A method and device for intelligent connection recommendation and dynamic networking mode switching

Abstract

The application discloses a method and device for intelligent connection recommendation and dynamic networking mode switching, and relates to the technical field of software-defined wide area network, and the method comprises the following steps: collecting multidimensional data according to a preset period, wherein the multidimensional data comprises basic attributes of each site, performance parameters of each link, network historical operation data and target service requirements; when a preset trigger condition is met, generating an optimal recommended configuration based on the multidimensional data by using a weighted scoring and AI-assisted decision mechanism, wherein the optimal recommended configuration comprises an optimal topology structure and an optimal connection scheme; and displaying the optimal recommended configuration through a visual interface. The application can ensure that the network is continuously stable and usable, and the service is continuously operated without interruption.

Description

Technical Field

[0001] This application relates to the field of software-defined wide area network (SDWAN) technology, and in particular to a method and apparatus for intelligent connection recommendation and dynamic networking mode switching. Background Technology

[0002] As enterprise networks continue to expand and their structures become increasingly complex, SD-WAN (Software-Defined Wide Area Network) is widely used in multi-branch, large-scale, and heterogeneous network scenarios, bringing users more flexible and intelligent connectivity and management capabilities.

[0003] Currently, in an SD-WAN environment, during the initial deployment phase of a network, network administrators manually specify the connection method (such as star, full mesh, Hub-Spoke, point-to-point, etc.) between each site (CPE node) using network topology design tools or the SD-WAN console, based on business needs and network scale, and configure link parameters and routing policies one by one.

[0004] Network changes or expansions using this type of solution typically rely on manual topology readjustment and manual configuration of new links. In large-scale networking or multi-branch dynamic expansion scenarios, manual operations are cumbersome, maintenance workload is large, and the risk of misoperation and configuration oversight is high, which can easily affect network availability and service continuity. Summary of the Invention

[0005] The purpose of this application is to provide a method and apparatus for intelligent connection recommendation and dynamic networking mode switching, which can ensure that the network is continuously stable and available and that services can continue to operate without interruption.

[0006] To achieve the above objectives, this application provides the following solution:

[0007] Firstly, this application provides a method for intelligent connection recommendation and dynamic networking mode switching, the method comprising:

[0008] Multi-dimensional data is collected according to a preset cycle. The multi-dimensional data includes: basic attributes of each site, performance parameters of each link, historical network operation data, and target service requirements.

[0009] When the preset triggering conditions are met, based on the multi-dimensional data, a weighted scoring and AI-assisted decision-making mechanism is used to generate the optimal recommended configuration, which includes: the optimal topology and the optimal connection scheme.

[0010] The optimal recommended configuration is displayed through a visual interface.

[0011] In one embodiment, the preset triggering conditions include: network node coming online, network node going offline, SLA declining, link anomaly, or administrator actively initiating configuration requests.

[0012] Clearly define trigger conditions for multiple scenarios to achieve precise and timely configuration optimization without requiring manual monitoring of network changes, thus avoiding the risk of network unavailability and business interruption caused by response delays.

[0013] In one embodiment, generating the optimal recommended configuration based on the multi-dimensional data, using a weighted scoring and AI-assisted decision-making mechanism, includes:

[0014] Determine the scoring criteria for each candidate topology;

[0015] The scoring indicators are weighted to obtain a score for each candidate topology;

[0016] The candidate network topologies are sorted according to their scores, and the top N candidate network topologies are selected as the target topologies.

[0017] Generate connection schemes corresponding to each target topology;

[0018] The target topology and its corresponding connection scheme are sampled and optimized through an AI-assisted decision-making mechanism in multiple rounds to obtain the optimal topology and the optimal connection scheme.

[0019] The optimal configuration generation is refined through a structured process. By using scoring indicators, weighted sorting, and AI-based multi-round correction, subjective bias and interference from invalid solutions are reduced, thereby improving the accuracy and feasibility of the configuration.

[0020] In one embodiment, the weighted scoring of the scoring metrics to obtain a score for each candidate topology includes:

[0021] Calculate a quantified value for each scoring metric for each candidate topology;

[0022] Obtain the weight value corresponding to each scoring indicator;

[0023] The score for each candidate topology is obtained by calculating the quantized value and weight value corresponding to each scoring metric for each candidate topology.

[0024] By quantifying the scoring indicators and combining them with weighted scores, the problem of traditional subjective evaluation is solved, making the merits and demerits of topologies comparable and objective, and reducing the risk of misselecting the optimal solution.

[0025] In one embodiment, obtaining the weight value corresponding to each scoring indicator includes:

[0026] The initial weights of each scoring indicator are obtained based on the current network size and geographical distribution.

[0027] Obtain the preference weight values ​​for each input rating indicator;

[0028] Reinforcement learning is used to obtain the weight adjustment of each scoring indicator;

[0029] Based on the initial scenario weight, preference weight value, and weight adjustment amount of each scoring indicator, obtain the temporary weight value of each scoring indicator;

[0030] The temporary weight values ​​of each scoring indicator are normalized to obtain the corresponding weight value for each scoring indicator.

[0031] By using a multi-dimensional weight acquisition method, the weights can be dynamically adapted to network size, geographical distribution, and business preferences, thereby improving the fit between topology assessment and actual scenarios.

[0032] In one embodiment, the step of sampling the target topology and the corresponding connection scheme using an AI-assisted decision-making mechanism to perform multiple rounds of correction and optimization to obtain the optimal topology and optimal connection scheme includes:

[0033] Simulate running each recommended configuration;

[0034] During simulation, obtain the weight values ​​of the scoring indicators to be adjusted in the target topology;

[0035] The score for each target topology is re-obtained based on the weight values ​​of the scoring indicators to be adjusted and the weight values ​​of the scoring indicators not adjusted in the target topology;

[0036] The scores for each reacquired target topology are reordered;

[0037] The target topology and corresponding connection schemes are sampled and optimized through multiple rounds of AI-assisted decision-making to obtain the optimal topology and optimal connection scheme.

[0038] This disclosure involves simulation, weight adjustment, and multiple rounds of AI optimization to reduce the risk of configuration implementation, avoid one-time decision-making errors, and improve the adaptability of the solution to the network.

[0039] In one embodiment, obtaining the weight values ​​of the scoring indicators to be adjusted in the target topology includes:

[0040] Receive the weight values ​​of the rating indicators to be adjusted from the user's input;

[0041] or,

[0042] Based on historical performance and current network status changes, the weight values ​​of the scoring indicators to be adjusted in the target topology are automatically obtained.

[0043] The two weight adjustment methods disclosed herein take into account both customization needs and automation efficiency, satisfying administrators' control over key scenarios while reducing manual intervention in routine scenarios.

[0044] In one embodiment, the step of sampling an AI-assisted decision-making mechanism to perform multiple rounds of correction and optimization on the reordered target topology and corresponding connection schemes to obtain the optimal topology and optimal connection scheme includes:

[0045] Obtain feature vectors of the current network scene characteristics through AI models;

[0046] Based on feature vectors, the AI ​​model performs multiple rounds of correction on the target topology and the corresponding index weights to obtain the optimal weights.

[0047] The score for each target topology is obtained again based on the corrected weight values;

[0048] The scores of each target topology obtained again are sorted again, and the target topology with the highest score is the optimal topology. The connection scheme corresponding to the optimal topology is the optimal connection scheme.

[0049] This disclosure uses AI to extract scene features and adjust weights to solve the problem of insufficient scene adaptation, making the optimal solution more suitable for actual network scenarios and further improving configuration accuracy.

[0050] In one embodiment, after displaying the optimal recommended configuration through a visual interface, the method further includes:

[0051] Upon receiving the change instruction, simulate the network configuration change according to the optimal recommended configuration;

[0052] After detecting a failure to change the network configuration, the system will automatically roll back to the configuration before the network configuration change was simulated.

[0053] After the network configuration change is detected, the network SLA is automatically verified to determine whether the network meets the service requirements after the switch.

[0054] If satisfied, switch directly to the optimal recommended configuration;

[0055] If the conditions are not met, the configuration will be automatically rolled back to the state before the network configuration change.

[0056] This disclosure constructs a change security closed loop, employing simulation verification, automatic rollback, and SLA verification to avoid network interruptions caused by changes, thus ensuring continuous network and service stability. Secondly, this application provides an apparatus for intelligent connection recommendation and dynamic networking mode switching, the apparatus comprising:

[0057] The data acquisition module is used to collect multi-dimensional data according to a preset period. The multi-dimensional data includes: basic attributes of each site, performance parameters of each link, historical network operation data, and target service requirements.

