BGP failure propagation modeling method and system based on control-data biplanes

Abstract

The invention relates to the technical field of Internet routing and security, and provides a BGP failure propagation modeling method and system based on control-data biplanes. The method comprises the following steps: step 1, constructing a control plane failure propagation mechanism, a data plane failure propagation mechanism and a coupling failure propagation mechanism between a control plane and a data plane; wherein the coupling failure propagation mechanism comprises the following steps: the failure of the control plane triggers the disconnection of a data plane link, and the continuous failure of the data plane triggers the disconnection of a control plane BGP (Border Gateway Protocol) session; and step 2, on the basis of the control plane failure propagation mechanism, the data plane failure propagation mechanism and the coupling failure propagation mechanism, modeling each router in the control plane as an agent by adopting an agent-based modeling method, and realizing synchronous updating and bidirectional feedback of a control-data dual-plane state.

Description

Technical Field

[0001] This invention relates to the field of Internet routing and security technology, and in particular to a BGP failure propagation modeling method and system based on a control-data dual-plane architecture. Background Technology

[0002] BGP is the primary inter-domain routing protocol used on the Internet, responsible for exchanging routing information between different Autonomous Systems (AS). BGP failures (such as link interruptions or router malfunctions) propagate through the network's control plane, causing routing table updates and path changes. These changes ultimately affect traffic forwarding in the data plane, potentially leading to network congestion, connection interruptions, and other problems.

[0003] In recent years, many researchers have modeled the cascading failure phenomenon of inter-domain routing systems from different perspectives. In the CFM model proposed by Guo Yi et al. (Guo Y, Wang Z, Luo S, et al. A cascading failure model for interdomain routing system[J]. International Journal of Communication Systems, 2012, 25(8): 1068-1076.), Liu Y, Peng W, Su J, et al. AssessingSurvivability ofInter-domain Routing System under Cascading IEEE, 2015: In the IRS models proposed in 382-385, node failure is considered the cause of cascading failures in inter-domain routing, and the node failure condition is assumed to be traffic overload. After determining the initial failed node, its load traffic is redistributed to other nodes according to certain rules. If these nodes exceed their rated load after receiving additional traffic, new failures are generated and used as input for the next round of the loop until no new failures occur. The difference between the three models lies in the initial load setting and traffic redistribution rules. However, this modeling approach does not reflect reality. When traffic exceeds the node's forwarding capacity, only packet dropping occurs, not a crash. Traffic overload should be the failure condition for edges, not nodes.

[0004] In response, Zhu et al. (Zhu Huihu, Qiu Han, Wang Qingxian, et al. Cascaded Failure Model of Inter-Domain Routing System Based on Dual Damage Factors [J]. Computer Engineering and Applications, 2019, 55(2):92-99.DOI:10.3778 / j.issn.1002-8331.1710-0258.) optimized the model and proposed the DDF-CFM model. The model distinguishes between node failure and edge failure, considering that node failure is caused by overload of control plane UPDATE messages, and edge failure is caused by overload of data plane service traffic. After determining the initially failed node or edge, the number of UPDATE messages received by each node at different times is calculated first, and it is determined whether the node is overloaded. If it is overloaded, it is added to the node failure set. After the UPDATE message propagation is completed, traffic redistribution is performed and it is determined whether the edge is overloaded. If it is overloaded, it is failed. Finally, it is determined whether there are still failed edges or nodes. If there are, a new round of loop is entered until the network is stable. Zhang et al. (Zhang Jun, Wang Yongjie, Zhang Jingye, et al. Cascading Failure Model of Inter-Domain Routing System Based on Optimal Effective Path [J]. Information Network Security, 2021(5):90-99.DOI:10.3969 / j.issn.1671-1122.2021.05.011.) proposed a cascading failure model based on optimal effective path on the basis of DDF-CFM, considering the business relationship of the inter-domain routing system and optimizing the calculation method of edge load based on the business relationship. Zhao et al. (Zhao Wendian. Research on Damage Mechanism of Cascading Failure of Inter-Domain Routing System [D]. National University of Defense Technology, 2021.) considered the promoting effect of failure node and edge recovery on cascading failure on the basis of DDF-CFM model, and constructed a cascading failure model of inter-domain routing system considering the recovery feedback effect. In existing failure propagation models such as DDF-CFM, there are fundamental limitations in their characterization of inter-domain routing systems. Specifically, the system's control plane and data plane are modeled as two independent processes that are only sequential in time: the UPDATE message propagation process and the subsequent traffic redistribution process. This model assumes that the traffic redistribution process begins precisely after the control plane update has fully converged.

