.

[0088] Scenario 6: Figure 12 As shown in Figure 2, if the change is to delete a single-link node in the cluster, the nodes from other nodes in the cluster and nodes in another cluster to the deleted node will be recalculated, and the nodes from other nodes in the cluster to another cluster and nodes in another cluster do not need to be calculated.

[0089] Scenario 7: Figure 13 As shown in Figure 2, if the change is to add a multi-link node within a cluster, two-way calculations are required between nodes within the cluster and between nodes within the cluster and nodes in another cluster, but no calculation is required between nodes in another cluster.

[0090] Scenario 8: Figure 14As shown in the figure, if the change is to delete a multi-link node in the cluster, the SPT tree between other nodes in the cluster and all nodes in the other cluster to the node is recalculated in one direction. Only the SPT tree between other nodes in the cluster and nodes in the other cluster that passes through the node needs to be recalculated, and no calculation is required between nodes in the other cluster.

[0091] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention, and equivalent changes or substitutions made on the basis of the above technical solutions fall within the scope of protection of the claims of the present invention.

Claims

1. An IP network routing simulation reasoning method based on community detection is characterized by: The method comprises the following steps: Step 1: Based on the network simulation scenario, build an IP network simulation environment through visualization or programming interfaces, extract the basic model of network elements and links, instantiate simulation network elements, add links between network elements, and form the initial network topology. Step 2: Configure the IP simulation protocol for the simulated network element through the configuration channel, including the configuration of the interconnection interface between network elements, the configuration of the ISIS protocol neighbor relationship, and the configuration of the network routing prefix import. After the initial configuration is completed, start the simulation task. Step 3: The routing calculation module performs a full calculation of the entire network using the Dijkstra algorithm based on the initial network topology and network element configuration information, obtains an SPT tree with each node as the root node, and constructs a node routing library based on this SPT tree. Step 4: Take the simulated network topology and the SPT tree set with each node as the root node as the input of the GN algorithm. The GN algorithm cuts the network topology to obtain several network clusters. Each network cluster is a sub-topology of the entire simulated network topology. The final effect of the cutting is that there is no path from any node in each cluster to the SPT tree of other nodes in the cluster that passes through another cluster. Step 5: Based on the steady-state IP simulation network, perform changes to the network topology, including adding or deleting links, adding or deleting nodes, and configuration changes of network elements. Step 6: After the change is completed, the change analysis module accurately identifies the network topology change operation. Based on the network elements involved in the change point and the network cluster information segmented by the GN algorithm, it further arranges the network element node pairs that need to be recalculated. The routing calculation module recalculates the SPT tree of the node pairs and updates the routes. Step 4 is as follows: Step 4.1: Convert the network topology G = (V, E) into the adjacency matrix A for easy calculation. Step 4.2: Calculate the shortest paths between all nodes and obtain the edge betweenness of each edge in turn. The following is the edge betweenness adjacency matrix B of topology G. Step 4.3: Delete the edge e with the largest betweenness max If deleted edges are generated in step 4.5, it is necessary to ensure Then set the corresponding edge value in the adjacency matrix A to zero, Step 4.4: Determine whether there is a new independent cluster structure in the topology after deleting the edge, R n Represents the n independent regions divided out, If no independent cluster structure is obtained, repeat steps 4.2 and 4.

3. Step 4.5: Determine whether the shortest path between the internal nodes of the independent cluster obtained in step 4.4 in the original topology passes through the external nodes of the cluster. Use P(u,v) to represent the shortest path between nodes u and v. For all edges e∈P(u,v) (where e=(x,y) is an edge in the path P(u,v)), satisfy x,y∈R i , then the cluster R i Meet the routing independence judgment. If it does not meet the routing independence, restore the deleted edge e re , stored in set E re and repeat steps 4.3 and 4.

4. Step 4.6: Calculate the modularity Q(n) of the split topology and store it in the set Q total middle, Step 4.7: Determine whether all nodes have degenerated into an independent cluster, that is, the adjacency matrix is ​​divided into E re The values ​​of all other elements are zero, and the formula is expressed as If the goal is not achieved, repeat steps 4.2-4.

6. Step 4.8: Q total Get the value of modularity Q closest to 1 The cluster corresponding to this value is the optimal partitioning scheme for the current topology.

2. The method for simulating and reasoning IP network routing based on community detection according to claim 1, wherein: In step 6, after the change is completed, the change analysis module accurately identifies the change operations of the network topology, including adding intra-cluster links, deleting intra-cluster links, adding inter-cluster links, deleting inter-cluster links, adding single-link nodes within a cluster, deleting single-link nodes within a cluster, adding multi-link nodes within a cluster, and deleting multi-link nodes within a cluster. Different routing inference strategies are executed for the above scenarios.

3. An IP network routing simulation and reasoning device based on community detection, characterized in that: Used to implement the IP network routing simulation and reasoning method based on community detection as described in any one of claims 1-2, the device includes a topology management module, a network configuration module, a network segmentation module, a change analysis module and a routing calculation module to implement the topology segmentation and routing simulation and reasoning process of the IP network.

4. The IP network routing simulation and reasoning device based on community detection according to claim 3 is characterized in that: The topology management module is the functional basis of this simulation equipment. It includes the management of basic network models such as network elements and links. It supports the arrangement of network models through a graphical interface or programming interface, defines the network topology structure, generates specific simulation network element instances such as routers and switches, and the connection between them. It can build a network environment according to the specific network simulation scenario. The network configuration module configures and manages the IP network protocol for the simulation network element instance created by the topology management module, including the configuration of the link interconnection interface IP, the configuration of the ISIS protocol neighbor relationship, the configuration of the network routing prefix import, etc., providing the network control plane configuration capability for the network simulation scenario. The network segmentation module is based on the improved GN algorithm. It divides the cluster structure of the simulation network topology and provides the division results to the change analysis module to support the acceleration of routing reasoning calculation. The change analysis module analyzes and judges the network topology adjustment actions of the topology management module during the simulation process, and identifies which clusters and network elements are specifically targeted for changes and adjustments in the case of multiple network clusters, including adding and deleting links, adding and deleting nodes, etc. Combined with the division of network clusters, the change analysis module analyzes and identifies the specific computing tasks required for routing convergence, thereby achieving the purpose of narrowing the calculation scope and increasing the convergence speed. The routing calculation module includes two aspects of calculation. The first is the initial full calculation when the IP simulation task is started. The Dijkstra algorithm is used to calculate the SPT tree with each node as the root node, and the node routing library is constructed based on this SPT tree. Second, after the simulation task is started, based on the topology changes, the calculation content obtained by the change analysis module is used as input to execute the node SPT tree calculation and routing library construction on demand.

Images