A multi-link intelligent routing method and system for financial transactions

Abstract

This application discloses a multi-link intelligent routing method and system for financial transactions, relating to the field of communication network technology. The method includes: extracting the transaction type, transaction amount, real-time requirements, and data sensitivity characteristics of financial transactions; classifying service types and assigning transmission priorities; deploying an SRv6 network and collecting link status parameters using a combination of periodic monitoring and event-triggered monitoring; constructing an evaluation system based on the analytic hierarchy process (AHP); assigning differentiated weights to indicators such as link latency, packet loss rate, and bandwidth utilization according to transmission priority; selecting the optimal transmission path; configuring a hybrid queue scheduling strategy of priority scheduling followed by weighted fair scheduling; encapsulating the segment identifier sequence corresponding to the optimal path into the segment routing header; and writing the transmission priority into the traffic category field before transmission. This application achieves dynamic adaptation between financial transaction service awareness and network resource scheduling, improving the transmission quality and network resource utilization of core transactions.

Description

Technical Field

[0001] This invention relates to the field of communication network technology, and more specifically, to a multi-link intelligent routing method and system for financial transactions. Background Technology

[0002] With the comprehensive digital transformation of financial services, core transactions such as online banking, mobile payment, and real-time settlement have placed stringent demands on the transmission quality, reliability, and intelligence level of financial backbone networks. To address these challenges, the industry has begun to explore combining IPv6 segment routing traffic engineering policies (SRv6 TE Policy) with software-defined networking (SDN) technology to achieve more flexible and efficient traffic scheduling.

[0003] A typical packet forwarding method (for example, see Chinese patent application CN202511398333.9) includes: a head node receiving a service packet, obtaining a target service attribute matching the packet flow characteristics, and obtaining at least one Intelligent Policy Routing (IPR) template matching the destination address; then selecting a target IPR template matching the target service attribute from the at least one IPR template, and obtaining the target quality conditions of the template; furthermore, selecting an SRv6 TE Policy that meets the target quality conditions and has the highest link priority from the target IPR template; and finally forwarding the service packet through the candidate paths included in the SRv6 TE Policy. In addition, this method periodically detects the path quality parameters of each candidate path and performs path switching or backswitching when conditions are met to avoid affecting service quality due to deteriorated forwarding path quality.

[0004] The aforementioned existing technologies improve the quality of service for packet forwarding to some extent by matching corresponding IPR templates to different business attributes and selecting the optimal forwarding path based on path quality conditions. However, this method still has the following technical shortcomings when facing the specific scenario of financial transactions:

[0005] First, this method has a coarse-grained approach to identifying business attributes. Its business attributes are typically mapped based on packet header information such as 5-tuples or Differential Service Code Points (DSCPs), failing to extract the deep characteristics unique to financial transactions, such as transaction amount, real-time requirements, and data sensitivity. This results in the transmission priority allocated to different financial transactions not truly reflecting their business value and service quality requirements. For example, a large real-time transfer and a small, ordinary query might be mapped to the same business attribute, thus matching the same IPR template. This prevents core transactions from receiving differentiated priority protection, making them vulnerable to bandwidth consumption by ordinary services.

[0006] Secondly, the path selection mechanism of this method lacks dynamic adaptation to service priorities. Although its IPR template includes quality conditions corresponding to service attributes (such as bandwidth utilization limits) and uses link priority as the basis for selecting SRv6 TE Policy, this link priority is statically configured and cannot be dynamically adjusted according to real-time network conditions (such as congestion levels and link quality fluctuations) and the actual service quality requirements of the service. When network conditions change, the static priority configuration cannot ensure that core transactions always obtain the optimal transmission path, nor can it proactively adjust low-priority services to idle paths to ensure the transmission stability of core services when network congestion occurs.

[0007] Therefore, how to provide a multi-link intelligent routing method and system that is oriented towards financial transactions, can perform refined business perception based on the deep characteristics of financial transactions, and can realize dynamic adaptation of business priority and path selection and resource scheduling has become a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0008] This application provides a multi-link intelligent routing method and system for financial transactions to solve the technical problems in the prior art, such as coarse business perception granularity, inability to perform refined business perception based on deep characteristics of financial transactions, lack of dynamic adaptation between path selection mechanism and business priority, and difficulty in ensuring the quality of core transaction services.

[0009] Firstly, this application provides a multi-link intelligent routing method for financial transactions, the method comprising:

[0010] Acquire financial transaction data to be transmitted, and extract transaction feature information from the financial transaction data, including transaction type, transaction amount, real-time requirements, and data sensitivity;

[0011] The financial transactions are divided into multiple business types based on the transaction feature information, and each business type is assigned a corresponding transmission priority, service quality requirement parameters, and data sensitivity level.

[0012] Deploy an SRv6 network and use a combination of periodic monitoring and event-triggered monitoring to collect link status parameters of each SRv6 path and each path segment in real time. The monitoring period is dynamically adjusted according to the transmission priority of the service type.

[0013] Candidate paths that meet the service quality requirements and data sensitivity levels are selected based on the service type and the link status parameters.

[0014] A path selection evaluation system is constructed based on the analytic hierarchy process (AHP). The path selection evaluation system includes multiple evaluation indicators such as link latency, packet loss rate, bandwidth utilization, link reliability, and security level. Differentiated weights are assigned to the multiple evaluation indicators according to the transmission priority of the service type. The comprehensive evaluation score of each candidate path is calculated, and the path with the highest comprehensive evaluation score is selected as the optimal transmission path.

[0015] Configure a hybrid queue scheduling strategy on the SRv6 source node. The hybrid queue scheduling strategy includes multiple priority queues corresponding to the transmission priority. The logic of priority scheduling followed by weighted fair scheduling is used to perform differentiated queue scheduling for financial transactions with different priorities.

[0016] The segment identifier sequence corresponding to the optimal transmission path is encapsulated into the segment routing header of the IPv6 header of the financial transaction data, and a traffic category field is added to the IPv6 header. The transmission priority is written into the traffic category field before transmission.

[0017] In one possible implementation, the method of combining periodic monitoring and event-triggered monitoring to collect link status parameters of each SRv6 path and each path segment in real time includes: collecting link status parameters of financial transactions corresponding to a first transmission priority in a first cycle, and collecting link status parameters of financial transactions corresponding to a second transmission priority in a second cycle, wherein the first transmission priority is higher than the second transmission priority, and the first cycle is shorter than the second cycle; when a sudden change in link status parameters is detected to exceed a preset threshold, event monitoring is triggered, and the monitoring frequency is increased to a third cycle, wherein the third cycle is shorter than the first cycle.

[0018] In one possible implementation, the path selection evaluation system based on the analytic hierarchy process (AHP) includes: establishing a hierarchical model comprising a target layer, a criterion layer, and a scheme layer, wherein the target layer is for optimal path selection, the criterion layer includes the multiple evaluation indicators, and the scheme layer includes the candidate paths; based on the transmission priority of the service type, a pairwise importance scale between each evaluation indicator is preset, a judgment matrix is ​​constructed based on the scale, the maximum eigenvalue and consistency ratio of the judgment matrix are calculated, and if the consistency ratio is less than a preset threshold, the consistency test is passed; the weights of each evaluation indicator are calculated based on the judgment matrix that has passed the consistency test, and the normalized parameter values ​​of each candidate path under each evaluation indicator are weighted and summed with the corresponding weights to obtain the comprehensive evaluation score of each candidate path.

[0019] In one possible implementation, the differential queue scheduling of financial transactions with different priorities using the logic of priority scheduling followed by weighted fair scheduling includes: setting up multiple priority queues corresponding to the transmission priorities, prioritizing the scheduling of the highest priority queue until that queue is empty, and then sequentially scheduling the next highest priority queue; assigning queue-level bandwidth weights to each priority queue; dividing the financial transaction data in each priority queue into multiple data streams according to 5-tuples, allocating bandwidth shares to each data stream proportionally within the bandwidth weight range of the queue, and scheduling each data stream using a weighted round-robin method.

[0020] In one possible implementation, the method further includes: when network status fluctuations are detected, determining the type of network fluctuation, and setting differentiated fluctuation response rules for financial transactions of different priorities based on the transmission priority.