[0005] However, actual network behavior shows that traffic redistribution in the data plane occurs gradually as UPDATE messages propagate hop-by-hop, with significant overlap and interleaving between these two processes. Furthermore, existing models completely overlook the resource contention and dynamic feedback mechanisms arising from the coexistence of the control plane and data plane within the same network device entity. This deficiency manifests on two levels: First, when a network failure occurs, the control plane (consuming CPU, memory, and other computing resources) becomes overloaded due to handling massive route updates, preventing routers from generating and distributing new forwarding table entries in a timely manner. At this point, the data plane (consuming link bandwidth resources) is still forced to forward traffic based on outdated and invalid forwarding tables, directing a large amount of service traffic to invalidated paths, creating a persistent traffic black hole and service interruption. Second, this process can trigger a vicious cycle: data plane errors in forwarding can cause secondary congestion, which further exacerbates the transmission and processing delays of control plane protocol messages (such as UPDATE messages), thus further worsening the control plane's convergence process. Therefore, existing models, by failing to characterize this crucial internal coupling mechanism, inevitably and systematically underestimate the actual scope, duration, and business impact of network failure propagation. Summary of the Invention

[0006] To address the simulation distortion problem caused by neglecting the dynamic coupling between the control plane and the data plane in existing failure propagation modeling methods, this invention proposes a BGP failure propagation modeling method and system based on the control-data dual plane.

[0007] In a first aspect, the present invention provides a BGP failure propagation modeling method based on a control-data dual-plane approach, comprising: Step 1: Construct a control plane failure propagation mechanism, a data plane failure propagation mechanism, and a coupling failure propagation mechanism between the control plane and the data plane; wherein the coupling failure propagation mechanism includes: the failure of the control plane triggers the disconnection of the data plane link, and the continuous failure of the data plane triggers the disconnection of the control plane BGP session; Step 2: Based on the control plane failure propagation mechanism, the data plane failure propagation mechanism, and the coupled failure propagation mechanism, an agent-based modeling method is adopted to model each router in the control plane as an agent, thereby realizing the synchronous update and bidirectional feedback of the control-data dual-plane states.

[0008] Furthermore, in step 1, the load of the control plane is UPDATE messages, and the failure condition of the control plane is: the number of UPDATE messages received by any router of the control plane in a short period of time exceeds its processing capacity. The load of the data plane is the data traffic flowing through the links, and the failure condition of the data plane is: the actual data traffic transmitted by any link of the data plane exceeds its rated link capacity.

[0009] Furthermore, in step 1, the control plane failure propagation mechanism specifically includes: when If any router in the control plane fails, its neighboring routers will update their routing tables and send UPDATE messages to their neighboring routers after a preset maximum hold time (HOLDTIME).

[0010] Furthermore, in step 1, the data plane failure propagation mechanism specifically includes: if If any link in the time-space data plane fails, in The load on the remaining links remains unchanged at any given time.

[0011] Further, in step 1, the control plane failure triggers the data plane link disconnection, specifically including: like If any router in the control plane fails, the link in the data plane connected to the failed router will be disconnected, and the remaining links in the data plane will also be interrupted. The traffic carried at any given time is redistributed according to the optimal effective path.

[0012] Further, in step 1, the persistent failure of the data plane triggers the disconnection of the control plane BGP session, specifically including: If the duration of failure of any link in the data plane reaches the preset HOLDTIME, the routing tables of the two routers corresponding to that link in the control plane will be updated and a new UPDATE message will be generated.

[0013] Furthermore, step 2 specifically includes: Step 2.1: Initialize each router in the control plane as an independent agent. Each agent calculates the capacity of the connected links and neighbor router information based on the network topology. Step 2.2: Based on the capacity of each link and router information in the network topology, the initial load and capacity of each link are obtained by calculating the optimal effective path, and failed routers and links are removed according to the preset initial failure policy. Step 2.3: All agents execute the workflow simultaneously to identify and remove failed routers and their corresponding links; Step 2.4: Based on the current network topology, redistribute traffic to the remaining links in the data plane, recalculate the load of each link according to the optimal effective path calculation method, and determine if there are any new failed routers and links. If so, repeat step 2.3; otherwise, repeat step 2.5. Step 2.5: Check if an UPDATE message is being propagated. If so, repeat step 2.3. Otherwise, it means the message has failed and will no longer be propagated.

