SRv6-based financial intelligent backbone network traffic scheduling method and system

Abstract

The application discloses a financial intelligent backbone network traffic scheduling method and system based on SRv6, and relates to the technical field of network scheduling.The method comprises the following steps: step S1, identifying and dividing the entry service traffic into predefined service types; step S2, matching the corresponding SRv6 traffic tunnel model for each service type; step S3, allocating differentiated QoS strategies for different types of service flows, and performing corresponding queue scheduling; step S4, constructing the corresponding optimization objective function for path calculation according to the SLA requirements of different service types; step S5, automatically triggering the intelligent optimization process when detecting that the link utilization rate exceeds the preset threshold; and step S6, issuing the optimal SRv6 strategy calculated to the entry device for execution.The application builds an intelligent and efficient financial backbone network architecture, and realizes fine traffic scheduling strategies, thereby improving the network resource utilization rate and operation and maintenance efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network scheduling, in particular to a financial intelligent backbone network traffic scheduling method and system based on SRv6. BACKGROUND

[0002] In recent years, with the rapid development and wide application of financial technology, the business support capability of financial data centers has become a hot spot of industry concern. As an important bridge for interconnection between multiple data centers, the backbone network has become increasingly important to ensure its stability and robustness. In response to the national strategy of deploying IPv6, Postal Savings Bank has adopted the industry-leading "SDN+SRv6" technology to build a new generation of backbone network covering two sites, four centers and provincial branches nationwide, and has improved the reliability of the backbone network in all directions and multiple dimensions. Online banking, mobile payment, real-time trading and other businesses have put forward very high requirements for the transmission quality, reliability and intelligent level of the financial backbone network.

[0003] Traditional financial backbone networks are mostly built based on MPLS (Multi-Protocol Label Switching) technology, which lacks flexibility in traffic scheduling and relies on distributed protocols for path calculation, making it difficult to achieve end-to-end precise scheduling based on business intent. Network operation usually relies on static configuration and manual intervention, which is slow in response and cannot effectively cope with network congestion and sudden traffic, resulting in low utilization of network resources and difficulty in meeting the differentiated service quality (SLA) needs of different financial businesses. Although the software-defined network (SDN) technology brings the advantage of centralized control, its combination with data plane forwarding technology is still not close enough to achieve efficient service chain construction and real-time performance measurement in the IPv6 network environment. Therefore, there is an urgent need for a new generation of financial backbone network solution that can deeply integrate advanced network technologies and achieve intelligent and differentiated traffic scheduling.

[0004] A network health degree evaluation method and device for a financial cloud backbone network are disclosed in Chinese Patent No. CN116248563B, which includes determining the traffic of different paths between the head office and branch offices in the financial cloud backbone network, and determining the grades of the different path traffic according to the bandwidth overhead required by the determined different path traffic. The different paths include a main path and a protection path corresponding to the main path. Based on the determined grades of the different path traffic, the network reliability and network carrying capacity of each grade traffic between each head office and each branch office on different paths are calculated. The health degree of each grade traffic is calculated based on the calculated network reliability and network carrying capacity, and the health degrees of each grade traffic are summed to obtain the network health degree of each grade traffic. The network health degree of the overall network is obtained by weighted sum of the proportion of each grade traffic.

[0005] A SRv6 backbone network traffic congestion risk assessment method is disclosed in a Chinese patent with the publication number CN115604724B. It mainly considers two indicators: the path bandwidth between the source and destination nodes, which measures the network's carrying capacity for traffic; and the predicted network traffic in the future period. Since network congestion assessment serves future optimization decisions, evaluating only the current traffic is insufficient. The demand traffic from the destination node to the source node is predicted using a recurrent neural network in machine learning, and the prediction results are used to represent the average level of traffic burst in the future period. SUMMARY

[0006] The present application aims to solve the problems of inflexible traffic scheduling, low intelligent operation and maintenance level, and difficulty in meeting the differentiated business SLA requirements in existing financial backbone networks.