[0021] In one possible implementation, the network fluctuation types include bandwidth reduction, increased node response latency, packet loss, and network congestion. The step of setting differentiated fluctuation response rules for financial transactions of different priorities based on the transmission priority includes: if the network fluctuation type is bandwidth reduction, then according to descending priority, locking dedicated bandwidth, compressing bandwidth ratio, and suspending bandwidth allocation for the corresponding financial transactions; if the network fluctuation type is increased node response latency, then according to descending priority, switching to backup nodes, increasing the number of verification nodes, extending processing time, and suspending processing for the corresponding financial transactions; if the network fluctuation type is packet loss, then according to descending priority, enabling redundant transmission, shortening retransmission interval, retransmission, and pending retransmission for the corresponding financial transactions; if the network fluctuation type is network congestion, then allocating resources from transactions below a preset level to transactions above the preset level, and restoring resource allocation according to a preset order after congestion is relieved.

[0022] In one possible implementation, the step of encapsulating the segment identifier sequence corresponding to the optimal transmission path into the segment routing header of the IPv6 header of the financial transaction data, adding a traffic category field to the IPv6 header, and writing the transmission priority into the traffic category field before transmission includes: sequentially encapsulating the segment identifier sequence corresponding to the optimal transmission path into the segment routing header of the IPv6 header of the financial transaction data, wherein the segment routing header includes a segment identifier list and a segment left shift pointer; adding a traffic category field to the IPv6 header, and writing the transmission priority into the traffic category field so that forwarding nodes can identify the priority and perform corresponding queue scheduling.

[0023] Secondly, this application provides a multi-link intelligent routing system for financial transactions, the system comprising:

[0024] The business awareness module is configured to acquire financial transaction data to be transmitted, extract transaction feature information of the financial transaction data, including transaction type, transaction amount, real-time requirements and data sensitivity; and classify the financial transaction into multiple business types according to the transaction feature information, and assign corresponding transmission priority, service quality requirement parameters and data sensitivity level to each business type.

[0025] The link monitoring module is configured to deploy an SRv6 network and uses a combination of periodic monitoring and event-triggered monitoring to collect link status parameters of each SRv6 path and each path segment in real time. The monitoring period is dynamically adjusted according to the transmission priority of the service type.

[0026] The path optimization module is configured to filter candidate paths that meet the service quality requirements and data sensitivity levels based on the service type and the link status parameters; construct a path optimization evaluation system based on the analytic hierarchy process (AHP), which includes multiple evaluation indicators such as link latency, packet loss rate, bandwidth utilization, link reliability, and security level; assign differentiated weights to the multiple evaluation indicators according to the transmission priority of the service type; calculate the comprehensive evaluation score of each candidate path; and select the path with the highest comprehensive evaluation score as the optimal transmission path.

[0027] The queue scheduling module, deployed on the SRv6 source node, includes multiple priority queues corresponding to the transmission priority, and is configured to perform differentiated queue scheduling for financial transactions of different priorities using a logic of priority scheduling followed by weighted fair scheduling.

[0028] The encapsulation and transmission module is configured to encapsulate the segment identifier sequence corresponding to the optimal transmission path into the segment routing header of the IPv6 header of the financial transaction data, add a traffic category field to the IPv6 header, and transmit the data after writing the transmission priority into the traffic category field.

[0029] In one possible implementation, the link monitoring module includes: a path segment monitoring unit configured to collect the port status and forwarding latency of the forwarding nodes corresponding to each path segment in each SRv6 path; a period adjustment unit configured to dynamically adjust the monitoring period according to the transmission priority of the service type, wherein the higher the transmission priority, the shorter the corresponding monitoring period; and an event triggering unit configured to increase the monitoring frequency to a preset high-frequency period when a sudden change in link status parameters is detected to exceed a preset threshold.

[0030] In one possible implementation, the path optimization module includes: a hierarchical analysis unit configured to establish a hierarchical structure model including a target layer, a criterion layer, and a scheme layer; to preset a pairwise importance scale between each evaluation indicator based on the transmission priority of the service type; to construct a judgment matrix based on the scale; to calculate the maximum eigenvalue and consistency ratio of the judgment matrix; and to pass the consistency test if the consistency ratio is less than a preset threshold; a weight allocation unit configured to calculate the weight of each evaluation indicator based on the judgment matrix that has passed the consistency test; and a scoring calculation unit configured to sum the normalized parameter values ​​of each candidate path under each evaluation indicator with the corresponding weight to obtain the comprehensive evaluation score of each candidate path.

[0031] In one possible implementation, the queue scheduling module includes: a priority scheduling unit configured to set multiple priority queues corresponding to the transmission priority, prioritize scheduling the highest priority queue, and then schedule the next highest priority queue in turn when the queue is empty; a fair scheduling unit configured to assign queue-level bandwidth weights to each priority queue; divide the financial transaction data in each priority queue into multiple data streams according to quintuples, allocate bandwidth shares to each data stream proportionally within the bandwidth weight range of the queue, and schedule each data stream using a weighted round-robin method.

[0032] In one possible implementation, the encapsulation transmission module includes: a path encapsulation unit configured to sequentially encapsulate the segment identifier sequence corresponding to the optimal transmission path into the segment routing header of the IPv6 header of the financial transaction data, wherein the segment routing header includes a segment identifier list and a segment left shift pointer; and a priority marking unit configured to add a traffic category field to the IPv6 header and write the transmission priority into the traffic category field so that forwarding nodes can identify the priority and perform corresponding queue scheduling.

[0033] In one possible implementation, the system further includes a fluctuation response module configured to determine the type of network fluctuation when network status fluctuations are detected, and to set differentiated fluctuation response rules for financial transactions of different priorities based on the transmission priority.

[0034] Compared with the prior art, this application has the following beneficial effects:

[0035] I. Implement business classification and prioritization based on deep characteristics of financial transactions

[0036] Existing technologies rely on message header information such as 5-tuples or differential service code points to identify business attributes. Their classification is limited to attributes at the message transmission level and cannot distinguish the intrinsic value differences between different business transactions in financial transactions. For example, a large real-time transfer and a small account query may have the same protocol type and port number at the message header level, and existing technologies map them to the same business attribute.

[0037] This application extracts feature information from four dimensions: transaction type, transaction amount, real-time requirements, and data sensitivity. Based on this, financial transactions are divided into core transaction categories, important business categories, and ordinary business categories, and corresponding transmission priorities, service quality requirements, and data sensitivity levels are assigned to each category. Through these technical means, the business value of financial transactions can be quantified into calculable priority parameters, providing a more refined business awareness foundation for subsequent path selection, link monitoring frequency adjustment, and queue scheduling.

[0038] II. Establish a link status monitoring mechanism linked to business priorities.

[0039] Existing link status monitoring technologies use a fixed-period approach, applying the same monitoring frequency to all services. Under this method, core transactions and ordinary services receive the same monitoring accuracy, failing to reflect differences in service priority.

[0040] This application employs a combined approach of periodic monitoring and event-triggered monitoring. The monitoring cycle is dynamically adjusted based on the transmission priority of the service type: the higher the transmission priority, the shorter the corresponding monitoring cycle. Simultaneously, when a sudden change in link status parameters exceeds a preset threshold, event monitoring is triggered, increasing the monitoring frequency to an even higher level. Through these technical means, core transactions can achieve a higher monitoring frequency, enabling faster responses to changes in link status. Furthermore, the event-triggered mechanism can promptly capture status changes when the link is abnormal, providing data support for subsequent path adjustments.

[0041] III. Implementing Dynamic Weighted Path Optimization Based on Business Priority

[0042] Existing technologies typically rely on a single metric or multiple metrics with fixed weights for path selection. Under the fixed-weight model, different business types share the same optimization objective, failing to reflect differences in business priorities.

[0043] This application constructs a path optimization system based on the Analytic Hierarchy Process (AHP), incorporating five evaluation indicators: link latency, packet loss rate, bandwidth utilization, link reliability, and security level. Differential weights are assigned to these five indicators according to the transmission priority of the service type. The comprehensive evaluation score of each candidate path is calculated, and the path with the highest score is selected as the optimal transmission path. Through these techniques, services of different priorities can obtain routing strategies that match their needs: high-priority services receive higher weights for latency and packet loss rate indicators during routing, while low-priority services receive higher weights for bandwidth utilization indicators.

[0044] IV. Achieving a balance between priority protection and bandwidth fairness in scheduling

[0045] There are two typical schemes for queue scheduling in the existing technology: single priority queue scheduling may cause packets in low priority queues to be unserviced for a long time; single weighted fair queue scheduling cannot guarantee low-latency transmission of high priority packets.