[0058] The intelligent recommendation engine module is used to generate the optimal recommendation configuration based on the multi-dimensional data when the preset trigger conditions are met, using a weighted scoring and AI-assisted decision-making mechanism. The optimal recommendation configuration includes: the optimal topology and the optimal connection scheme.

[0059] A visual interface is used to display the optimal recommended configuration.

[0060] According to the specific embodiments provided in this application, the following technical effects are disclosed:

[0061] This system proactively collects multi-dimensional information from each site according to a preset cycle, including basic attributes, link performance parameters, historical network operation data, and target business requirements. Unlike traditional solutions that rely solely on initial business needs and network scale for static planning, this system provides real-time and comprehensive data support for subsequent decisions. When preset trigger conditions are met, administrators do not need to manually readjust the topology and configure links. Instead, based on the collected multi-dimensional data, a weighted scoring method is used to initially screen reasonable solutions. Then, combined with an AI-assisted decision-making mechanism, the optimal topology and connection scheme are automatically generated. This significantly reduces the complexity of manual operations, lowers the workload of maintenance, and reduces the risk of errors and configuration omissions. It also avoids network interruptions or service disruptions during changes, thus solving the drawbacks of traditional manual configuration in large-scale networking or dynamic expansion, ensuring that the network remains stable and available, and services continue uninterrupted. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a flowchart illustrating a method for intelligent connection recommendation and dynamic networking mode switching according to an exemplary embodiment;

[0064] Figure 2 This is a flowchart illustrating a method for intelligent connection recommendation and dynamic networking mode switching according to an exemplary embodiment;

[0065] Figure 3 A schematic diagram of the functional modules of a device for intelligent connection recommendation and dynamic networking mode switching provided in an embodiment of this application;

[0066] Figure 4 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0067] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0068] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0069] Figure 1 This is a flowchart illustrating a method for intelligent connection recommendation and dynamic networking mode switching according to an exemplary embodiment, such as... Figure 1 As shown, the method includes the following steps S101-S103:

[0070] In step S101, multi-dimensional data is collected according to a preset period. The multi-dimensional data includes: basic attributes of each site, performance parameters of each link, historical network operation data, and target service requirements.

[0071] The basic attributes include: geographical location, node type, CPE model, port resources and cost; the performance parameters include: bandwidth utilization, latency, SLA indicators and historical health indicators; the network historical operation data includes: historical abnormal event logs and network change event logs; the target service requirements include service SLA requirements, service priority classification and key service types.

[0072] The basic attributes of a site reflect the inherent capabilities of network nodes and are the hardware constraints for network topology:

[0073] Geographic location records the physical location of a site, which directly affects topology design;

[0074] Node type: Distinguish between core nodes, edge nodes, etc., and determine the function of the node in the topology (core nodes need to serve as the traffic forwarding hub).

[0075] The CPE model clearly indicates the protocols supported by the device (such as IPsec / GRE), maximum bandwidth, and other hardware capabilities, avoiding the recommendation of connection solutions that exceed the device's performance (e.g., if a CPE only supports 100Mbps, then a 200Mbps link is not recommended).

[0076] Port resource statistics include the number of idle / used ports (e.g., 4 gigabit Ethernet ports, 1 idle), and the link expansion capability is limited (non-critical link resources should be released first when there are insufficient idle ports).

[0077] Costs include equipment procurement, link leasing, and other expenses, balancing performance and economy (e.g., ordinary branches prioritize low-cost internet links).

[0078] Link performance parameters reflect the real-time operational quality of the network and are the core basis for judging whether the current network meets the standards.

[0079] Bandwidth utilization is calculated as the ratio of actual bandwidth used to total bandwidth (e.g., 80%). If it exceeds the threshold (e.g., 70%), expansion or traffic splitting solutions should be recommended.

[0080] The latency record shows the data transmission time (e.g., 60ms). For latency-sensitive services (e.g., video conferencing), the latency must be kept below the service threshold (e.g., 50ms).

[0081] SLA metrics include committed metrics such as link availability (e.g., 99.9%) and packet loss rate (e.g., 1.5%), and optimization requirements are triggered when these metrics are not met.

[0082] Historical health metrics are stability scores (e.g., 1-10) calculated based on 3 months of data. Links with low health scores should be replaced or have backups added.

[0083] Historical anomaly event logs in network historical operation data are used to record past faults (such as link interruption time, cause, and scope of impact). For example, if a certain link is always interrupted during the morning peak on the 5th of each month, it can be recommended to switch to a backup link in advance. Network change event logs are used to record historical topology adjustments (such as changing from a star topology to a Hub-Spoke topology) and their effects (such as a 10% improvement in SLA), providing a reference for similar scenarios (topologies with good historical performance are preferred for networks of the same size).

[0084] The target business requirements are used to clarify the service targets of the network and determine the priority and direction of network topology.

[0085] Business SLA requirements are customized standards for different businesses (e.g., core transaction business requires latency ≤30ms, general office business requires latency ≤100ms), and the solution must meet the corresponding requirements;

[0086] Business priority classification: differentiate between Level 1 (core transactions), Level 2 (video conferencing), and Level 3 (employee internet access) services, and prioritize high-priority services when allocating resources (e.g., if bandwidth is insufficient, Level 1 services will be prioritized first).

[0087] Key business types are classified according to transmission characteristics (such as real-time interactive business and big data transmission), and matched with link characteristics (real-time business prioritizes low-latency dedicated lines, and big data business prioritizes high-bandwidth links).

[0088] The "preset period" is not a fixed value, but is flexibly adjusted according to the "frequency of change" of the data type and "decision requirements":

[0089] High-frequency data collection items (high real-time requirements): Link performance parameters (bandwidth utilization, latency, etc.) are collected every 1-5 minutes to ensure the capture of real-time network fluctuations (such as bandwidth congestion caused by sudden traffic surges).

[0090] Mid-frequency data collection items (periodic changes): Target business needs (such as business priority adjustments) and node port resources (such as port occupancy status) are collected once a day to adapt to daily changes in business and resources;

[0091] Low-frequency data collection items (relatively static): Basic site attributes (such as geographical location, CPE model) are collected once a week or month to avoid unnecessary resource consumption;

[0092] Triggered data collection: When network anomalies occur (such as link interruption) or business changes occur (such as the addition of new core services), supplementary data collection is initiated immediately to ensure that no critical data is missed.

[0093] In step S102, when the preset triggering conditions are met, based on the multi-dimensional data, a weighted scoring and AI-assisted decision-making mechanism is used to generate the optimal recommended configuration, which includes the optimal topology and the optimal connection scheme.

[0094] The preset triggering conditions include: network node coming online, network node going offline, SLA declining, link anomaly, or administrator actively initiating configuration requests.

[0095] The triggering conditions cover key change scenarios for the network and services, ensuring timely response when adjustments are needed:

[0096] Network node online / offline: When adding a branch node, the topology needs to be expanded (e.g., connecting the new node to the Hub); when a node is offline, redundant links need to be deleted (to avoid unnecessary resource occupation).

[0097] SLA decline: When business SLA metrics (such as availability and latency) fail to meet the standards for three consecutive times, optimization (such as replacing links or adjusting topology) is initiated.

[0098] Link anomalies: Frequent link interruptions, persistent bandwidth utilization exceeding thresholds, and other abnormal states require recommendations for repair or replacement solutions.

[0099] Administrator-initiated: Supports manual triggering (such as before planned expansion) to meet proactive optimization needs.

[0100] In terms of mode switching execution, the system has a flexible automatic triggering mechanism. When real-time SLA detection finds that any indicator in the current network mode deviates significantly from the threshold (such as primary and backup link congestion, latency jitter, node downtime, etc.), AI judges that the current service is facing high risk, or automatic inspection before a major business event triggers a switch, the system can automatically enter the switch process; of course, it also supports administrators to initiate the switch through interface approval or manual operation.

[0101] In one embodiment, the step of generating the optimal recommended configuration based on the multi-dimensional data, using a weighted scoring and AI-assisted decision-making mechanism, includes the following sub-steps A1-A5:

[0102] A1. Determine the scoring criteria for each candidate topology.

[0103] Determine the scoring metrics (latency, bandwidth utilization, geographical distribution, cost, etc.) for each candidate topology m (star, fully interconnected, hub-spoke, dual-center, etc.).