[0014] Furthermore, in step 2.3, the workflow includes a message parsing stage and a message sending stage; wherein, During the message parsing phase: the Agent checks whether the number of UPDATE messages received in the previous round exceeds its capacity threshold. If it does, the Agent marks the router as in a failed state and executes step 2.4; if it does not exceed the threshold, the Agent removes duplicate UPDATE messages and puts them into the forwarding queue. The Agent parses the received KEEPALIVE message and checks the liveness status of neighboring routers: if no KEEPALIVE message is received from a neighboring router within the preset HOLDTIME, the neighboring router is determined to be invalid. When the Agent determines that a neighboring router has failed, it clears all routing entries learned from that neighboring router, marks the neighboring router as unreachable, and puts the UPDATE message announcing the failure of the neighboring router into the forwarding queue. During the message sending phase: the Agent sends KEEPALIVE messages to all neighboring routers and forwards UPDATE messages in the forwarding queue to specific neighboring nodes according to the routing forwarding rules based on business relationships.

[0015] Secondly, the present invention provides a BGP failure propagation modeling system based on a control-data dual-plane architecture, comprising: The failure propagation mechanism construction unit is used to construct a control plane failure propagation mechanism, a data plane failure propagation mechanism, and a coupled failure propagation mechanism between the control plane and the data plane; wherein the coupled failure propagation mechanism includes: the failure of the control plane triggers the disconnection of the data plane link, and the continuous failure of the data plane triggers the disconnection of the control plane BGP session. The dual-plane failure propagation modeling unit is used to model each router in the control plane as an agent based on the control plane failure propagation mechanism, the data plane failure propagation mechanism and the coupled failure propagation mechanism, thereby realizing synchronous updates and bidirectional feedback of the control-data dual-plane states.

[0016] Thirdly, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method as described in the first aspect.

[0017] The beneficial effects of this invention are as follows: This invention starts with the control-data dual-plane structure of inter-domain routers, clarifying the coupling mechanism between the two in terms of resource sharing and state. Furthermore, it distinguishes the essential differences in resource consumption between the two planes, and performs refined modeling of the load and failure conditions of the control plane and data plane respectively. Based on this, it constructs failure propagation rules based on the aforementioned coupling mechanism to define how the states of the two planes influence each other in real time and bidirectionally. Finally, to efficiently simulate this complex system, agent-based modeling (ABM) technology is adopted, using the autonomous behavior and interaction of agents to run a computer simulation with a synchronous coupling mechanism.

[0018] This invention fundamentally solves the simulation timing distortion problem caused by traditional serial processing models by constructing a synchronous coupling mechanism for parallel interaction and bidirectional real-time feedback between the control plane and the data plane, significantly improving the overall fidelity and accuracy of system simulation. Regarding the coupling mechanism between the two planes in terms of resource sharing and state, by distinguishing the essential differences in resource consumption between the two, the load and failure conditions of the control plane and the data plane are modeled in detail.

[0019] This invention employs agent-based modeling and a bidirectional failure propagation mechanism. By defining the autonomous behavior and interaction rules of agents at the micro level, it achieves a synchronous coupling mechanism between the control and data planes, thereby enabling a bottom-up characterization of macro-emergent natural phenomena caused by micro-behavior. This not only avoids the limitations of traditional models that rely on manually defined macro-patterns, but also endows the model with the crucial ability to proactively discover unknown and complex failure modes. Attached Figure Description

[0020] Figure 1 A flowchart illustrating the BGP failure propagation modeling method based on a control-data dual-plane provided in this embodiment of the invention; Figure 2 A flowchart of the network model provided in this embodiment of the invention; Figure 3 This is a flowchart of the workflow of the agent provided in an embodiment of the present invention; Figure 4 This is a structural diagram of the BGP failure propagation modeling system based on a control-data dual-plane provided in an embodiment of the present invention; Figure 5 This is a structural block diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] First, the control-data dual-plane coupling mechanism is described: The control and data planes of inter-domain routing control systems exhibit strong coupling. Control plane update messages rely on data plane links, while data plane traffic requires control plane update messages to determine routes. This coupling also affects failure propagation in inter-domain routing systems, causing failures to not only spread within one plane but also "overflow" to another, potentially contributing to its failure. It's important to note that the control and data planes are abstractions based on the different functions of the inter-domain routing system. While they share a common physical network for data forwarding, this is fundamentally different from two coupled networks (e.g., the internet and power grid).