[0046] This application employs a two-level scheduling logic: priority scheduling followed by weighted fair scheduling. The first level sets up multiple priority queues corresponding to transmission priorities, prioritizing the scheduling of the highest-priority queue until it is empty, after which the next highest-priority queues are scheduled in turn. The second level assigns bandwidth weights to each priority queue, dividing the transaction data within each priority queue into multiple data streams using five-tuples. Each data stream is allocated a bandwidth share corresponding to its queue bandwidth weight, and a weighted round-robin method is used to schedule each data stream. Through these technical means, both priority transmission of high-priority services and fair bandwidth allocation among data streams within the same priority queue are ensured.

[0047] V. Implement hop-by-hop transmission of priority information on the SRv6 path

[0048] Existing technologies lack the means to combine business priorities with the SRv6 path encapsulation depth, and the priority decision of the source node cannot be passed to the intermediate forwarding node.

[0049] This application encapsulates the segment identifier sequence corresponding to the optimal transmission path into the segmentation routing header of the IPv6 header of financial transaction data, achieving precise path control. Simultaneously, a traffic category field is added to the IPv6 header, and the transmission priority is written into this field. Through these technical means, priority information can be transmitted hop-by-hop along the transmission path with the packet, and each forwarding node can perform corresponding queue scheduling based on the priority after parsing the traffic category field.

[0050] VI. Overall Synergistic Effect

[0051] The aforementioned technical features of this application do not exist in isolation, but rather form a cohesive whole through explicit input-output relationships and logical progression, specifically manifested in the following five aspects:

[0052] (i) Synergy between business awareness and link monitoring

[0053] The transmission priority assigned in the business awareness step serves as the basis for dynamically adjusting the monitoring period in the link monitoring step. Specifically, core transactions have the highest transmission priority, and the link monitoring step allocates them with the shortest monitoring period (e.g., 100ms) to achieve higher link status awareness accuracy; ordinary transactions have lower transmission priority, and the link monitoring step allocates them with a longer monitoring period (e.g., 500ms) to avoid wasting monitoring resources. Through this coordination, monitoring resources are allocated differentiatedly according to business priority.

[0054] (II) Synergy between Link Monitoring and Path Optimization

[0055] The link status parameters collected during the link monitoring step, including latency, packet loss rate, bandwidth utilization, link reliability, and security level for each SRv6 path and path segment, serve as input data for the analytic hierarchy process (AHP) evaluation system in the path optimization step. The path optimization step calculates the comprehensive evaluation score for each candidate path based on the real-time collected link status parameters, rather than relying on static configurations or historical data. Through this collaboration, the path optimization decision reflects the current real-time network status, improving the accuracy and timeliness of routing results.

[0056] (III) Collaboration between business perception and path optimization

[0057] The transmission priority defined in the business awareness step serves as the basis for the weight allocation of the analytic hierarchy process (AHP) in the path optimization step. Specifically, for core transactions, the path optimization step assigns higher weights (30% each) to latency and packet loss rate, prioritizing paths with lower latency and packet loss rates during routing. For general services, the path optimization step assigns higher weights to bandwidth utilization, prioritizing paths with ample bandwidth. Through this coordination, the path optimization objectives are aligned with the actual needs of the business.

[0058] (iv) Coordination between path optimization and queue scheduling

[0059] The optimal transmission path selected in the path optimization step serves as the forwarding path for packets after the queue scheduling step. The queue scheduling step, based on the transmission priority determined in the service awareness step, sends packets to the corresponding priority queue for scheduling. After scheduling, packets are forwarded according to the path determined in the path optimization step. Through this coordination, the result of path optimization can be implemented through the queue scheduling mechanism, forming a closed loop between routing decisions and forwarding execution.

[0060] (v) Coordination between service awareness and encapsulated transmission

[0061] The transmission priority identified in the service awareness step is written into the traffic category field of the IPv6 header through the encapsulation transmission step. The encapsulation transmission step determines the corresponding traffic category value based on the transmission priority (e.g., 46 for P1, 26 for P2, and 0 for P3) and writes this value into the traffic category field. Through this coordination, the priority decision of the source node can be passed hop-by-hop along the transmission path with the packet. After parsing the traffic category field, each forwarding node can perform corresponding queue scheduling according to the priority, forming end-to-end quality of service coordination.

[0062] Overall synergistic technical effects

[0063] Through the synergistic effect of the above five levels, the technical solution of this application forms a complete processing flow from service classification, link monitoring, path optimization, queue scheduling to encapsulated transmission. The input and output relationships between each step are clear, achieving the following overall technical effects:

[0064] First, business priorities are applied throughout the entire process. The transmission priorities assigned in the business awareness step successively affect the monitoring frequency of link monitoring, the weight allocation of path optimization, the queue selection of queue scheduling, and the traffic category labeling of encapsulated transmission, thus achieving consistent transmission and application of business priorities throughout the entire process from awareness to transmission.

[0065] Secondly, real-time status-driven decision-making. The real-time link status parameters collected during the link monitoring step serve as input data for the path optimization step, enabling route selection decisions to be based on the current network status rather than static configuration, thus improving the timeliness and accuracy of the decisions.

[0066] Third, the decision-making and execution closed loop. The path selection step determines the optimal path, the queue scheduling step schedules according to priority, and the encapsulation and transmission step encapsulates the path and priority information before sending it, forming a complete closed loop from decision-making to execution, ensuring that the routing results can be executed accurately.

[0067] Fourth, end-to-end quality of service coordination. The encapsulation and transmission steps write priority information into the traffic category field of the IPv6 header, enabling priority decisions to be passed hop-by-hop along the transmission path. Each forwarding node performs corresponding queue scheduling based on this information, achieving end-to-end quality of service coordination from the source node to the destination node. Attached Figure Description

[0068] Figure 1 This application provides an overall flowchart of a multi-link intelligent routing method for financial transactions.

[0069] Figure 2 A flowchart illustrating the SRv6 path segment-level monitoring mechanism provided in this application embodiment;

[0070] Figure 3A schematic diagram of the path optimization process based on the analytic hierarchy process provided in this application embodiment;

[0071] Figure 4 A flowchart illustrating the WFQ+PQ hybrid queue scheduling mechanism provided in this application embodiment;

[0072] Figure 5 This is a schematic diagram of the architecture of a multi-link intelligent routing system for financial transactions provided in an embodiment of this application. Detailed Implementation

[0073] To make the objectives, technical solutions, and advantages of this application clearer, the application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0074] I. Definitions

[0075] 1. SRv6 (Segment Routing over IPv6): A segmented routing technology based on IPv6. It adopts a header node path selection mechanism, pre-encapsulating the segment identifiers of the segments the path will pass through in the header node. When a packet passes through an SR node, the SR node forwards the packet according to the segment identifiers of the packet.

[0076] 2. Segment Identifier (SID): A path segment identifier in an SRv6 network, used to uniquely identify a forwarding instruction or a forwarding path segment.

[0077] 3. Segment Routing Header (SRH): An IPv6 extension header used to carry information such as a segment identifier list and a segment left shift pointer, enabling precise path control.

[0078] 4. Traffic Class (TC): A field in the IPv6 header used to identify the priority and quality of service class of a packet, which can be used by forwarding nodes to identify and perform corresponding queue scheduling.

[0079] 5. Analytic Hierarchy Process (AHP): A multi-objective decision analysis method that constructs a hierarchical structure model and judgment matrix, calculates the weights of each evaluation index, and then comprehensively evaluates the alternative solutions.

[0080] 6. Weighted Fair Queuing (WFQ): A queue scheduling algorithm that assigns bandwidth weights to each queue and schedules packets in each queue according to the weight ratio to achieve fair bandwidth allocation.

[0081] 7. Priority Queuing (PQ): A queue scheduling algorithm that sets up multiple priority queues. The scheduler strictly serves each queue in priority order, and packets in the higher priority queue are scheduled first.

[0082] 8. Five-tuple: refers to the source IP address, destination IP address, source port number, destination port number, and protocol type, used to uniquely identify a data stream.

[0083] 9. Deep Packet Inspection (DPI): A network packet inspection technology that identifies service types by parsing the packet payload.

[0084] II. Method Examples

[0085] This embodiment provides a multi-link intelligent routing method for financial transactions, applied to the backbone network scenario of financial institutions. For example... Figure 1 As shown, the method includes the following steps:

[0086] Step S1: Classification and Prioritization of Financial Transaction Businesses

[0087] Collect financial transaction data to be transmitted and extract transaction characteristic information such as transaction type, transaction amount, real-time requirements, and data sensitivity.