[0104] Here, a table can be pre-defined to indicate the correspondence between candidate topologies and their corresponding scoring metrics.

[0105] A2. Apply weighted scoring to the scoring indicators to obtain the score for each candidate topology.

[0106] In one embodiment, the weighted scoring of the scoring metrics to obtain a score for each candidate topology includes the following sub-steps A21-A23:

[0107] A21. Calculate the quantitative value for each scoring metric for each candidate topology.

[0108] For each candidate topology m, calculate the quantized value f_i(m) for each scoring index i, and obtain the quantized score {f_1(m),f_2(m),...,f_n(m)} for each index.

[0109] Latency: f_latency(m) = 1 - (average latency of topology m / maximum allowable latency of service) (if average latency ≤ maximum latency, score ≥ 0; otherwise, score < 0, marked as risk).

[0110] Bandwidth utilization: f_bandwidth(m) = 1 - average bandwidth utilization of topology m (the lower the utilization, the higher the score);

[0111] Geographical distribution: f_geography(m) = 1 - (average distance between core nodes and edge nodes / preset reasonable distance threshold) (the closer the distance, the higher the score);

[0112] Cost: f_cost(m) = 1 - (total cost of topology m / total budget) (the lower the cost, the higher the score).

[0113] A22. Obtain the weight value corresponding to each scoring indicator.

[0114] In one embodiment, obtaining the weight value corresponding to each scoring indicator includes the following sub-steps A221-A225:

[0115] A221. Obtain the initial weights of each scoring indicator based on the current network size and geographical distribution.

[0116] A222. Obtain the preference weight values ​​for each input rating indicator.

[0117] A223. Use reinforcement learning to obtain the weight adjustment amount of each scoring indicator.

[0118] A224. Based on the initial scenario weight, preference weight value, and weight adjustment amount of each scoring indicator, obtain the temporary weight value of each scoring indicator.

[0119] A225. Normalize the temporary weight values ​​of each scoring indicator to obtain the weight value corresponding to each scoring indicator.

[0120] A. Generate the initial weight b_i for the i-th metric in the scenario.

[0121] The scene classification module automatically outputs the initial weights b_i based on the current network size (such as decision tree), geographical distribution, and other characteristics.

[0122] B. Overlay the administrator's preference weight value a_i for the i-th indicator:

[0123] Administrators can configure this through the interface (e.g., manually increasing a_latency = +0.08, decreasing a_cost = -0.05).

[0124] The weights after the offset are temporarily set to: w_i_temp1 = b_i + a_i.

[0125] C. The weight adjustment Δw_i of the i-th indicator obtained by superimposing reinforcement learning:

[0126] The reinforcement learning module (Q-Learning) uses the "historical topology switching effect" as the reward signal (e.g., SLA improvement rate +0.3 points, failure rate increase -0.2 points).

[0127] With "current weight combination w_i_temp1" as the state and "adjustment direction of Δw_i (+0.02 / -0.02 / unchanged)" as the action, the optimal Δw_i is found through training;

[0128] The adjusted weights are temporarily set as: w_i_temp2 = w_i_temp1 + Δw_i.

[0129] D. Constraints and Normalization:

[0130] Use the `clamp` function to limit the range: `w_i_clamp=clamp(w_i_temp2,L_i,U_i)` (e.g., `L_cost=0.1`, `U_cost=0.3`).

[0131] Normalization ensures that the sum is 1: w_i = w_i_clamp / Σw_i_clamp.

[0132] A23. Obtain the score for each candidate topology based on the quantized value and weight value corresponding to each scoring indicator of each candidate topology.

[0133] Topology score calculation (S(m)=Σw_i・f_i(m)):

[0134] For each candidate topology m, calculate the comprehensive score according to the formula: S(m)=w_1·f_1(m)+w_2·f_2(m)+...+w_n·f_n(m);

[0135] The higher the score, the better the topology m is adapted to the current scenario, preferences, and historical effects.

[0136] A3. Sort the candidate network topologies according to their scores, and select the top N candidate network topologies as the target topologies.

[0137] The candidate network topologies are sorted according to the S(m) score, and the top N candidate network topologies are selected as the target topologies.

[0138] A4. Generate the connection schemes corresponding to each target topology.

[0139] Based on the current node capabilities, CPE models, geographical locations, bandwidth status, and business requirements of each site, specific connection schemes are generated for each target topology. Each two-line scheme includes: which nodes serve as primary and backup nodes, which nodes establish connections with each other, and how bandwidth is allocated. Recommended configurations include: the target topology and its corresponding specific connection scheme.

[0140] I. Implementation of star topology connection scheme.

[0141] The core of a star topology is determining the unidirectional connection between a central node and edge nodes. The scheme generation follows a four-step logic: "Center Optimization → Edge Association → Link Adaptation → Bandwidth Allocation".

[0142] Central node selection: Select central nodes based on comprehensive node capabilities and stability: Prioritize core nodes with the highest CPE model performance score (e.g., Huawei AR650 is better than ordinary edge devices), the lowest CPU / memory load (utilization <50%), and the longest historical fault-free period (continuous online >90 days); if there are multiple candidates, combine geographical location (select the node with the closest average distance to other nodes to reduce overall link latency).

[0143] Edge Nodes and Connection Relationships: All nodes except the central node are edge nodes, and each edge node establishes only a single connection with the central node. Link Selection Criteria: Existing links with latency <50ms and bandwidth utilization <70% (with redundancy reserved) are given priority. If multiple links are available, they are selected based on the following priority order: "fiber optic link > microwave link > 5G wireless link" (in descending order of stability).

[0144] Bandwidth allocation strategy: Based on service priority, 60% of the link bandwidth is reserved for high-priority services (e.g., level 5), 40%-50% for medium-priority services (levels 3-4), and 20%-30% for low-priority services (levels 1-2). The total allocated bandwidth for a single link shall not exceed 80% of its contracted bandwidth (to avoid congestion). If the total service demand of the edge node exceeds the link's carrying capacity, a "service priority degradation reminder" will be automatically triggered (e.g., temporarily limiting the rate of low-priority services).

[0145] II. Implementation of the dual-center topology connection scheme.

[0146] The dual-center topology requires the construction of an architecture of "guaranteed interconnection bandwidth between primary and backup centers + dual-homing of edge nodes", with the core logic being "primary / backup optimization → interconnection reinforcement → dual edge connections → differentiated bandwidth allocation".

[0147] Primary and backup center selection: Primary center node: Select the core node with the highest CPE performance score and the most high-priority services (level 5); Backup center node: Select the node with the closest geographical distance to the primary center (<500 km for cross-city, <50 km for intra-city), interconnection link latency <30ms and SLA compliance rate ≥99.9% from the top 3 candidate nodes in performance score (to ensure primary and backup switching efficiency).

[0148] Connection design:

[0149] Interconnection between primary and backup centers: At least two independent links (such as fiber optic cables from different operators) must be provided. The primary link carries the data synchronization traffic between the primary and backup (bandwidth ≥ 30% of the total traffic of both), and the backup link serves as redundancy (triggering condition: primary link utilization > 80% or failure).

[0150] Edge node connection: Each edge node connects to both the primary and backup centers simultaneously. The primary link prioritizes links with low latency (carrying 80% of the traffic), while the backup link selects links with high stability (carrying 20% ​​of the traffic and switching only when the primary link fails).

[0151] Bandwidth allocation strategy: The bandwidth of the main center link is reserved according to "service priority × 1.2" (high redundancy), and the bandwidth of the backup center link is reserved according to "service priority × 0.8"; the main and backup interconnection links are allocated separately to ensure that the data synchronization bandwidth is not less than 20% of the total business volume of the main center (to avoid data loss during switching).

[0152] III. Implementation of Hub-Spoke Topology Connection Scheme.

[0153] The Hub-Spoke topology needs to be divided into regions based on geographical clusters to achieve "centralized operation within a region and interconnection between regions". The solution generation logic is "region division → Hub node selection → Spoke affixing → cross-region link planning".

[0154] Regional division and Hub node selection: Cluster nodes based on their geographical location (e.g., the same city or a region with a radius of less than 200 kilometers is divided into a Spoke cluster); select 1 Hub node within each cluster - prioritize nodes with CPE models that support 10 Gigabit optical ports, port resources ≥ the number of cluster nodes (to meet all Spoke connections), and located in the geographical center of the cluster (to reduce the link length within the region).