[0023] The biplane coupling failure mechanism disclosed in this invention manifests as a vicious cycle spanning the plane boundaries. This mechanism begins with initial failure in either plane and forms a closed loop through the following path: Forward Path (Control Plane Failure Triggers Data Plane Problems): When the control plane fails due to overload of node CPU, memory, etc., the direct consequence is that routing updates and forwarding table entry distribution are delayed or interrupted. This causes the data plane to continuously forward traffic based on outdated and invalid paths for an extended period, resulting in widespread traffic black holes or incorrect routing. This abnormal traffic distribution can quickly impact some links, triggering or exacerbating data plane congestion and secondary failures.

[0024] Feedback Path (Data Plane Problems Backlash Control Plane Convergence): When the data plane fails due to link congestion, it severely impacts the timely and reliable transmission of control plane protocol messages (such as BGP UPDATE and KEEPALIVE). This transmission delay or packet loss is equivalent to artificially amplifying the scale and severity of the network failure, causing the control plane to process more and more complex update messages, further exhausting its already strained CPU resources and greatly slowing down or even disrupting its convergence process.

[0025] like Figure 1 As shown, this embodiment of the invention provides a BGP failure propagation modeling method based on a control-data dual-plane approach, including: S1: Construct a control plane failure propagation mechanism, a data plane failure propagation mechanism, and a coupled failure propagation mechanism between the control plane and the data plane; the coupled failure propagation mechanism includes: control plane failure triggers data plane link disconnection, and persistent data plane failure triggers control plane BGP session disconnection.

[0026] S2: Based on the control plane failure propagation mechanism, data plane failure propagation mechanism and coupled failure propagation mechanism, an agent-based modeling method is adopted to model each router in the control plane as an agent, so as to realize the synchronous update and bidirectional feedback of the control-data plane states.

[0027] This invention provides a BGP failure propagation modeling method based on a control-data dual-plane architecture. Starting from the control-data dual-plane structure of inter-domain routers, it clarifies the coupling mechanism between the two planes in terms of resource sharing and state. Using agent-based modeling (ABM) technology, it runs a computer simulation with a synchronous coupling mechanism through the autonomous behavior and interaction of agents. This method accurately characterizes the "spillover" effect between the control plane and data plane during failure propagation for the first time, overcoming the fundamental shortcomings of existing research and providing a high-fidelity method for characterizing failure propagation.

[0028] Based on the above embodiments, the present invention provides a detailed explanation of the control plane failure propagation mechanism, data plane failure propagation mechanism, and coupling failure propagation mechanism in S1.

[0029] 1. Control plane failure propagation mechanism The load on the control plane is UPDATE (routing update) messages. A control plane failure occurs when any router on the control plane receives more UPDATE messages than its processing capacity within a short period. Nodes in the control plane are routers.

[0030] Specifically, some router nodes in the control plane cannot handle the surge in UPDATE packets in a short period of time, resulting in the inability to perform normal routing and forwarding. Furthermore, if a large number of UPDATE packets are injected into a node in a short period of time, it will cause the node to run out of memory and fail. Therefore, the load on the control plane is mainly UPDATE packets, and the resources consumed are the CPU and memory of the router nodes in the inter-domain routing system.

[0031] After receiving an UPDATE message, the router processes and forwards it. If the number of UPDATE messages received in a short period exceeds the processing capacity of the node's CPU and memory, it will cache them. Therefore, the maximum number of messages a node can process and the maximum number of messages it can queue depend on its CPU and memory resources. For any router in an inter-domain routing system... Assuming the maximum number of UPDATE messages it can process instantaneously, i.e. the maximum instantaneous load. Assuming the maximum number of UPDATE messages that can be cached is the maximum cache load. ,router Short-term processing capacity, i.e., the maximum number of messages that can be processed. For routers, the maximum instantaneous load and maximum cache load are constant values, which can be calculated based on the CPU and memory resources of the inter-domain routers.