[0088] Transaction data originates from various business messages within the financial institution's bus system, including standardized transaction messages sent from various business channels of the financial institution. Taking a commercial bank as an example, its business channels include the counter system, mobile banking APP, online banking system, and third-party payment interface. The transaction messages sent from each channel comply with the financial industry's ISO8583 or SWIFT message specifications. The message header contains a transaction type code (e.g., "101" represents real-time transfer, "202" represents fund redemption, and "303" represents balance inquiry), and the message body contains a transaction amount field (e.g., "5,000,000" represents 5 million yuan), a transaction real-time indicator (e.g., "RT" represents real-time, "BT" represents batch), and a data sensitivity indicator (e.g., "H" represents high sensitivity, and "L" represents low sensitivity).

[0089] Based on the aforementioned characteristics, financial transactions are categorized into three business types: core transactions (including real-time transfers and payment settlements), important transactions (including batch clearing and fund redemptions), and general transactions (including online banking inquiries and account loss reporting). Each business type is assigned a corresponding transmission priority (P1 highest, P2 medium, P3 lowest) and QoS requirements (latency threshold, packet loss rate threshold, bandwidth requirements). Simultaneously, data sensitivity levels (high, medium, low) are marked, and corresponding link security level requirements (encrypted transmission, authentication, and standard protection) are clearly defined.

[0090] The specific calibration standards are as follows:

[0091]

[0092] Taking a real-time interbank transfer transaction of 5 million yuan as an example, its transaction type code is "101", the transaction amount is 5 million yuan, the real-time identifier is "RT", and the data sensitivity identifier is "H". According to the above classification rules, this transaction is classified as a core transaction, with a transmission priority of P1, a latency requirement of ≤10ms, a packet loss rate requirement of ≤0.01%, and a high data sensitivity level, requiring encrypted transmission.

[0093] Step S2: SRv6 Network Deployment and Real-time Link Status Monitoring

[0094] The SRv6 network architecture is deployed, including SRv6 source nodes (core data centers of financial institutions), SRv6 forwarding nodes (key nodes in multiple links), and SRv6 destination nodes (branch offices or third-party data centers). Each link corresponds to a different SRv6 path, and each path consists of multiple consecutive SRv6 path segments. Each path segment is assigned a unique segment identifier (SID), forming a precisely manageable path system. Taking a financial institution as an example, its head office core data center deploys SRv6 source nodes, each first-tier branch data center deploys SRv6 forwarding nodes, and second-tier branches and outlets deploy SRv6 destination nodes. Between the head office and first-tier branches, two different physical paths are deployed: a dedicated line link and a 5G private network link.

[0095] Figure 2 The details of link status monitoring in step S2 are shown below:

[0096] A combination of periodic monitoring and event-triggered monitoring is employed. The SRv6 source node collects link status parameters (latency, packet loss rate, bandwidth utilization, link load, and security level) for each SRv6 path and its segments in real time. The monitoring cycle is dynamically adjusted based on the service priority specified in step S1; higher transmission priority corresponds to a shorter monitoring cycle. For the highest priority service type, the shortest cycle is used to collect the link status parameters involved in its transmission path. For example, the monitoring cycle for P1 service is 100ms, for P2 service it is 200ms, and for P3 service it is 500ms. When a sudden change in link parameters exceeds a preset threshold (e.g., a sudden increase in latency exceeding 20%), event monitoring is triggered, increasing the monitoring frequency to 50ms.

[0097] To achieve fine-grained monitoring of SRv6 path segments, a collaborative monitoring architecture is constructed, consisting of a monitoring center, data acquisition agents, and path detection probes. The monitoring center is deployed at the control layer and is responsible for the unified distribution of monitoring tasks, the aggregation and processing of monitoring data, and data interaction with the path calculation module.

[0098] The path segment-level monitoring adopts a dual-mode mechanism that combines active detection and passive data acquisition:

[0099] Passive data collection mode: A data collection agent is deployed on each SRv6 node (including source node, forwarding node, and destination node). This agent interacts with the node's operating system via NETCONF / YANG or gRPC protocols, subscribing in real-time to hardware counters and performance data for each interface corresponding to each SID, including the number of bytes, packets, dropped packets, and CRC errors in both inbound and outbound directions. The collection agent performs preliminary calculations on this raw data to derive metrics such as interface utilization and packet loss rate. This data is then aggregated at the SID granularity and pushed to the monitoring center via the Telemetry protocol at millisecond intervals (e.g., 100ms). For example, the collection agent on the SRv6 source node in the head office core computer room collects outbound interface data from the dedicated link between the head office and first-level branches every 100ms, calculating a real-time bandwidth utilization of 45% and a packet loss rate of 0.005% for this link.

[0100] Active Probe Mode: For complex multi-hop path segments, the monitoring center dynamically instantiates a lightweight probe session based on the STAMP protocol. A STAMP session sender is initiated at the head node of each path segment to be monitored, constructing a probe packet with a timestamp and forwarding it to the tail node along the path specified by the SID of that path segment. Upon receiving the probe packet, the STAMP session reflector at the tail node immediately constructs a response packet carrying the reception timestamp and returns along the original path. The head node calculates the one-way latency and jitter of the path segment based on the send and receive timestamps. For example, for a dedicated line path segment from the head office to a first-level branch, the head node sends a STAMP probe packet every 100ms, and by calculating the difference between the send and receive timestamps, the one-way latency of this path segment is found to be 8.5ms.

[0101] After receiving raw data from passive acquisition and active detection, the monitoring center performs a data preprocessing process: providing a unified clock reference for all nodes through the Precise Time Protocol (PTP), stamping the received data with arrival timestamps, and aligning asynchronous data to a unified time-domain grid using interpolation algorithms; removing abnormal values ​​caused by network interruptions and smoothing continuous packet loss data; and encapsulating the aligned and cleaned data according to a unified data model to form standardized link state tuples containing timestamps, SID identifiers, latency, packet loss rate, bandwidth utilization, and link status, which are then pushed to the path calculation module in real time via a high-speed message bus (such as Kafka).

[0102] Step S3: SRv6 candidate path generation

[0103] Based on the business type, QoS requirements, and security level requirements of the current financial transaction determined in step S1, and combined with the link status parameters of each SRv6 path monitored in step S2, SRv6 paths that meet all requirements are selected as candidate paths. The selection rules are: link latency ≤ current service latency threshold, packet loss rate ≤ current service packet loss rate threshold, bandwidth utilization ≤ 80% (preset), and security level ≥ security level corresponding to the sensitivity of the current service data.

[0104] Taking the aforementioned real-time transfer transaction of 5 million yuan (P1 service, latency threshold of 10ms, packet loss rate threshold of 0.01%, high security requirement) as an example, there are three SRv6 paths from the head office to the destination branch: Path A (dedicated line, latency of 8ms, packet loss rate of 0.005%, bandwidth utilization of 75%, high security level), Path B (5G private network, latency of 12ms, packet loss rate of 0.02%, bandwidth utilization of 40%, medium security level), and Path C (Internet VPN, latency of 15ms, packet loss rate of 0.1%, bandwidth utilization of 30%, low security level). Path A meets all the screening criteria (latency 8ms≤10ms, packet loss rate 0.005%≤0.01%, bandwidth utilization of 75%≤80%, high security level meets the requirements), while Path B does not meet the latency and packet loss rate requirements, and Path C does not meet the packet loss rate and security level requirements. Therefore, Path A is selected as the candidate path.

[0105] Step S4: Dynamic Optimization of SRv6 Path

[0106] A path optimization evaluation system is constructed based on the Analytic Hierarchy Process (AHP), with link latency, packet loss rate, bandwidth utilization, link reliability, and security level as core evaluation indicators. Based on the QoS requirements of each service type defined in step S1, differentiated weights are assigned to each evaluation indicator, as follows:

[0107]

[0108] See Figure 3The construction process of the AHP evaluation system is as follows: First, a hierarchical model is established, clearly dividing the target layer (optimal path selection), criterion layer (latency, packet loss rate, bandwidth utilization, link reliability, security level), and scheme layer (candidate paths). Second, the network management system or policy controller compares each pair of criterion indicators pairwise based on historical data and policy templates, assigning a scale value of 1-9 to form a judgment matrix. Third, a consistency verification method is established: the maximum eigenvalue and consistency ratio (CR) of the matrix are calculated; if CR < 0.1, the verification is passed; otherwise, the judgment matrix is ​​adjusted. Finally, the eigenvector method or geometric mean method is used to calculate the weight of each criterion. The normalized parameter values ​​of each candidate path under each criterion (based on real-time monitoring data) are weighted and summed with the corresponding weights to obtain a comprehensive evaluation value. The path with the highest score is selected as the optimal SRv6 transmission path.