[0155] Connection design:

[0156] Within the region: All Spoke nodes connect only to the Hub node of this cluster, and the link selection prioritizes using fiber optic cables within the same city (latency <20ms).

[0157] Inter-regional: Each Hub node is interconnected with the core Hub node (usually the Hub in the cluster where the headquarters is located), forming a hierarchical structure of "core Hub → regional Hub → Spoke". Cross-regional links must meet the requirement of bandwidth utilization of <60% (to ensure cross-regional business).

[0158] Bandwidth allocation strategy: Links within the region are allocated according to "total demand of this cluster's services × 1.1" (with 10% redundancy reserved); links across regions are allocated according to "inter-regional interactive service volume × 1.3" (considering the stability fluctuations of cross-city links), and high-priority cross-regional services (such as headquarters-branch video conferencing) occupy 50% of the link bandwidth separately.

[0159] IV. Implementation of the fully interconnected topology connection scheme.

[0160] A fully interconnected topology requires all nodes to be directly connected to each other, and is suitable for small-scale networks (number of nodes < 10). The solution generation logic is "full link coverage → priority sorting → dynamic bandwidth adaptation":

[0161] Connection design: Traverse all node pairs to ensure that there is at least one link between any two nodes (prioritize reusing existing links, and plan new links if no existing links exist). Link type selection: Use fiber optic cables for nodes within the same city (low cost, low latency), and use dedicated lines from carriers for nodes across cities (guaranteed SLA).

[0162] Bandwidth allocation strategy: Dynamic allocation based on the frequency of inter-node business interactions: For node pairs with the highest interaction frequency (e.g., daily average data volume > 100GB), link bandwidth is allocated at a ratio of "interaction volume × 1.5" (high redundancy); for node pairs with low interaction frequency (e.g., daily average < 10GB), bandwidth is allocated at a ratio of "interaction volume × 1.2". Simultaneously, 20% emergency bandwidth is reserved for all links (to handle sudden traffic surges).

[0163] Redundancy control: If the number of nodes exceeds 8, the top 30% of node pairs with the highest interaction frequency will be automatically selected to retain full connectivity, while the remaining node pairs will communicate indirectly through "relay nodes" (to avoid an explosion in the number of links and control costs).

[0164] A5. The target topology and its corresponding connection scheme are sampled and optimized through multiple rounds of AI-assisted decision-making mechanism to obtain the optimal topology and the optimal connection scheme.

[0165] In one embodiment, the step of sampling the target topology and the corresponding connection scheme using an AI-assisted decision-making mechanism to perform multiple rounds of correction and optimization to obtain the optimal topology and optimal connection scheme includes the following sub-steps A51-A55:

[0166] A51. Simulate running each recommended configuration.

[0167] A52. During simulation, obtain the weight values ​​of the scoring indicators to be adjusted in the target topology.

[0168] For example, it can receive the weight values ​​of the scoring indicators to be adjusted from user input; or, it can automatically obtain the weight values ​​of the scoring indicators to be adjusted in the target topology based on historical performance and current network status changes.

[0169] After generating specific connection schemes for each target topology, the system enters the evaluation phase of the recommended configuration. The core of this phase is to combine manual intervention with automatic optimization and dynamically calibrate the scoring logic.

[0170] Simulate and run each recommended configuration, and evaluate each recommended configuration:

[0171] During simulation, administrators can flexibly configure the weight of each indicator according to actual business needs. For example, when a region experiences sudden bandwidth shortage during simulation, the administrator can temporarily increase the weight of the "bandwidth utilization" indicator. For financial business scenarios, the weight of "reliability (SLA / historical health)" can be manually increased.

[0172] During simulation, the system can also automatically and dynamically adjust the weights based on historical performance (such as the actual SLA compliance rate and fault repair time of a certain type of solution in the past 3 months) and current network status changes (such as sudden increases in node load and fluctuations in link latency). If historical data shows that "the actual failure rate of the solution increases when the cost weight is too high", the system will fine-tune Δw_i through the algorithm to reduce the proportion of cost weight. If the link latency in a certain area is detected to be continuously increasing, the weight of the "latency" indicator will be automatically increased to ensure that the evaluation logic matches the current network status.

[0173] A53. Re-obtain the score for each target topology based on the weight values ​​of the scoring indicators to be adjusted and the weight values ​​of the scoring indicators not adjusted in the target topology.

[0174] A54. Reorder the scores for each reacquired target topology.

[0175] A55. The target topology and corresponding connection scheme after reordering are sampled and optimized through multiple rounds of AI-assisted decision-making mechanism to obtain the optimal topology and optimal connection scheme.

[0176] In one embodiment, the step of sampling the AI-assisted decision-making mechanism to perform multiple rounds of correction and optimization on the reordered target topology and the corresponding connection scheme to obtain the optimal topology and optimal connection scheme includes the following sub-steps A551-A554:

[0177] A551. Obtain feature vectors of the current network scene characteristics through AI models.

[0178] A552. Based on feature vectors, the AI ​​model performs multiple rounds of corrections on the target topology and the corresponding index weights to obtain the optimal weights.

[0179] A553. Obtain the score for each target topology again based on the corrected weight values.

[0180] A554. The scores of each target topology obtained again are sorted again, and the target topology with the highest score is the optimal topology. The connection scheme corresponding to the optimal topology is the optimal connection scheme.

[0181] Based on the evaluation and weight adjustment of recommended configurations, the system introduces an AI-assisted decision-making mechanism. Through multiple rounds of correction and optimization, it finally generates an accurate list of recommended solutions.

[0182] Each recommended configuration, after evaluation and correction, is further refined using AI-assisted decision-making.

[0183] The AI ​​model first performs in-depth identification of the characteristics of the current network scene, including node size (small / medium / large, such as <10 nodes being small and 10-50 nodes being medium), geographical distribution (clustered in the same city / dispersed across cities / mixed distribution), and business characteristics (high bandwidth requirements / low latency sensitivity / high reliability requirements), and constructs a scene feature vector.

[0184] Based on scene characteristics, the AI ​​model performs multiple rounds of adjustments to the target topology and indicator weights:

[0185] At the topology level: If a scenario of "medium-sized + cross-city distribution + mixed business" is identified, and historical data shows that the SLA compliance rate of the dual-center topology is 12% higher than that of the star topology in this scenario, the recommendation priority of the dual-center topology will be increased, even if its initial score is slightly lower.

[0186] At the weighting level: If the model finds that the current “geographical distribution” weight deviates from the historical best solution in the same scenario by more than 0.1, it will automatically fine-tune Δw_i (e.g., from 0.15 to 0.2) and recalculate the S(m) score of all solutions.

[0187] After 3-5 rounds of iterative optimization, the system finally generates a list of recommended solutions sorted by comprehensive score. Each solution in the list contains complete structured information: topology type, primary and backup node list, link connection relationship, bandwidth allocation strategy, cost details, risk level and countermeasures, and comes with AI optimization instructions (such as "due to the characteristics of cross-city scenarios, the weight of dual-center topology is increased by 5%), providing a clear and reliable basis for administrator decision-making and subsequent mode switching.

[0188] In step S103, the optimal recommended configuration is displayed through a visual interface.

[0189] The topology diagram presents the optimal topology in a visual way: different colors / icons are used to distinguish core nodes (red) and edge nodes (blue), solid / dashed lines are used to mark main links and backup links, and key parameters (such as "dedicated line - latency 20ms - bandwidth 100Mbps") are marked next to the links, allowing administrators to understand the network architecture at a glance.

[0190] This disclosure collects multi-dimensional information such as basic site attributes, link performance, historical operational data, and business requirements at preset intervals. It then combines preset trigger conditions with weighted scoring and AI-assisted decision-making mechanisms to generate optimal topology and connection schemes, which are displayed through a visual interface. This enables comprehensive perception and dynamic response to network status, significantly improving the intelligence level of network topology decision-making. It avoids the rigidity of traditional static configurations or fixed templates, and generates optimal solutions adapted to the current network scenario based on real-time data. Simultaneously, the visual display facilitates administrators' intuitive understanding of configuration content, reducing manual analysis costs and operational errors, effectively improving network operation and maintenance efficiency and flexibility, ensuring business continuity in scenarios such as node changes and link anomalies, and ultimately achieving precise configuration and performance optimization of network resources.

[0191] In one embodiment, after displaying the optimal recommended configuration through a visual interface, the method further includes the following steps S104-S108:

[0192] S104. Upon receiving the change instruction, perform a simulated network configuration change according to the optimal recommended configuration.