[0032] The specific conversion methods for maximum instantaneous load and maximum cache load are as follows: For example, during the Code-Red worm incident, the mainstream inter-domain router brand used by operators was Cisco 7000. The node's UPDATE packet processing capacity was approximately 300 packets / second, and its cache capacity was approximately 130,000 packets. The processor was a Motorola 68040, single-core, 25MHz, with 64MB of memory. For comparison, the Huawei NetEngine 8000 X16, a commonly used backbone router in current inter-domain routing systems, is selected. According to the product manual, the main control board responsible for control plane data processing has a 24-core 2.5GHz CPU and 32GB of memory. Assuming that processing capacity is proportional to processor frequency and cache capacity is proportional to memory size, the instantaneous processing capacity of the HUAWEI NetEngine8000 X16 is 720,000 packets / second. Assuming that 25% of the memory space can be used for packet caching, the cache capacity is 16,640,000 packets.

[0033] Control plane failure is caused by the router receiving more UPDATE messages than it can process within a short period. Let... Represents a node (router) exist The number of UPDATE messages received at any given time can be set initially. It can also be configured according to the actual situation. Therefore, the control plane failure condition is: router exist The number of UPDATE messages received at any given time exceeds its short-term processing capacity, i.e. .

[0034] Inter-domain routers A failure will directly interrupt all its BGP sessions, causing neighboring routers to detect the interruption and trigger the generation and propagation of UPDATE messages. Therefore, the control plane failure propagation mechanism specifically includes: when... If any router in the control plane fails, At that moment, its neighboring routers will update their routing tables and send UPDATE messages to their neighboring routers. To maintain the maximum hold time preset by the holdtimer.

[0035] 2. Data plane failure propagation mechanism The load of the data plane is the data traffic flowing through the links. The failure condition of the data plane is: the actual data traffic transmitted by any link of the data plane exceeds its rated link capacity.

[0036] Specifically, the core router's data plane forms a forwarding table based on the routing table and forwards service data. Therefore, the ports of the data plane can be abstracted as router interfaces, and the traffic on the router interfaces is the load on the data plane.

[0037] While using betweenness as the calculation method for router interface load is simple and efficient, some calculated paths violate the "No Valley" principle of inter-domain routing, potentially leading to routing loops and free transit issues. Based on the business relationship constraints of inter-domain routing systems (Yang B, Zhang Y, Lu Y. A New Methods for Cascading Failures Analysis in Inter-domain Routing System[C] / / 2015 Fifth International Conference on Instrumentation and Measurement, Computer, Communication and Control (IMCCC). IEEE, 2015: 382-385.), the shortest path satisfying the "No Valley" principle is defined as the optimal effective path. Therefore, referring to the definition of the optimal effective path, assuming... Indicates in Always passing through the router and router Links between interfaces The number of optimal effective paths, Indicates in Always passing through the router and router Links between interfaces The load traffic.

[0038] Assuming that the traffic on each optimal effective path is the same, then the link exist The real-time load at time t can be represented as Then the link The initial load is If the rated load is proportional to the initial load, then the router... and router Links between interfaces The rated load can be expressed as ,in This is the tolerance factor. According to (Qiu H, Li Y, Lv P, et al. Provisioning core link bandwidths to support streaming media[C] / / 2010 2nd International Conference on FutureComputer and Communication. IEEE, 2010, 1: 543-547.), since QoS can be guaranteed when link traffic is within 80% of the link capacity, in practice, when router interface traffic exceeds 80% of the link capacity, the link needs to be expanded. The tolerance factor can be set to a reference value of 0.25, or it can be dynamically set according to the actual situation. When router interface traffic exceeds the link capacity, data packets will accumulate in the router port queue. If continuous failures lead to the exhaustion of the output queue, the router will discard subsequent data packets according to tail drop or active queue management strategies, causing network congestion and a decrease in transmission performance. Therefore, the router... and router Links between interfaces The failure condition is: .

[0039] The data plane failure propagation mechanism specifically includes: if Any link in the time data plane Failure, in At any given time, the load on the remaining links remains unchanged, i.e. ,in .

[0040] 3. Failure propagation mechanism of two-plane coupling When the control plane fails due to overload of node CPU, memory, etc., the direct consequence is that routing updates and the distribution of forwarding table entries are delayed or interrupted. Therefore, after a control plane node failure occurs, the failure will not only propagate in the control plane but also in the data plane.