[0109] To overcome the limitation of static weights in responding to network changes in real time, this solution introduces a combined weight generation mechanism based on "basic service weights" and "network dynamic adjustment factors." The bandwidth adjustment factor alpha_bw is calculated based on the maximum link utilization U_max(t) of the entire network. When U_max(t) is low (e.g., less than 50%), alpha_bw is close to 1; when U_max(t) approaches the congestion threshold (e.g., 80%), alpha_bw gradually increases to 1.5. The security adjustment factor alpha_sec is calculated based on the current security event level S_level(t). alpha_sec is 1 when the security level is normal, increases to 1.3 during alert periods, and increases to 2.0 during high-risk periods. The final comprehensive weight vector W_final is obtained by multiplying the basic service weight vector W_base element-wise with the network dynamic adjustment factor and then normalizing.

[0110] Taking the aforementioned real-time transfer transaction of 5 million yuan as an example, the normalized parameter values ​​of the current path A (dedicated line) are: latency 0.9, packet loss rate 0.95, bandwidth utilization 0.6, link reliability 0.98, and security level 1.0; the normalized parameter values ​​of path D (backup dedicated line) are: latency 0.85, packet loss rate 0.92, bandwidth utilization 0.4, link reliability 0.96, and security level 1.0.

[0111] Calculated using P1 service weights (latency 30%, packet loss rate 30%, security level 20%, link reliability 10%, bandwidth utilization 10%):

[0112] The overall score for path A is calculated as follows: 0.9 × 0.3 + 0.95 × 0.3 + 1.0 × 0.2 + 0.98 × 0.1 + 0.6 × 0.1 = 0.27 + 0.285 + 0.2 + 0.098 + 0.06 = 0.913.

[0113] The overall score for path D is calculated as follows: 0.85×0.3+0.92×0.3+1.0×0.2+0.96×0.1+0.4×0.1=0.255+0.276+0.2+0.096+0.04=0.867.

[0114] Path A scored higher and was selected as the optimal transmission path.

[0115] Step S5: WFQ+PQ multi-queue scheduling processing

[0116] Configure a WFQ+PQ hybrid queue scheduling strategy on the SRv6 source node, adopting a "priority first, fairness later" scheduling logic, and combine it with the business priority marked in step S1 to achieve differentiated scheduling of financial transactions.

[0117] See Figure 4 The scheduling logic consists of two steps:

[0118] The first step is PQ scheduling: Three priority queues are set up, corresponding to the priorities P1, P2, and P3 as defined in step S1. The scheduler prioritizes scheduling queue P1 (core transactions) until queue P1 is empty, then schedules queue P2, and finally queue P3. For example, when there are real-time transfer transaction messages in queue P1, the scheduler prioritizes processing the messages in that queue, even if there are a large number of messages waiting in queues P2 and P3, it must wait for queue P1 to finish processing before processing its own messages.

[0119] The second step is WFQ scheduling: Default bandwidth weights (5:3:2) are assigned to the three queues, which can be dynamically adjusted based on business traffic. Transaction data within each priority queue is divided into multiple data streams based on a five-tuple (source IP, destination IP, protocol number, source port, destination port). Each data stream is allocated a bandwidth share corresponding to its queue bandwidth weight, and a weighted round-robin scheduling method is used to ensure fair bandwidth allocation among data streams within the same priority queue. For example, in queue P2, there are two data streams: a batch clearing stream and a fund redemption stream. With a queue bandwidth weight of 3, each data stream receives a bandwidth share of 1.5. Weighted round-robin scheduling ensures fair bandwidth usage for both.

[0120] To achieve adaptive dynamic adjustment of queue bandwidth weights, a closed-loop feedback mechanism based on proportional-integral-derivative (PID) control is introduced into the WFQ scheduler. Each priority queue (P1, P2, P3) is configured with an independent PID controller to adjust its bandwidth weight in WFQ in real time. The input variable of the PID controller is the error value e_i(t), defined as the difference between the queue's real-time average delay D_avg_i(t) and the target QoS delay threshold D_target_i calibrated in step S1. The controller output variable u_i(t) is the bandwidth weight adjustment amount for queue i, using a positional PID control algorithm. After each queue's PID controller independently calculates its output, these adjustment amounts are collaboratively normalized to ensure that the sum of the total bandwidth weights of all queues is 1.

[0121] Step S6: SRv6 path encapsulation and data transmission

[0122] The SRv6 source node encapsulates the SID sequence corresponding to the optimal SRv6 transmission path determined in step S4 into the segmented routing header (SRH) of the IPv6 header of the financial transaction data. The SRH contains a segment list (storing the SID sequence) and a segment left shift pointer (pointing to the SID to be forwarded).

[0123] To achieve end-to-end QoS assurance, the priority decision of the source node is transmitted to all forwarding nodes along the entire SRv6 path through the Traffic Class (TC) field in the IPv6 header. At the SRv6 source node, after the WFQ+PQ scheduler in step S5 completes packet scheduling, it determines the priority queue (P1, P2, or P3) to which the packet belongs. The system maintains an internal QoS mapping table, pre-configured with mapping rules based on the DiffServ (Differentiated Services) model: P1 (core transactions) corresponds to an IPv6 TC value of 46 and a PHB (hop-by-hop behavior) of EF (accelerated forwarding); P2 (important services) corresponds to an IPv6 TC value of 26 and a PHB of AF31 (guaranteed forwarding); P3 (normal services) corresponds to an IPv6 TC value of 0 and a PHB of BE (best-effort forwarding). When constructing an SRv6 packet, the source node writes the determined TC value into the TrafficClass field of the IPv6 basic header according to the above mapping relationship.

[0124] Taking the aforementioned 5 million yuan real-time transfer transaction as an example, this transaction is marked as P1 priority, and the source node sets its TC field to 46. The optimal path for this transaction is the head office → first-tier branch dedicated line, with a corresponding SID sequence of [2001:db8:1::1,2001:db8:2::1,2001:db8:3::1]. This is encapsulated in the SRH Segment list, with the SL pointer initially set to 2. After encapsulation, the message is sent to the P1 queue to await scheduling.

[0125] When a packet carrying the above-described encapsulation arrives at any SRv6 intermediate forwarding node, the node's forwarding plane parses the SRH, looks up the local forwarding table based on the SID pointed to by the SL pointer, determines the next-hop outgoing interface and the next-hop IPv6 address, updates the SL pointer (decrements the SL value by 1), and forwards the packet to the next hop. Simultaneously, the node's queue scheduling module parses the TC field in the IPv6 header, identifies a TC value of 46, and sends the packet to the EF queue according to the local QoS configuration for priority forwarding.

[0126] Step S7: Dynamic Path Adjustment and Fault Switching

[0127] During the transaction data transmission process, the SRv6 source node continuously monitors the optimal SRv6 transmission path and the link status parameters of each path segment (continuing the monitoring mechanism of step S2).

[0128] If the detected link status parameters exceed the preset threshold (such as sudden increase in latency, excessive packet loss rate, or link interruption), the path is immediately marked as unavailable, triggering the path reselection process. Steps S3-S6 are repeated to re-select candidate paths and determine a new optimal path. Simultaneously, priority is given to retaining transaction data with P1 and P2 priorities that have not been fully transmitted, ensuring that core transactions and important services are not lost or interrupted. Transaction data with P3 priority can be retransmitted or delayed depending on the link recovery status.

[0129] Taking the aforementioned real-time transfer transaction of 5 million yuan as an example, when the system detects a sudden increase in the dedicated line link latency from 8ms to 25ms (exceeding the P1 service latency threshold of 10ms), it immediately marks the path as unavailable and triggers path reselection. During the reselection process, the system finds that the backup 5G dedicated network link currently has a latency of 9ms and a packet loss rate of 0.008%, meeting the P1 service requirements, and switches the transaction traffic to the 5G dedicated network link. During the switching process, the system prioritizes retaining the incomplete data of this real-time transfer transaction to ensure that the transaction is not interrupted.