[0193] Once the administrator confirms the optimal recommended configuration and issues a change instruction through the visual interface, the system initiates a simulated change process. This step does not directly modify the actual network configuration, but rather reproduces the current network's real-world state (including node connectivity, link parameters, and service traffic distribution) in a virtual environment, applying the optimal recommended configuration (such as topology adjustments and connection scheme updates) to this simulated environment. During the simulation, the system monitors the execution of configuration commands in real time, such as whether nodes have successfully connected to the new topology, whether link parameters have taken effect correctly, and whether routing policies conflict, thereby verifying the feasibility of the optimal configuration and avoiding risks to actual business operations from direct manipulation.

[0194] S105. After detecting a failure in the network configuration change, automatically roll back to the configuration before the simulated network configuration change.

[0195] If a configuration failure occurs during the simulation change process (such as nodes failing to recognize new topology instructions, link parameter configuration conflicts causing connection interruptions, routing policy loops, etc.), the system will immediately trigger an automatic rollback mechanism. The rollback operation restores the network state in the simulation environment to its initial state before the change, including node connection relationships, link parameters, routing rules, etc., all restored to their state before the simulation started. This design ensures that the simulation environment remains controllable, preventing failed configurations from accumulating in the simulation environment and affecting subsequent retrying or analysis, while also preventing misconfigurations from indirectly affecting the actual network.

[0196] S106. After the network configuration change is detected, the network SLA is automatically verified to determine whether the network meets the service requirements after the switch.

[0197] Once the simulated configuration changes are fully executed in the virtual environment (i.e., all topology adjustments and parameter configurations are completed without conflict or interruption), the system will automatically initiate a network SLA (Service Level Agreement) verification. Verification includes, but is not limited to, whether link bandwidth meets standards, whether latency and packet loss rate are within acceptable business limits, the stability of inter-node communication, and the transmission quality of critical services. These indicators are all tied to target business requirements (such as the low latency requirements for video conferencing and the high bandwidth requirements for large data transmission). By comparing the actual verification results with preset SLA thresholds, it is determined whether the simulated network topology can meet business operation requirements.

[0198] S107. If satisfied, switch directly to the optimal recommended configuration.

[0199] If the SLA verification results show that all network indicators after the simulated change meet or exceed the thresholds required by the business, it indicates that the optimal recommended configuration has practical application value. At this time, the system will automatically synchronize the optimal configuration in the simulated environment to the actual network, completing a full update of the topology, connection scheme, routing policies, etc. The entire switchover process does not require manual configuration item by item; automated commands ensure that the actual network is completely consistent with the optimal configuration, quickly realizing the upgrade or adjustment of the network mode.

[0200] S108. If not satisfied, automatically roll back to the configuration before the simulated network configuration change.

[0201] If the SLA verification results fail to meet business requirements (e.g., critical link latency exceeds limits, insufficient bandwidth causing service disruptions), it indicates that the optimal recommended configuration may not match actual business needs. In this case, the system will terminate the actual switchover process and automatically restore the simulated environment to its pre-change state, while maintaining the current configuration of the actual network. This operation prevents configurations that do not meet business requirements from being applied to the actual network, ensuring the continuity and stability of existing services and providing a foundation for subsequent re-optimization of the recommended configuration.

[0202] Build a closed loop for change security, and avoid network interruptions caused by changes through simulation verification, automatic rollback and SLA verification, so as to ensure the continuous stability of network and services.

[0203] Currently, in enterprise networking, SD-WAN, and wide area network environments, networking architecture and node connection relationships mostly adopt traditional solutions based on static planning and manual configuration.

[0204] The specific approach is as follows: During the initial network deployment phase, network administrators, based on business needs and network scale, manually specify the connection methods (such as star, full mesh, hub-and-sponge, point-to-point, etc.) between each site (CPE node) using network topology design tools or the SD-WAN console, and configure link parameters and routing policies one by one. Network changes or expansions using this approach typically rely on manual topology readjustment and manual configuration of new links. Some products allow semi-automatic configuration through predefined templates, but still require manual confirmation and adjustment of parameters such as connection relationships between sites, bandwidth allocation, and primary / backup links.

[0205] This approach has the following four drawbacks:

[0206] 1. Lack of intelligent and dynamic adaptive capabilities: Due to the use of static configuration, once the network experiences scenarios such as node expansion, link failure, sudden changes in traffic distribution, or dynamic changes in SLA, the system cannot identify the optimal connection and networking mode in real time. Administrators need to manually analyze and execute network reconstruction, which is slow and prone to errors.

[0207] This disclosure addresses this deficiency by collecting multi-dimensional data at preset intervals and combining it with a closed-loop mechanism of automatic triggering decision-making and AI dynamic optimization. The system continuously collects dynamic information such as basic site attributes, link performance parameters, historical operational data, and business requirements. It sets preset trigger conditions for scenarios such as node online / offline status, SLA decline, and link anomalies. Once these conditions are met, the analysis process is automatically initiated—without manual intervention. Candidate topologies are initially screened through weighted scoring, and then an AI model models network scenario characteristics, revising indicator weights and topology schemes multiple times to generate the optimal configuration in real time. This process achieves real-time network status awareness and automatic response, replacing the traditional static mode that relies on manual analysis, significantly improving response speed and reducing human error.

[0208] 2. Manual operation is complex and inefficient: In large-scale networking or multi-branch dynamic expansion scenarios, manual operation is complicated, the workload of operation and maintenance is large, the risk of misoperation and configuration omission is high, and it is very easy to affect network availability and business continuity.

[0209] This disclosure replaces tedious manual operations with full-process automation, eliminating the need for manual intervention from configuration generation to change implementation. The system automatically generates and visualizes the optimal topology and connection scheme through algorithms, and administrators only need to confirm the change instruction to initiate the subsequent process. The change phase supports simulated network configuration and has a built-in automatic rollback mechanism—if the configuration fails or the SLA verification fails, the system automatically restores to the original state without manual troubleshooting or repair. For large-scale network or expansion scenarios, there is no need to manually adjust the link parameters and routing policies of each node, completely simplifying the operation and maintenance process, reducing the risk of misoperation, and significantly improving configuration efficiency and network continuity.

[0210] 3. Fixed network topology and inflexible switching: If the network structure needs to be switched as a whole (such as from star to full mesh, or upgrading to multi-center redundancy), the topology needs to be completely reconstructed. The parameters of each node and link need to be manually configured repeatedly, which is complicated and there is no one-click switching or intelligent suggestion mechanism.

[0211] This disclosure breaks the static and fixed limitations of network topology, constructing a technical system for intelligent recommendation and flexible switching. Through candidate topology ranking and AI multi-round optimization, the system can automatically generate solutions for different modes such as star, full mesh, and multi-center redundancy, without requiring manual topology reconstruction. The switching process relies on simulated configuration and automatic verification mechanisms; the administrator can trigger mode switching after receiving the optimal recommendation, without repeatedly configuring node and link parameters. For example, when upgrading from Hub-Spoke mode to full mesh mode, the AI ​​model will adjust weights based on current network characteristics and generate an adaptation solution, enabling one-click, seamless mode switching with a visual interface, solving the pain point of complex traditional switching processes.

[0212] 4. Failure to consider the actual network operating status and node capability differences: This solution is mostly based on static business requirements, often ignoring the equipment capabilities, geographical distribution, real-time bandwidth load and historical health status of each site node, resulting in a deviation between the network topology and actual business needs, causing resource waste or performance bottlenecks.

[0213] This disclosure deeply integrates actual network operation data into the entire configuration decision-making process, achieving precise matching between the solution and actual needs. Multi-dimensional data collection covers the equipment capabilities, geographical distribution, real-time bandwidth load, and historical health status of each site, avoiding the limitations of relying solely on static business requirements. In the scoring stage, each indicator is quantified, and the weight values ​​are dynamically calculated through initial scenario weights, preference weights, and reinforcement learning adjustments, ensuring that differences in node capabilities, link loads, and other factors are fully considered. The AI ​​model also corrects the topology and weights based on the current network scenario feature vectors, and the final optimal solution can adapt to the performance differences of different nodes, avoiding waste or performance bottlenecks caused by resource mismatch.

[0214] Currently, some newer SD-WAN systems have introduced templated networking and policy-triggered connection switching.