[0041] Data plane failure, if it reaches a certain level, will lead to control plane failure. When the data plane link load exceeds its rated capacity, data transmission is not interrupted; instead, subsequent data packets are dropped according to strategies such as tail drop or active queue management. KEEPALIVE packets may be dropped or transmitted normally. Whether a KEEPALIVE packet is dropped depends on whether the link traffic at the moment the KEEPALIVE packet is sent exceeds its rated traffic. The router will only determine that the BGP session has timed out, perform route updates, and send UPDATE packets if it does not receive a new KEEPALIVE packet before its HOLDTIMER timer expires. Considering the long interval between KEEPALIVE packet transmissions and the extremely small data volume compared to the total traffic, it can be assumed that if the link traffic failure continues for more than the preset maximum hold time (HOLDTIME), all KEEPALIVE packets will be dropped, the BGP session will be disconnected, and the failure will propagate to the control plane. Therefore, the coupled failure propagation mechanism of the two planes is as follows: (1) Control plane failure triggers the disconnection of the data plane link, specifically including: like Any router in the control plane at any time If it fails, then the data plane contains information related to the failed router. The connected link is broken, and the remaining links in the data plane are... The traffic carried at any given time is redistributed according to the optimal effective path, that is... ,in .

[0042] (2) The continuous failure of the data plane triggers the disconnection of the control plane BGP session, specifically including: If the duration of failure of any link in the data plane reaches a preset HOLDTIME value, then the routing tables of the two corresponding routers in the control plane will be updated and new UPDATE messages will be generated. For example, in a routing network, if the link... from The continuous overload time from time to time reaches the HOLDTIME value, i.e., the data plane , Router i and router j in the control plane will update their routing tables and generate new UPDATE messages.

[0043] Building upon the above embodiments, this invention provides steps for implementing synchronous coupled simulation using Agent Based Model (ABM). In the ABM framework, by distinguishing between the environment and agents, the problem is described as multiple agents interacting with the environment and making autonomous decisions to solve it. Each router can act as an agent, possessing sensors and actuators. Sensors are used to perceive the environment and receive messages from other nodes, while actuators determine their own state and behavior and interact with the environment. Thus, by assigning an agent to each router, simulation of the control plane is achieved. The inter-domain routing system can be abstracted as a network model, or environment, carrying service traffic, composed of all agents. The actions of these agents will affect the environment; therefore, simulation of the data plane is achieved through the network model.

[0044] like Figure 2 As shown, S2 specifically includes: S2.1: Initialize each router in the control plane as an independent agent. Each agent calculates the capacity of the connected links and neighbor router information based on the network topology.

[0045] S2.2: Based on the capacity of each link and router information in the network topology, the initial load and capacity of each link are obtained by calculating the optimal effective path, and failed routers and links are removed according to the preset initial failure policy.

[0046] Specifically, S2.1 and S2.2 are the initialization phases, in which each node (router) is initialized as an independent agent, and information such as link capacity and neighbor information is calculated based on the network topology. Then, the initial load and capacity of each link are obtained by calculating the optimal effective path. Finally, the corresponding nodes and edges are selected for removal according to the initial failure policy.

[0047] S2.3: All agents execute the workflow simultaneously to identify and remove failed routers and their corresponding links.

[0048] S2.4: The network model performs traffic redistribution on the remaining links in the data plane according to the current network topology, recalculates the load of each link according to the optimal effective path calculation method, and determines whether there are any new failed routers and edges. If so, S2.3 is executed again; otherwise, S2.5 is executed.

[0049] S2.5: Check if an UPDATE message is being propagated. If so, repeat S2.3. Otherwise, it indicates failure and will no longer be propagated.

[0050] Specifically, S2.3 to S2.5 constitute the execution phase. First, each Agent follows the workflow to identify and remove failed nodes. At this time, edges connected to failed nodes will also become invalid. Then, traffic redistribution is performed, and the load of each edge is recalculated based on the optimal effective path calculation method. Failed edges are then identified and removed. If new failures occur in this round or UPDATE messages are still propagating in the network, the next iteration begins. Otherwise, the cascading failure process ends, and the network tends to stabilize.