[0130] To achieve second-level failover, this solution introduces a pre-calculated backup path mechanism in addition to primary path optimization. After completing an optimal path calculation based on AHP in step S4, the path calculation module outputs a candidate list containing the top N (e.g., top three) optimal paths. Each path in this list includes its complete SID sequence, AHP score, and corresponding link status snapshot timestamp, cached in the controller's local high-speed storage. When the link parameters of the current optimal path exceed the preset threshold multiple times consecutively, the controller directly retrieves the currently available backup path with the highest score from the local cache and sends the SRv6 policy containing the backup path's SID list to the source node PE device via the PCEP or NETCONF protocol. The source node immediately switches the affected core and important service traffic to the new path. After the switchover is complete, the controller asynchronously triggers a complete path re-optimization process in the background, re-executing steps S3 and S4, recalculating the optimal path set based on the latest network-wide link status, and updating the cache of the pre-calculated backup path list.

[0131] If bandwidth contention is detected within queues of the same priority, the queue bandwidth weights are dynamically adjusted through the WFQ scheduling mechanism (continuing the scheduling logic of step S5).

[0132] Step S8: Data Verification and Security Audit

[0133] After receiving transaction data, the SRv6 destination node parses the SID sequence in the SRH (corresponding to the encapsulation content in step S6), compares it with the preset valid path SID sequence, and verifies the correctness of the transmission path. It uses the CRC32 algorithm to verify data integrity. It also verifies the digital signature in the transaction data (using a preset algorithm from the financial institution) to confirm the legitimate data source. If the verification passes, data reception is complete; if the verification fails, an exception message is sent to the SRv6 source node, triggering data retransmission.

[0134] Digital signatures employ a Public Key Infrastructure (PKI)-based signature scheme: The sender (SRv6 source node) signs the transaction data hash (e.g., SHA-256) using its own private key, employing either RSA-PSS or ECDSA as the signature algorithm. The signature value, along with the certificate (or public key identifier), is encapsulated in the packet header or payload. The receiver (SRv6 destination node) first verifies the validity of the sender's certificate chain (including validity period, issuer, and revocation status), then decrypts the signature using the sender's public key, calculates the hash value of the received data, and compares the two for consistency.

[0135] The system records the entire process information for each financial transaction, including transaction ID, business type (classification result in step S1), transmission path SID sequence (encapsulated content in step S6), queue scheduling information (scheduling result in step S5), link status parameters (monitoring data in step S2), verification result, and transmission time, forming a complete security audit log. Log entries are in a structured format and written in real-time to a distributed storage system with redundancy protection, and encrypted storage (such as AES-GCM) is used to prevent tampering. To meet compliance retention requirements (≥1 year), logs can be automatically tiered and stored on low-cost media with integrity protection implemented. The audit interface provides a role-based access control (RBAC)-based query API, supports multi-dimensional retrieval and export, and generates compliance reports.

[0136] Taking the aforementioned 5 million yuan real-time transfer transaction as an example, after receiving the message, the destination node parses the SID sequence [2001:db8:1::1,2001:db8:2::1,2001:db8:3::1] in the SRH and compares it with the preset legal path SID sequence to confirm consistency; it calculates the CRC32 value of the message and compares it with the check value carried by the sender to confirm consistency; it verifies the sender's digital signature to confirm the legality of the transaction source. After the verification is passed, the system generates an audit log, recording the transaction ID "TR202501150001", business type "core transaction", transmission path "head office → first-level branch dedicated line", queue scheduling information "P1 queue priority scheduling", link status parameters "latency 8ms, packet loss rate 0.005%", verification result "passed", and transmission time "2025-01-15 10:30:25.123". The log is saved to the distributed storage system.

[0137] III. System Implementation Examples

[0138] This embodiment provides a multi-link intelligent routing system for financial transactions, which is implemented based on the method embodiment described above. For example... Figure 5 As shown, the system consists of three layers: network layer, control layer, and application layer.

[0139] The network layer adopts a dual-ring, dual-plane topology. Each data center deploys dual-machine backbone core routers and dual-plane backbone edge routers. High-bandwidth optical transmission network lines are primarily used between data centers, ensuring full interconnection between each other. The production backbone network also employs a dual-plane architecture, with dual-machine P-PE devices deployed in each data center, providing higher availability. The non-production backbone network shares the same architecture as the production backbone network, but with reduced equipment and link resources, deployed on a single plane. The overall architecture design employs a three-layer structure: core layer, aggregation layer, and access layer, with each layer using a dual-plane architecture.

[0140] The core layer consists of interconnected data center P backbone routers, providing high-speed forwarding and traffic scheduling between data centers; the aggregation layer consists of DC-PE and branch aggregation PE nodes, serving as the backbone network service access aggregation point; the access layer consists of data center core switches and branch intranet area switches, serving the data center and branch service sides.

[0141] The control layer employs an off-site disaster recovery architecture deployed SDN controller cluster, encompassing network management, network control, and network analysis. The control layer includes the following modules:

[0142] The SRv6 policy management module is used for the generation, distribution, and management of SRv6 policies. This module includes: a policy generation submodule, used to generate a list of SRv6 SIDs based on business requirements; a policy verification submodule, used to verify the syntactic and semantic correctness of policies; a policy distribution submodule, used to distribute policies to network devices via the NETCONF / YANG protocol; and a policy monitoring submodule, used to monitor the execution status and effectiveness evaluation of policies.

[0143] The intelligent path calculation module is used for path calculation based on multiple constraints, and implements the path filtering and optimization functions of steps S3 and S4 in the above method embodiment.

[0144] The QoS policy management module is used for the formulation and execution of QoS policies, and implements the queue scheduling policy configuration in step S5 of the above method embodiment.

[0145] The dynamic optimization module is used for automatic optimization when the link is congested, and realizes the path dynamic adjustment and fault switching functions in step S7 of the above method embodiment.

[0146] The data acquisition module is used for real-time acquisition of network status data, realizing the link status monitoring function in step S2 of the above method embodiment.

[0147] The application layer includes various network applications developed based on the controller's northbound interface.

[0148] The system also includes functional modules corresponding to the above method steps:

[0149] The business awareness module is configured to acquire financial transaction data to be transmitted, extract transaction feature information from the financial transaction data, including transaction type, transaction amount, real-time requirements, and data sensitivity; and classify the financial transactions into multiple business types based on the transaction feature information, assigning corresponding transmission priority, service quality requirement parameters, and data sensitivity levels to each business type. This module corresponds to step S1 in the method embodiment.

[0150] The link monitoring module, configured to deploy an SRv6 network, uses a combination of periodic monitoring and event-triggered monitoring to collect link status parameters of each SRv6 path and each path segment in real time. The monitoring period is dynamically adjusted according to the transmission priority of the service type. This module corresponds to step S2 in the method embodiment. The link monitoring module includes: a path segment monitoring unit, used to collect the port status and forwarding latency of the forwarding nodes corresponding to each path segment in each SRv6 path; a period adjustment unit, used to dynamically adjust the monitoring period according to the transmission priority of the service type, wherein a higher transmission priority corresponds to a shorter monitoring period; and an event triggering unit, used to increase the monitoring frequency to a preset high-frequency period when a sudden change in link status parameters exceeds a preset threshold.

[0151] The path optimization module is configured to filter candidate paths that meet the service quality requirements and data sensitivity levels based on the service type and link status parameters; construct a path optimization evaluation system based on the analytic hierarchy process (AHP), which includes multiple evaluation indicators such as link latency, packet loss rate, bandwidth utilization, link reliability, and security level; assign differentiated weights to the multiple evaluation indicators according to the transmission priority of the service type; calculate the comprehensive evaluation score for each candidate path; and select the path with the highest comprehensive evaluation score as the optimal transmission path. This module corresponds to steps S3 and S4 in the method embodiment. The path selection module includes: a hierarchical analysis unit, used to establish a hierarchical structure model including a target layer, a criterion layer, and a scheme layer; based on the transmission priority of the service type, a pairwise importance scale between each evaluation indicator is preset; a judgment matrix is ​​constructed based on the scale; the maximum eigenvalue and consistency ratio of the judgment matrix are calculated; if the consistency ratio is less than a preset threshold, the consistency test is passed; a weight allocation unit, used to calculate the weight of each evaluation indicator based on the judgment matrix that has passed the consistency test; and a scoring calculation unit, used to sum the normalized parameter values ​​of each candidate path under each evaluation indicator with the corresponding weight to obtain the comprehensive evaluation score of each candidate path.

[0152] The queue scheduling module, deployed on the SRv6 source node, includes multiple priority queues corresponding to the transmission priority. It is configured to perform differentiated queue scheduling for financial transactions of different priorities using a logic of priority-based scheduling followed by weighted fair scheduling. This module corresponds to step S5 in the method embodiment. The queue scheduling module includes: a priority scheduling unit, used to set multiple priority queues corresponding to the transmission priority, prioritizing the scheduling of the highest priority queue until that queue is empty, and then sequentially scheduling the next highest priority queue; a fair scheduling unit, used to allocate queue-level bandwidth weights to each priority queue; dividing the financial transaction data within each priority queue into multiple data streams according to a 5-tuple; allocating bandwidth shares proportionally to each data stream within the bandwidth weight range of that queue; and scheduling each data stream using a weighted round-robin method.