[0215] Users can select fixed network topologies (such as "star topology", "Hub-Spoke mode", "dual-center mode", etc.) in the system and configure the connection methods between nodes, primary and backup routes, bandwidth allocation, and SLA policies according to certain rules. When a link anomaly occurs or the SLA fails to meet expectations, some systems can automatically switch traffic to the backup link or switch some network paths according to preset heuristic rules.

[0216] This approach has the following four drawbacks:

[0217] 1. Insufficient intelligence and rigid recommended solutions: The templates are fixed and the system lacks the ability to analyze real-time multi-dimensional data (node ​​capabilities, current bandwidth, geographical location, historical faults, business priorities, etc.), making it unable to provide truly optimal connection suggestions and network topology recommendations "tailored to local conditions".

[0218] This disclosure overcomes the limitations of fixed templates through dynamic perception and intelligent decision-making mechanisms. The system collects real-time data from multiple dimensions, including node capabilities, bandwidth status, geographical location, historical faults, and service priorities, at preset intervals, rather than relying on preset templates. Based on this data, a weighted scoring method is used to initially screen candidate topologies. Then, an AI-assisted decision-making mechanism models network scenario characteristics, and reinforcement learning is used to dynamically adjust indicator weights. After multiple rounds of refinement and optimization, the optimal topology and connection scheme adapted to the current network state is generated. This process achieves a shift from "fixed template application" to "adapting to local conditions based on real-time data," ensuring that the recommended scheme accurately matches actual network needs.

[0219] 2. Limited scope of automation: Although it supports partial primary / standby backup / traffic switching, it is difficult to achieve one-click mode switching and complete configuration synchronization across the entire network. Batch multi-node and large-scale structural changes still require manual intervention, and there is no transactional rollback mechanism during the switching process.

[0220] This disclosure expands the scope of automation, enabling network-wide mode switching and security management. The system not only automatically generates optimal configurations but also supports one-click switching of network modes based on these configurations, eliminating the need for manual adjustments to each node. During the switching process, feasibility is verified in advance by simulating network configuration changes, and an automatic rollback mechanism is built-in—if the configuration fails or the SLA verification fails, the system automatically reverts to the state before the change, forming a closed-loop transactional operation. For batch multi-node expansion or large-scale structural adjustments, no manual intervention is required throughout the process, completely solving the problem that automation in existing technologies is limited to switching only some links.

[0221] 3. Lack of integrated visualization and simulation assessment: Template changes / link switching are mostly black-box automatic behaviors, and operation and maintenance personnel cannot intuitively predict the specific impact of changes on business / traffic. Without simulation or historical comparison, risks cannot be identified in advance.

[0222] This disclosure deeply integrates visualization and simulation assessment into the network topology decision-making process. The optimal recommended configuration is intuitively displayed through a visual interface, allowing operations and maintenance personnel to clearly understand the topology, connection relationships, and key parameters. Before executing network changes, the system supports simulating the optimal configuration, presenting in real time the potential impact of the changes on service traffic, link load, and SLA metrics, while also comparing historical operational data to help assess risks. This design breaks the "black box operation" model, enabling operations and maintenance personnel to anticipate and mitigate risks in advance, thus improving the security of network changes.

[0223] 4. Low API integration capability: It is difficult to achieve deep integration with third-party orchestration systems and automated operation and maintenance platforms, which affects the scalability of automated operation and maintenance.

[0224] This disclosure includes standardized interfaces in its architectural design to support deep integration with third-party systems. Core functionalities such as optimal recommended configurations, topology data, and switching commands can be exported via APIs, facilitating access by third-party orchestration systems and automated operations and maintenance platforms. For example, external platforms can obtain real-time network status data, trigger intelligent recommendation processes, or execute mode switching commands through these interfaces, achieving automated cross-system collaboration. This design significantly enhances integration capabilities with the external ecosystem, providing a foundation for large-scale automated operations and maintenance expansion.

[0225] Figure 2 This is a flowchart illustrating a method for intelligent connection recommendation and dynamic networking mode switching according to an exemplary embodiment, such as... Figure 2 As shown, the method includes the following steps:

[0226] Event Trigger (S1): When events such as network node going online / offline, SLA decline, link anomaly, or administrator actively initiating configuration requests are detected, the intelligent recommendation process is automatically started.

[0227] Generate Recommendation Scheme (S2): The system calls the intelligent recommendation engine to generate the optimal networking mode and connection suggestions based on the current network status, business requirements and historical data, and form a recommended configuration.

[0228] User Confirmation (S3): The recommended solution is pushed to the management console, where the administrator can choose to execute it automatically, confirm manually, or ignore the suggestion. If confirmed, proceed to the next step; otherwise, the process ends.

[0229] Perform atomic switchover (S4): The system simulates the configuration change, assesses the risk, and saves a snapshot. Then, it atomically distributes the network configuration changes in batches, such as establishing / disconnecting tunnels, updating routes, adjusting bandwidth, and activating primary / standby operations. If any step fails, it can automatically roll back to the snapshot before the change.

[0230] Verify SLA (S5): After the switchover is completed, automatically verify the network SLA (such as latency, packet loss, bandwidth availability, etc.) to determine whether the network topology after the switchover meets the service requirements.

[0231] Complete the switchover (S6) or roll back to the snapshot (S7): If the verification is successful, the switchover is confirmed to be successful and the topology is updated; if the verification is unsuccessful, a rollback is triggered, all network changes are restored to the previous state, and the topology is updated again.

[0232] Topology information update (S8): After each switch or rollback, the system automatically refreshes the global topology view and operation and maintenance logs to ensure configuration consistency and traceability.

[0233] Through the above steps, this invention achieves real-time intelligent identification of network status, autonomous recommendation of optimal strategies, and automated switching with high security levels. It provides strong support for SD-WAN networking and operation and maintenance management in large-scale, dynamic, and heterogeneous environments, significantly reducing labor costs and business risks, and improving network resource utilization efficiency and overall business continuity.

[0234] The switching process begins with a full configuration simulation exercise in a sandbox or virtual environment. Recommended network configurations are automatically imported, and real business data flows are reproduced. Metrics such as route convergence, port connectivity, QoS policy delivery, and service availability are seamlessly detected, and a simulation effect report is automatically generated to assess the security of the implementation. Subsequently, the system predicts the impact on services based on the simulation evaluation results. Combining local business models, SLA requirements, and current traffic templates, it calculates changes in latency, jitter, and packet loss for critical services during the switching window and generates a list of affected objects and potential alerts for administrators to use in decision-making.

[0235] Before the official switch, the platform performs atomic snapshot backups of all current network configurations, topology tables, and policy forwarding information, and stores the snapshot files in a distributed and encrypted manner to ensure future process traceability and rapid rollback capabilities.

[0236] The following are examples of this disclosure in practical use scenarios.

[0237] Example 1:

[0238] During a company network deployment, a new branch site was built at headquarters. This site connects to the company's existing SD-WAN system via a Customer Premises Equipment (CPE). The system automatically collects information on the site's geographical location, port resources, CPE model, node type, bandwidth utilization, and other attributes, and compiles data on network link latency, historical health metrics, and costs. The system also monitors the geographical distribution, latency, bandwidth utilization, historical health metrics, and costs between this branch site and the company's various data centers. An intelligent recommendation algorithm combines these scoring parameters with current business needs and SLA requirements to calculate the comprehensive score for different network topologies (such as star, fully interconnected, and Hub-Spoke) under each metric, and automatically recommends the highest-scoring network topology. For example, for a star topology, a primary connection bandwidth allocation of 2Gbps is suggested. The system displays the recommended solution on the operations and maintenance management interface, clearly listing the reasons for the recommendation (e.g., lowest latency, optimal bandwidth utilization, high redundancy, best historical health metrics, optimal cost, and compliance with business needs and SLAs). After the administrator confirms with one click, the system will automatically complete the connection deployment, bandwidth allocation and route distribution, ensuring that the entire process is consistent with the above scoring indicators.

[0239] Example 2:

[0240] During the continuous operation of the enterprise SD-WAN network, the system monitored that the bandwidth utilization of a certain link consistently exceeded the threshold, and historical health indicators declined, while the corresponding branch point issued an SLA warning. The system immediately performed intelligent analysis and comprehensive scoring on parameters such as the current network topology, bandwidth utilization, geographical distribution, cost, redundancy, historical health indicators, business requirements, and SLA. If it was found that the existing model could not meet the demand, the system automatically popped up a recommendation to "upgrade to a multi-center redundant network topology." The recommendation included: adding redundant links to the new node for the specified branch node, optimizing the primary and backup nodes, balancing bandwidth utilization, improving redundancy, keeping costs under control, and reassessing the health of all links. After administrator confirmation, the system automatically configured the new network topology, primary and backup nodes, bandwidth allocation, and routing in batches according to the scoring results, and recorded all operations for traceability and secure rollback. The entire process always prioritized parameters such as latency, bandwidth utilization, cost, redundancy, historical health, SLA, and business requirements.