[0051] like Figure 3 As shown in S2.3, the Agent's workflow includes a message parsing phase and a message sending phase; among which, Message parsing phase: The Agent checks whether the number of UPDATE messages received in the previous round exceeds its capacity threshold. If it does, the Agent marks the router as in a failed state and executes S2.4. If it does not exceed the threshold, the Agent removes duplicate UPDATE messages and puts them into the forwarding queue. The Agent parses the received KEEPALIVE message and checks the liveness status of the neighboring router: if no KEEPALIVE message is received from the neighboring router within the preset maximum waiting time HOLDTIME, the neighboring router is determined to be invalid. When the Agent determines that a neighboring router has failed, it clears all routing entries learned from that neighboring router, marks the neighboring router as unreachable, and puts the UPDATE message declaring the neighboring router's failure into the forwarding queue.

[0052] Specifically, the agent first checks if the number of UPDATE messages received in the previous round exceeds a threshold. If it does, the node's state is set to invalid, and no further action is taken. If the node is not invalid, the received UPDATE messages are deduplicated and placed in the forwarding queue for processing. Subsequently, the agent checks the liveness status of neighboring nodes by parsing KEEPALIVE messages. According to the BGP-4 protocol, nodes periodically send KEEPALIVE messages to nodes with which they have established connections to determine the validity of the connection. If a node does not receive a KEEPALIVE message from a neighbor within the maximum waiting time (HOLDTIME) preset by the holdtimer, it determines that the neighbor is invalid. HOLDTIME is three times the KEEPALIVE message sending interval and can be manually modified by the user. By default, HOLDTIME is set to 180 seconds, and the KEEPALIVE message sending interval is 60 seconds. Once a neighbor node is determined to be invalid, the node will clear all route entries learned from that neighbor and mark the relevant routes as "unreachable." Simultaneously, it will generate an UPDATE message announcing the route's invalidation and place it in the forwarding queue for processing.

[0053] Message sending phase: The Agent sends KEEPALIVE messages to all neighboring routers and forwards UPDATE messages in the forwarding queue to specific neighboring nodes according to the routing forwarding rules based on business relationships.

[0054] Specifically, the Agent first sends a KEEPALIVE message to all reachable neighbors, and then sends the UPDATE messages in the forwarding queue to different neighbors according to the routing forwarding rules based on business relationships.

[0055] like Figure 4 As shown, this embodiment of the invention also provides a BGP failure propagation modeling system based on a control-data dual-plane architecture, comprising: The failure propagation mechanism construction unit is used to construct the control plane failure propagation mechanism, the data plane failure propagation mechanism, and the coupled failure propagation mechanism between the control plane and the data plane; the coupled failure propagation mechanism includes: control plane failure triggers data plane link disconnection, and continuous data plane failure triggers control plane BGP session disconnection. The dual-plane failure propagation modeling unit is used to model each router in the control plane as an agent based on the control plane failure propagation mechanism, data plane failure propagation mechanism and coupled failure propagation mechanism, thereby realizing synchronous update and bidirectional feedback of the control-data dual-plane states.

[0056] Figure 5 A schematic diagram of the physical structure of an electronic device is provided. This electronic device may include: a processor 501, a communications interface 502, a memory 503, and a communication bus 504. The processor 501, communications interface 502, and memory 503 communicate with each other via the communication bus 504. The processor 501 can call logical instructions from the memory 503 to execute a BGP failure propagation modeling method based on a control-data dual-plane architecture, which includes the methods described in embodiments S1-S2 above.

[0057] Furthermore, when the logical instructions in the aforementioned memory 503 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for modeling BGP failure propagation based on control-data dual plane, characterized in that, The application comprises the following steps: Step 1: constructing a control plane failure propagation mechanism, a data plane failure propagation mechanism and a coupling failure propagation mechanism between the control plane and the data plane; Wherein the coupling failure propagation mechanism comprises: the control plane failure triggers the data plane link disconnection, and the data plane continuous failure triggers the control plane BGP session disconnection; Step 2: based on the control plane failure propagation mechanism, the data plane failure propagation mechanism and the coupling failure propagation mechanism, an agent-based modeling method is adopted to model each router in the control plane as an agent, so as to realize the synchronous update and bidirectional feedback of the control-data double plane state.

2. The method of claim 1, wherein the method is based on a control-data plane dual plane BGP failure propagation modeling method. In step 1, the load of the control plane is the UPDATE message, and the failure condition of the control plane is that the number of UPDATE messages received by any router of the control plane within a short time exceeds the processing capacity of the router; The load of the data plane is the data flow through the link, and the failure condition of the data plane is that the actual transmission data flow of any link of the data plane exceeds the rated link capacity.