[0153] The encapsulation and transmission module is configured to encapsulate the segment identifier sequence corresponding to the optimal transmission path into the segment routing header of the IPv6 header of the financial transaction data, add a traffic category field to the IPv6 header, and write the transmission priority into the traffic category field before transmission. This module corresponds to step S6 in the method embodiment. The encapsulation and transmission module includes: a path encapsulation unit, used to sequentially encapsulate the segment identifier sequence corresponding to the optimal transmission path into the segment routing header of the IPv6 header of the financial transaction data, the segment routing header including a segment identifier list and a segment left shift pointer; and a priority marking unit, used to add a traffic category field to the IPv6 header and write the transmission priority into the traffic category field so that forwarding nodes can identify the priority and perform corresponding queue scheduling.

[0154] The fluctuation response module is configured to determine the type of network fluctuation when network status fluctuations are detected, and set differentiated fluctuation response rules for financial transactions of different priorities according to the transmission priority.

[0155] This module corresponds to the fluctuation response part in step S7 of the method embodiment. The bandwidth adjustment operation in the fluctuation response rule is implemented in the following way: the bandwidth weight in the hybrid queue scheduling strategy supports dynamic reallocation. When a network fluctuation such as bandwidth reduction is detected, the queue scheduling module performs the following operations on each priority queue in descending order of transmission priority: lock the current bandwidth weight of queue P1 and prevent compression, compress the bandwidth weight of queue P2 proportionally, and temporarily reduce the bandwidth weight of queue P3 to zero. The compression and locking operations are implemented by updating the bandwidth weight configuration of the WFQ scheduler, and the configuration update takes effect immediately.

[0156] IV. Extensions and Variations of the Implementation Examples

[0157] 1. Basic Simplified Implementation Example (Lightweight Deployment Scenario for Small and Medium-Sized Financial Institutions)

[0158] It is suitable for small and medium-sized financial institutions such as city commercial banks, rural commercial banks, and village banks, with small network deployment scale (number of links ≤ 10, transaction peak ≤ 1000TPS), single core transaction type (only real-time transfer and payment settlement are included), and limited operation and maintenance resources.

[0159] The proposed adjustments are as follows: Business classification is simplified to a two-tier system, retaining only core transactions (P1) and ordinary transactions (P2). Core transactions only include real-time transfers and payment settlements. The AHP evaluation system criteria are simplified to three core indicators (link latency, packet loss rate, and security level), with fixed weight allocations for P1 (latency 40%, packet loss rate 40%, security level 20%) and P2 (latency 20%, packet loss rate 20%, security level 60%). Complex consistency check iterations are eliminated, replaced by fixed-weight linear scoring. Monitoring cycles are fixed at P1=200ms and P2=500ms, eliminating high-frequency monitoring triggered by events. Queue bandwidth weights are fixed at 2:1 (P1:P2), eliminating dynamic adjustment functionality. Security checks retain path verification and integrity verification; digital signature verification is temporarily disabled, and audit logs only record core fields.

[0160] 2. High Availability Enhancement Implementation Example (Core Transaction Scenarios of a Nationwide Large Bank)

[0161] It is applicable to the core transaction systems of state-owned banks and national joint-stock banks (transaction peak ≥ 10000 TPS), involving cross-regional multi-link deployment (number of links ≥ 50, including heterogeneous links such as dedicated lines / 5G private networks / satellite links), and has a wide variety of core transaction types.

[0162] The solution is adjusted as follows: Business classification adds sub-priorities to core transactions, dividing them into P1-1 (large transfers / clearances, ≥1 million) and P1-2 (small core transactions, <1 million) based on transaction amount. P1-1 enjoys the highest level of monitoring, routing, and scheduling priority. Monitoring adds redundant monitoring across multiple nodes in different locations, setting up two backup monitoring nodes in addition to the primary monitoring node. The AHP optimization system adds a link heterogeneity evaluation index, prioritizing heterogeneous links as candidate paths. Fault switching implements pre-calculation of the optimal backup path, synchronously calculating and caching the top three optimal paths, reducing fault switching time to ≤50ms, and increasing local caching of transaction data and off-site disaster recovery. WFQ+PQ scheduling adds logical bandwidth isolation, allocating dedicated isolated bandwidth (30% of the total bandwidth) to the P1-1 sub-priority. Transmission encryption adopts the SM4 national cryptographic algorithm, and the audit log adds operation and maintenance records.

[0163] 3. Lightweight Edge Deployment Implementation Examples (Bank Offline Branches / Mobile Business Development Scenarios)

[0164] It is suitable for edge scenarios such as bank offline branches and mobile business terminals (such as mobile teller machines and outdoor card opening terminals). The deployment environment is lightweight hardware (edge ​​gateway / industrial router), with limited computing resources (CPU≤4 cores, memory≤8G), and the link is mainly wireless link (5G / 4G / WiFi).

[0165] The solution has been adjusted as follows: Core modules (service classification, link monitoring, simplified routing, and queue scheduling) are deployed on the edge gateway, while non-core modules (such as full-scale security auditing and off-site disaster recovery) are removed. The total memory usage of core modules is ≤1GB, and CPU utilization is ≤30%. Link monitoring adds a link signal strength monitoring indicator, with the monitoring period adaptively adjusted (200ms when signal strength is ≥-70dBm, and automatically increased to 100ms when it is <-70dBm). The routing algorithm eliminates the AHP (Analytic Hierarchy Process) and adopts a multi-threshold weighted fast routing algorithm, setting fixed thresholds and weights only for three indicators: latency, packet loss rate, and signal strength. SRv6 encapsulation adds adaptive link MTU adjustment, adjusting the encapsulation packet size based on the real-time MTU value of the wireless link. Fault handover adds a weak network speed reduction strategy, automatically reducing the transmission rate of ordinary services when the wireless link signal is weak.

[0166] 4. System Deployment Variants

[0167] This system can be adapted into a centralized deployment system variant and a distributed cloud-edge collaborative deployment system variant, depending on the financial institution's network architecture, cloud deployment strategy, and business coverage.

[0168] Centralized deployment system variant: All core modules are centrally deployed in the cloud server / physical server cluster in the core computer room of the financial institution’s head office, and the multi-link resources of all branch nodes across the country are uniformly managed. Branch nodes only deploy SRv6 forwarding nodes and data collection agents. All transaction data is uploaded to the head office core module for processing and then transmitted along the optimal SRv6 path.

[0169] A variant of the distributed cloud-edge collaborative deployment system adopts a layered deployment model of "cloud-based control + edge processing". The cloud (head office / provincial branch cloud platform) deploys a global routing control module, a full-scale security audit module, and a cross-regional link scheduling module; the edge (city branches / offline outlets / edge gateways) deploys a local business classification module, a local link monitoring module, a simplified routing module, and a WFQ+PQ scheduling module. Local transactions are processed independently by the edge, with results synchronized to the cloud; cross-regional transactions are uploaded from the edge to the cloud, where global routing is performed before being distributed to the edge for execution.

[0170] The embodiments described above are merely preferred embodiments of this application and are not intended to limit the scope of protection of this application. Those skilled in the art should understand that, inspired by the technical concepts and principles disclosed in this application, such as business perception based on deep characteristics of financial transactions, dynamic adjustment of link monitoring cycles based on business priorities, allocation of path evaluation index weights based on the analytic hierarchy process (AHP), hybrid queue scheduling combining priority scheduling followed by weighted fair scheduling, and priority hop-by-hop propagation through the SRv6 header traffic category field, modifications can be made to the technical solutions described in the above embodiments. For example, the granularity of business classification can be adjusted, the specific value of the monitoring cycle can be changed, the index composition or weight allocation scheme of the AHP evaluation system can be modified, the bandwidth weight ratio of the WFQ queue can be adjusted, or the specific implementation method of SRv6 encapsulation can be changed. Equivalent substitutions can also be made to some technical features, such as replacing the specific algorithm for dynamic adjustment of the monitoring cycle with other adaptive algorithms with the same function, or replacing the specific implementation method of PID control for adjusting queue weights with other closed-loop control algorithms. These modifications or equivalent substitutions do not depart from the technical concepts and principles of this application and should all fall within the scope of protection of this application.