[0241] Example 3:

[0242] During routine enterprise network expansion, the operations and maintenance platform manages branch nodes in batches. The system automatically obtains attributes such as geographical location, node type, CPE model, port resources, and bandwidth utilization for each node via API interfaces. It also summarizes parameters such as connectivity, geographical distribution, latency, bandwidth utilization, cost, and redundancy with the existing network. Based on current business needs and global SLA requirements, the intelligent recommendation engine generates optimal networking strategies and network structure patterns for each node, automatically optimizing primary / backup node allocation, bandwidth utilization and allocation, cost control, redundancy design, and health assessment. It also explicitly outputs networking suggestions that meet all the aforementioned scoring parameters. The recommendation results are displayed through a visual interface, reviewed, and automatically distributed to each node in batches, ensuring seamless alignment across all scoring indicators and system parameters during the network expansion process.

[0243] In summary, this disclosure automatically collects multi-dimensional data on network nodes, links, and historical operations, combined with intelligent algorithms to achieve connection recommendations and dynamic switching of network topologies. This fundamentally enhances the intelligence, automation, and adaptability of enterprise networks. Compared to existing static templates or manually configured networking methods, this invention can respond more promptly and accurately to complex scenarios such as node expansion, link anomalies, traffic surges, and SLA changes, automatically providing optimal connection and networking suggestions. After user confirmation or policy triggering, mode switching is completed with a single click, significantly reducing maintenance difficulty and configuration error risks. The entire switching process ensures uninterrupted service and configuration consistency through atomic operations and rollback mechanisms. Simultaneously, the system visually displays topology changes and optimization effects, improving the controllability and transparency of operations. Through API integration and batch operation capabilities, it can adapt to network operation and maintenance needs in large-scale, complex, and heterogeneous environments, effectively improving network resource utilization, high availability, and service continuity. It also possesses continuous self-learning and optimization capabilities, demonstrating strong practical application value and commercial prospects.

[0244] Based on the same inventive concept, this application also provides an apparatus for implementing the above-described method of intelligent connection recommendation and dynamic networking mode switching. The solution provided by this apparatus is similar to the implementation described in the above-described method. Therefore, the specific limitations of one or more apparatus embodiments for intelligent connection recommendation and dynamic networking mode switching provided below can be found in the limitations of the method for intelligent connection recommendation and dynamic networking mode switching described above, and will not be repeated here.

[0245] In one exemplary embodiment, such as Figure 3 As shown, a device for intelligent connection recommendation and dynamic networking mode switching includes:

[0246] The data acquisition module is used to collect multi-dimensional data according to a preset period. The multi-dimensional data includes: basic attributes of each site, performance parameters of each link, historical network operation data, and target service requirements.

[0247] The intelligent recommendation engine module is used to generate the optimal recommendation configuration based on the multi-dimensional data when the preset trigger conditions are met, using a weighted scoring and AI-assisted decision-making mechanism. The optimal recommendation configuration includes: the optimal topology and the optimal connection scheme.

[0248] A visual interface is used to display the optimal recommended configuration.

[0249] In one embodiment, the preset triggering conditions include: network node coming online, network node going offline, SLA declining, link anomaly, or administrator actively initiating configuration requests.

[0250] In one embodiment, regarding the aspect of generating the optimal recommendation configuration based on the multi-dimensional data using a weighted scoring and AI-assisted decision-making mechanism, the intelligent recommendation engine is specifically used for:

[0251] Determine the scoring criteria for each candidate topology;

[0252] The scoring indicators are weighted to obtain a score for each candidate topology;

[0253] The candidate network topologies are sorted according to their scores, and the top N candidate network topologies are selected as the target topologies.

[0254] Generate connection schemes corresponding to each target topology;

[0255] The target topology and its corresponding connection scheme are sampled and optimized through an AI-assisted decision-making mechanism in multiple rounds to obtain the optimal topology and the optimal connection scheme.

[0256] In one embodiment, regarding the aspect of applying a weighted score to the scoring metric to obtain a score for each candidate topology, the intelligent recommendation engine is specifically used for:

[0257] Calculate a quantified value for each scoring metric for each candidate topology;

[0258] Obtain the weight value corresponding to each scoring indicator;

[0259] The score for each candidate topology is obtained by calculating the quantized value and weight value corresponding to each scoring metric for each candidate topology.

[0260] In one embodiment, regarding the acquisition of the weight value corresponding to each rating metric, the intelligent recommendation engine is specifically used for:

[0261] The initial weights of each scoring indicator are obtained based on the current network size and geographical distribution.

[0262] Obtain the preference weight values ​​for each input rating indicator;

[0263] Reinforcement learning is used to obtain the weight adjustment of each scoring indicator;

[0264] Based on the initial scenario weight, preference weight value, and weight adjustment amount of each scoring indicator, obtain the temporary weight value of each scoring indicator;

[0265] The temporary weight values ​​of each scoring indicator are normalized to obtain the corresponding weight value for each scoring indicator.

[0266] In one embodiment, regarding the aspect of performing multiple rounds of correction and optimization on the target topology and corresponding connection schemes using an AI-assisted decision-making mechanism to obtain the optimal topology and optimal connection scheme, the intelligent recommendation engine is specifically used for:

[0267] Simulate running each recommended configuration;

[0268] During simulation, obtain the weight values ​​of the scoring indicators to be adjusted in the target topology;

[0269] The score for each target topology is re-obtained based on the weight values ​​of the scoring indicators to be adjusted and the weight values ​​of the scoring indicators not adjusted in the target topology;

[0270] The scores for each reacquired target topology are reordered;

[0271] The target topology and corresponding connection schemes are sampled and optimized through multiple rounds of AI-assisted decision-making to obtain the optimal topology and optimal connection scheme.

[0272] In one embodiment, regarding the acquisition of the weight values ​​of the scoring metrics to be adjusted in the target topology, the intelligent recommendation engine is specifically used for:

[0273] Receive the weight values ​​of the rating indicators to be adjusted, input by the user;

[0274] or,

[0275] Based on historical performance and current network status changes, the weight values ​​of the scoring indicators to be adjusted in the target topology are automatically obtained.

[0276] In one embodiment, regarding the aspect of performing multiple rounds of correction and optimization on the reordered target topology and corresponding connection schemes using an AI-assisted decision-making mechanism to obtain the optimal topology and optimal connection scheme, the intelligent recommendation engine is specifically used for:

[0277] Obtain feature vectors of the current network scene characteristics through AI models;

[0278] Based on feature vectors, the AI ​​model performs multiple rounds of correction on the target topology and the corresponding index weights to obtain the optimal weights.

[0279] The score for each target topology is obtained again based on the corrected weight values;

[0280] The scores of each target topology obtained again are sorted again, and the target topology with the highest score is the optimal topology. The connection scheme corresponding to the optimal topology is the optimal connection scheme.

[0281] In one embodiment, a mode switching execution module is further included, specifically for:

[0282] Upon receiving the change instruction, simulate the network configuration change according to the optimal recommended configuration;

[0283] After detecting a failure to change the network configuration, the system will automatically roll back to the configuration before the network configuration change was simulated.

[0284] After the network configuration change is detected, the network SLA is automatically verified to determine whether the network meets the service requirements after the switch.

[0285] If satisfied, switch directly to the optimal recommended configuration;

[0286] If the conditions are not met, the configuration will be automatically rolled back to the state before the network configuration change.

[0287] In one embodiment, an API interface is also included, which facilitates integration with third-party platforms and batch operations for maintenance.

[0288] In the above embodiments, the intelligent recommendation engine module analyzes and processes the collected multi-dimensional data, and automatically generates one or more candidate network topologies (including but not limited to star topology, full interconnection, dual-center, etc.) by combining target business requirements and site changes. During the recommendation process, the system employs a weighted scoring mechanism, comprehensively considering factors such as bandwidth utilization, cost, geographical distribution, and historical health indicators to obtain a score for each model. This engine supports intelligent recommendation of the optimal network topology based on historical operating data, AI models, self-learning mechanisms, and administrator preferences.