3. The BGP failure propagation modeling method based on control-data dual plane according to claim 2, characterized in that, In step 1, the control plane failure propagation mechanism specifically comprises: When Any router of the time control plane fails, its adjacent routers will update the routing table and send UPDATE packet to the neighbor routers after the preset maximum holding time HOLDTIME.

4. The method of claim 2, wherein the method is based on a control-data dual plane BGP failure propagation modeling method. In step 1, the data plane failure propagation mechanism specifically includes: if If any link in the time-space data plane fails, in The load on the remaining links remains unchanged at any given time.

5. The method of claim 2, wherein the method is based on a control-data dual plane BGP failure propagation modeling method. In step 1, the control plane failure triggers the data plane link disconnection, specifically comprising: If any router in the control plane fails, the links in the data plane that are connected to the failed router are disconnected, and the remaining links in the data plane have their traffic redistributed according to the optimal effective paths. carried by the links in the data plane are redistributed according to the optimal effective paths.

6. The method of claim 2, wherein the method is based on a control-data dual plane BGP failure propagation modeling method. In step 1, the data plane continuous failure triggers the control plane BGP session disconnection, specifically comprising: If the continuous failure time of any link in the data plane reaches the preset HOLDTIME, the link updates the routing table and generates a new UPDATE message in the corresponding two routers of the control plane.

7. The method of claim 1, wherein the method is based on a control-data plane dual plane BGP failure propagation modeling method. The step 2 specifically comprises: Step 2.1: each router in the control plane is initialized as an independent agent, and each agent calculates the connected link capacity and neighbor router information according to the network topology; Step 2.2: based on the link capacity and router information in the network topology, the initial load and capacity of each link are obtained by calculating the optimal effective path, and the failed routers and links are removed according to the preset initial failure strategy; Step 2.3: all agents execute the work flow simultaneously, and the failed routers and corresponding links are counted and removed; Step 2.4: according to the current network topology, the remaining links in the data plane are executed for flow redistribution, the load of each link is recalculated according to the optimal effective path calculation method, whether there is a new failed router and link is judged, if yes, step 2.3 is executed again, if not, step 2.5 is executed; Step 2.5: whether there is an UPDATE message propagation is detected, if yes, step 2.3 is repeated, if not, it means that the failure no longer propagates.

8. The method of claim 7, wherein the method is based on a control-data dual plane BGP failure propagation modeling method. In the step 2.3, the work flow comprises a message parsing stage and a message sending stage; wherein, In the message parsing stage, the agent checks whether the number of UPDATE messages received in the last round exceeds the threshold, if yes, the router is marked as a failure state and step 2.4 is executed, if not, the received UPDATE messages are de-duplicated and put into the forwarding queue; The agent analyzes the received KEEPALIVE message, checks the alive state of the neighbor router: if the KEEPALIVE message from the neighbor router is not received within the preset HOLDTIME, the neighbor router is determined to be invalid; When the agent determines that a neighbor router is invalid, all route entries learned from the neighbor router are cleared, the neighbor router is marked as unreachable, and an UPDATE message announcing the invalidity of the neighbor router is put into the forwarding queue; The message sending stage: the agent sends KEEPALIVE messages to all neighbor routers, and forwards the UPDATE messages in the forwarding queue to specific neighbor nodes according to the routing forwarding rules based on commercial relations.

9. A system for modeling BGP failure propagation based on control-data dual plane, characterized in that, Comprise: A failure propagation mechanism construction unit is configured to construct a control plane failure propagation mechanism, a data plane failure propagation mechanism, and a coupling failure propagation mechanism between the control plane and the data plane; The coupling failure propagation mechanism comprises: the control plane failure triggers the data plane link disconnection, and the data plane continuous failure triggers the control plane BGP session disconnection; A dual-plane failure propagation modeling unit is configured to model each router in the control plane as an agent based on the control plane failure propagation mechanism, the data plane failure propagation mechanism, and the coupling failure propagation mechanism, and to realize the synchronous update and bidirectional feedback of the control-data dual-plane state by using an agent-based modeling method.

10. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the method of any one of claims 1 to 8 when executing the program.

Images