[0171] The scope of protection of this application is determined by the appended claims, and the description and drawings are used to interpret the content of the claims. Any modifications, equivalent substitutions, improvements, etc., made to the technical solution of this application within the technical concept and principle of this application are included within the scope of protection of this application.

Claims

1. A multi-link intelligent routing method for financial transactions, characterized in that, include: Acquire financial transaction data to be transmitted, and extract transaction feature information from the financial transaction data, including transaction type, transaction amount, real-time requirements, and data sensitivity; The financial transactions are divided into multiple business types based on the transaction feature information, and each business type is assigned a corresponding transmission priority, service quality requirement parameters, and data sensitivity level. Deploy an SRv6 network and use a combination of periodic monitoring and event-triggered monitoring to collect link status parameters of each SRv6 path and each path segment in real time. The monitoring period is dynamically adjusted according to the transmission priority of the service type. Candidate paths that meet the service quality requirements and data sensitivity levels are selected based on the service type and the link status parameters. A path selection evaluation system is constructed based on the analytic hierarchy process (AHP). The path selection evaluation system includes multiple evaluation indicators such as link latency, packet loss rate, bandwidth utilization, link reliability, and security level. Differentiated weights are assigned to the multiple evaluation indicators according to the transmission priority of the service type. The comprehensive evaluation score of each candidate path is calculated, and the path with the highest comprehensive evaluation score is selected as the optimal transmission path. Configure a hybrid queue scheduling strategy on the SRv6 source node. The hybrid queue scheduling strategy includes multiple priority queues corresponding to the transmission priority. The logic of priority scheduling followed by weighted fair scheduling is used to perform differentiated queue scheduling for financial transactions with different priorities. The segment identifier sequence corresponding to the optimal transmission path is encapsulated into the segment routing header of the IPv6 header of the financial transaction data, and a traffic category field is added to the IPv6 header. The transmission priority is written into the traffic category field before transmission.

2. The multi-link intelligent routing method for financial transaction according to claim 1, wherein, The method employs a combination of periodic monitoring and event-triggered monitoring to collect link status parameters for each SRv6 path and each path segment in real time, including: Link status parameters of financial transactions corresponding to a first transmission priority are collected in a first cycle, and link status parameters of financial transactions corresponding to a second transmission priority are collected in a second cycle, wherein the first transmission priority is higher than the second transmission priority, and the first cycle is shorter than the second cycle. When a sudden change in link status parameters is detected that exceeds a preset threshold, event monitoring is triggered, and the monitoring frequency is increased to the third cycle, which is shorter than the first cycle.

3. The method of claim 1, wherein, The path selection evaluation system based on the analytic hierarchy process includes: A hierarchical model is established, comprising a target layer, a criterion layer, and a solution layer, wherein the target layer is for optimal path selection, the criterion layer includes the multiple evaluation indicators, and the solution layer includes the candidate paths; Based on the transmission priority of the service type, a pairwise importance scale between each evaluation index is preset, a judgment matrix is ​​constructed based on the scale, the maximum eigenvalue and consistency ratio of the judgment matrix are calculated, and if the consistency ratio is less than a preset threshold, the consistency test is passed. The weights of each evaluation index are calculated based on the judgment matrix that has passed the consistency test. The normalized parameter values ​​of each candidate path under each evaluation index are weighted and summed with the corresponding weights to obtain the comprehensive evaluation score of each candidate path.

4. The multi-link intelligent routing method for financial transactions according to claim 1, characterized in that, The logic of prioritizing scheduling followed by weighted fair scheduling for differentiated queue scheduling of financial transactions with different priorities includes: Set up multiple priority queues corresponding to the transmission priority, and prioritize scheduling the highest priority queue until that queue is empty, then schedule the next highest priority queue in turn. Assign queue-level bandwidth weights to each priority queue; divide the financial transaction data in each priority queue into multiple data streams according to 5-tuples; allocate bandwidth shares to each data stream proportionally within the bandwidth weight range of the queue; and schedule each data stream using a weighted round-robin method.

5. The multi-link intelligent routing method for financial transactions according to claim 4, characterized in that, The method further includes: The bandwidth weights of each priority queue are adjusted in real time by a proportional-integral-derivative controller. The input variable of the proportional-integral-derivative controller is the difference between the real-time average latency of the queue and the latency threshold in the service quality requirement parameter, and the output variable is the bandwidth weight adjustment amount.

6. The multi-link intelligent routing method for financial transactions according to claim 1, characterized in that, The step of encapsulating the segment identifier sequence corresponding to the optimal transmission path into the segmented routing header of the IPv6 header of the financial transaction data, adding a traffic category field to the IPv6 header, and writing the transmission priority into the traffic category field before transmission includes: The segment identifier sequence corresponding to the optimal transmission path is sequentially encapsulated into the segment routing header of the IPv6 header of the financial transaction data. The segment routing header includes a segment identifier list and a segment left shift pointer. A traffic category field is added to the IPv6 header, and the transmission priority is written into the traffic category field so that forwarding nodes can identify the priority and perform corresponding queue scheduling.

7. The multi-link intelligent routing method for financial transactions according to claim 1, characterized in that, The method further includes: When network status fluctuations are detected, the type of network fluctuation is determined, and differentiated fluctuation response rules are set for financial transactions of different priorities based on the transmission priority.

8. The multi-link intelligent routing method for financial transactions according to claim 7, characterized in that, The types of network fluctuations include reduced bandwidth, increased node response latency, packet loss, and network congestion. The step of setting differentiated volatility response rules for financial transactions of different priorities based on the transmission priority includes: If the network fluctuation type is bandwidth reduction, then according to the priority descending order, lock dedicated bandwidth, compress bandwidth ratio, and suspend bandwidth allocation for the corresponding financial transactions; If the network fluctuation type is an increase in node response latency, then according to the priority descending order, the corresponding financial transaction is set to switch to a backup node, increase the number of verification nodes, extend the processing time, and suspend processing. If the network fluctuation type is packet loss, then in descending order of priority, the corresponding financial transactions are set to enable redundant transmission, shorten retransmission interval, retransmission, and pending retransmission. If the network fluctuation type is network congestion, resources are allocated from transactions below the preset level to transactions above the preset level, and resource allocation is restored in a preset order after the congestion is relieved.

9. A multi-link intelligent routing system for financial transactions, characterized in that, The system includes: The business awareness module is configured to acquire financial transaction data to be transmitted, extract transaction feature information of the financial transaction data, including transaction type, transaction amount, real-time requirements and data sensitivity; and classify the financial transaction into multiple business types according to the transaction feature information, and assign corresponding transmission priority, service quality requirement parameters and data sensitivity level to each business type. The link monitoring module is configured to deploy an SRv6 network and uses a combination of periodic monitoring and event-triggered monitoring to collect link status parameters of each SRv6 path and each path segment in real time. The monitoring period is dynamically adjusted according to the transmission priority of the service type. The path optimization module is configured to filter candidate paths that meet the service quality requirements and data sensitivity levels based on the service type and the link status parameters; construct a path optimization evaluation system based on the analytic hierarchy process (AHP), which includes multiple evaluation indicators such as link latency, packet loss rate, bandwidth utilization, link reliability, and security level; assign differentiated weights to the multiple evaluation indicators according to the transmission priority of the service type; calculate the comprehensive evaluation score of each candidate path; and select the path with the highest comprehensive evaluation score as the optimal transmission path. The queue scheduling module, deployed on the SRv6 source node, includes multiple priority queues corresponding to the transmission priority, and is configured to perform differentiated queue scheduling for financial transactions of different priorities using a logic of priority scheduling followed by weighted fair scheduling. The encapsulation and transmission module is configured to encapsulate the segment identifier sequence corresponding to the optimal transmission path into the segment routing header of the IPv6 header of the financial transaction data, add a traffic category field to the IPv6 header, and transmit the data after writing the transmission priority into the traffic category field.

10. A multi-link intelligent routing system for financial transactions according to claim 9, characterized in that, The link monitoring module includes: The path segment monitoring unit is configured to collect the port status and forwarding latency of the forwarding nodes corresponding to each path segment in each SRv6 path; The period adjustment unit is configured to dynamically adjust the monitoring period according to the transmission priority of the service type, wherein the higher the transmission priority, the shorter the corresponding monitoring period. The event triggering unit is configured to increase the monitoring frequency to a preset high-frequency period when a sudden change in the link status parameter is detected to exceed a preset threshold.

Images