[0289] The mode switching execution module is used to atomically complete the network mode switch based on the recommendation results, upon user confirmation or automatic policy triggering. The switch execution includes closed-loop steps such as configuration simulation exercise, business impact prediction, network snapshot saving, atomic changes and anomaly rollback, and automatically verifies the overall network health status after implementation.

[0290] The visual interface and API interface are used to recommend solutions and the switching process. The management interface can be visualized and displayed, and it supports business simulation comparison, solution highlighting and historical log traceability. At the same time, the system opens up an automated API interface to facilitate integration with third-party platforms and batch operation of maintenance.

[0291] This disclosure utilizes multi-source, real-time network status awareness and data fusion, along with intelligent scoring and self-learning algorithms, to automatically evaluate and dynamically recommend optimal connection schemes and networking modes. Upon detecting anomalies or policy triggers, it can automatically or with a single click switch to the optimal networking mode. The entire switching process is equipped with atomic operation and rollback mechanisms, as well as visualization and change simulation functions, ensuring uninterrupted service and consistent, controllable configuration during configuration changes. Furthermore, this invention supports batch operations and third-party system integration via APIs or automated scripts, adapting to business needs in complex and heterogeneous environments. The aforementioned multi-dimensional data collection, intelligent recommendation, automatic switching, atomic rollback, visual interaction, and automated integration technologies collectively constitute the unique innovations and key technical content requiring protection of this invention.

[0292] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 4 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a method and apparatus for intelligent connection recommendation and dynamic network mode switching.

[0293] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0294] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0295] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0296] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0297] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0298] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0299] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0300] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0301] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for intelligent connection recommendation and dynamic networking mode switching, characterized in that, The method includes: Multi-dimensional data is collected according to a preset cycle. The multi-dimensional data includes: basic attributes of each site, performance parameters of each link, historical network operation data, and target service requirements. When the preset triggering conditions are met, based on the multi-dimensional data, a weighted scoring and AI-assisted decision-making mechanism is used to generate the optimal recommended configuration, which includes: the optimal topology and the optimal connection scheme. The optimal recommended configuration is displayed through a visual interface; Based on the multi-dimensional data, a weighted scoring and AI-assisted decision-making mechanism is used to generate the optimal recommended configuration, including: Determine the scoring criteria for each candidate topology; The scoring indicators are weighted to obtain a score for each candidate topology; The candidate network topologies are sorted according to their scores, and the top N candidate network topologies are selected as the target topologies. Generate connection schemes corresponding to each target topology; The target topology and its corresponding connection scheme are sampled and optimized through an AI-assisted decision-making mechanism in multiple rounds to obtain the optimal topology and the optimal connection scheme. The AI-assisted decision-making mechanism for sampling the target topology and its corresponding connection schemes undergoes multiple rounds of correction and optimization to obtain the optimal topology structure and optimal connection scheme, including: Simulate running each recommended configuration; During simulation, obtain the weight values ​​of the scoring indicators to be adjusted in the target topology; The score for each target topology is re-obtained based on the weight values ​​of the scoring indicators to be adjusted and the weight values ​​of the scoring indicators not adjusted in the target topology; The scores for each reacquired target topology are reordered; The AI-assisted decision-making mechanism is used to perform multiple rounds of correction and optimization on the reordered target topology and the corresponding connection scheme to obtain the optimal topology and the optimal connection scheme. The AI-assisted decision-making mechanism performs multiple rounds of correction and optimization on the reordered target topology and corresponding connection schemes to obtain the optimal topology and optimal connection scheme, including: Obtain feature vectors of the current network scene characteristics through AI models; Based on feature vectors, the AI ​​model performs multiple rounds of correction on the target topology and the corresponding index weights to obtain the optimal weights. The score for each target topology is obtained again based on the corrected weight values; The scores of each target topology obtained again are sorted again, and the target topology with the highest score is the optimal topology. The connection scheme corresponding to the optimal topology is the optimal connection scheme.

2. The method according to claim 1, characterized in that, The preset triggering conditions include: network node coming online, network node going offline, SLA declining, link anomaly, or administrator actively initiating configuration requests.

3. The method according to claim 1, characterized in that, The weighted scoring of the scoring indicators to obtain a score for each candidate topology includes: Calculate a quantified value for each scoring metric for each candidate topology; Obtain the weight value corresponding to each scoring indicator; The score for each candidate topology is obtained by calculating the quantized value and weight value corresponding to each scoring metric for each candidate topology.

4. The method according to claim 3, characterized in that, The step of obtaining the weight value corresponding to each scoring indicator includes: The initial weights of each scoring indicator are obtained based on the current network size and geographical distribution. Obtain the preference weight values ​​for each input rating indicator; Reinforcement learning is used to obtain the weight adjustment of each scoring indicator; Based on the initial scenario weight, preference weight value, and weight adjustment amount of each scoring indicator, obtain the temporary weight value of each scoring indicator; The temporary weight values ​​of each scoring indicator are normalized to obtain the corresponding weight value for each scoring indicator.

5. The method according to claim 1, characterized in that, The process of obtaining the weight values ​​of the scoring indicators to be adjusted in the target topology includes: Receive the weight values ​​of the rating indicators to be adjusted, input by the user; or, Based on historical performance and current network status changes, the weight values ​​of the scoring indicators to be adjusted in the target topology are automatically obtained.

6. The method according to claim 1, characterized in that, After displaying the optimal recommended configuration through a visual interface, the method further includes: Upon receiving the change instruction, simulate the network configuration change according to the optimal recommended configuration; After detecting a failure to change the network configuration, the system will automatically roll back to the configuration before the network configuration change was simulated. After the network configuration change is detected, the network SLA is automatically verified to determine whether the network meets the service requirements after the switch. If satisfied, switch directly to the optimal recommended configuration; If the conditions are not met, the configuration will be automatically rolled back to the state before the network configuration change.

7. A device for intelligent connection recommendation and dynamic networking mode switching, characterized in that, The device includes: The data acquisition module is used to collect multi-dimensional data according to a preset period. The multi-dimensional data includes: basic attributes of each site, performance parameters of each link, historical network operation data, and target service requirements. The intelligent recommendation engine module is used to generate the optimal recommendation configuration based on the multi-dimensional data when the preset trigger conditions are met, using a weighted scoring and AI-assisted decision-making mechanism. The optimal recommendation configuration includes: the optimal topology and the optimal connection scheme. A visual interface is provided to display the optimal recommended configuration. In terms of generating the optimal recommendation configuration based on the multi-dimensional data, using weighted scoring and AI-assisted decision-making mechanisms, the intelligent recommendation engine module is specifically used for: Determine the scoring criteria for each candidate topology; The scoring indicators are weighted to obtain a score for each candidate topology; The candidate network topologies are sorted according to their scores, and the top N candidate network topologies are selected as the target topologies. Generate connection schemes corresponding to each target topology; The target topology and its corresponding connection scheme are sampled and optimized through an AI-assisted decision-making mechanism in multiple rounds to obtain the optimal topology and the optimal connection scheme. In terms of performing multiple rounds of correction and optimization on the target topology and corresponding connection schemes using an AI-assisted decision-making mechanism to obtain the optimal topology and optimal connection schemes, the intelligent recommendation engine module is specifically used for: Simulate running each recommended configuration; During simulation, obtain the weight values ​​of the scoring indicators to be adjusted in the target topology; The score for each target topology is re-obtained based on the weight values ​​of the scoring indicators to be adjusted and the weight values ​​of the scoring indicators not adjusted in the target topology; The scores for each reacquired target topology are reordered; The AI-assisted decision-making mechanism is used to perform multiple rounds of correction and optimization on the reordered target topology and the corresponding connection scheme to obtain the optimal topology and the optimal connection scheme. In the aspect of performing multiple rounds of correction and optimization on the reordered target topology and corresponding connection schemes using an AI-assisted decision-making mechanism to obtain the optimal topology and optimal connection scheme, the intelligent recommendation engine module is specifically used for: Obtain feature vectors of the current network scene characteristics through AI models; Based on feature vectors, the AI ​​model performs multiple rounds of correction on the target topology and the corresponding index weights to obtain the optimal weights. The score for each target topology is obtained again based on the corrected weight values; The scores of each target topology obtained again are sorted again, and the target topology with the highest score is the optimal topology. The connection scheme corresponding to the optimal topology is the optimal connection scheme.

Images