[0007] To achieve the above purpose, the technical solution adopted by the present application is:

[0008] The SRv6-based financial intelligent backbone network traffic scheduling method comprises the following steps:

[0009] Step S1, based on the five-tuple information of deep packet inspection business flow, the ingress traffic is identified and divided into predefined business types;

[0010] Step S2, for each identified business type, a corresponding SRv6 traffic tunnel model is matched, which includes a single-path multi-segment routing model for latency-sensitive business and a multi-path single-segment routing model for bandwidth-sensitive business;

[0011] Step S3, different QoS policies are allocated to different categories of business flow, and the network nodes perform corresponding queue scheduling according to the QoS policies;

[0012] Step S4, based on the real-time collected network state information, the corresponding optimization objective function is constructed for the SLA requirements of different business types to perform path calculation;

[0013] Step S5, through real-time data flow monitoring, the network link state is monitored, and when the link utilization rate exceeds the preset threshold, the intelligent optimization process is automatically triggered;

[0014] Step S6, the optimal SRv6 strategy calculated is encoded into a SID list, and is issued to the ingress PE device for execution through the control protocol.

[0015] Further, in step S1, the predefined business types include operation and maintenance, transaction business, monitoring business, backup business, and other business.

[0016] Further, in the step S2, the single-path multi-segment routing model matches transaction business class services and operation and maintenance management class services, the multi-path single-segment routing model matches monitoring business class services, backup business class services and other business class services;

[0017] Among them, the single-path multi-segment routing model specifies the precise path through an ordered SID list; the multi-path single-segment routing model only specifies the destination SID and uses ECMP for load sharing.

[0018] Further, in the step S3, the QoS policy includes assigning different priority weights to different service categories, and configuring corresponding queue scheduling weights on all network devices according to the priority weights.

[0019] Further, in the step S4, corresponding optimization objective functions are constructed for the SLA requirements of different service types for calculation, including two modes of delay optimization calculation and bandwidth optimization calculation;

[0020] Among them, the delay optimization calculation is used for delay-sensitive services, including transaction business class services and operation and maintenance management class services, and adopts the objective function of minimizing the end-to-end delay path, which includes three components of link propagation delay, node processing delay and service queuing delay;

[0021] The bandwidth optimization calculation is used for bandwidth-sensitive services, including monitoring business class services, backup business class services and other business class services, and adopts the objective function of finding path combinations that meet the bandwidth requirements, and the sum of the available bandwidths of the paths is calculated to meet the bandwidth requirements of the services.

[0022] Further, in the step S5, the tuning process specifically includes:

[0023] Identify all service flows passing through congested links to form a set of flows to be adjusted, extract the service characteristics of each flow in the set of flows to be adjusted, including service category, service priority, bandwidth requirement and delay requirement;

[0024] Sort based on service priority and set a priority threshold, and include service flows with a priority lower than the threshold in the set of adjustable flows;

[0025] Recalculate the path for each flow in the set of adjustable flows, and the optimization objective is to minimize the maximum link utilization rate of the whole network while meeting the service SLA constraints;

[0026] After calculating the new path, the controller generates the corresponding SRv6 policy and issues the policy to the related PE devices for traffic switching;

[0027] Among them, the traffic switching adopts a seamless switching mechanism, and the switching time includes propagation delay, processing delay and buffer time.

[0028] Further, the recalculation of the path for each flow in the adjustable flow set adopts a multi-stage heuristic algorithm, including a single-path search stage and a multi-path load sharing stage;

[0029] In the single-path search stage, a single alternative path is found for each flow, and all links on the new path have utilization lower than the threshold after increasing the bandwidth demand of the flow to be adjusted;

[0030] If the single-path cannot meet the requirement, the multi-path load sharing stage is entered, and the sum of the available bandwidths of multiple paths is calculated to meet the bandwidth demand of the service, and the load of each path is increased to have utilization lower than the threshold;

[0031] In the single-path search stage, a single alternative path is found for each flow, and all links on the new path have utilization lower than the threshold after increasing the bandwidth demand of the flow to be adjusted;

[0032] The multi-path load sharing stage specifically includes the following steps:

[0033] The controller generates K feasible paths leading to the same destination based on the network topology to form a candidate path set;

[0034] The bottleneck bandwidth of each path is calculated;

[0035] The initial splitting ratio is calculated according to the available bandwidth capacity of each path;

[0036] Whether the utilization constraint is met under the initial splitting ratio is verified path by path and link by link;

[0037] If any link on any path does not meet the utilization constraint, adjustment is needed until a set of paths and their splitting ratios are found to meet all constraints.

[0038] The SRv6-based financial intelligent backbone network traffic scheduling system is implemented based on the SRv6-based financial intelligent backbone network traffic scheduling method, and includes three levels of network layer, control layer and application layer;

[0039] The network layer adopts a dual-ring dual-plane topology structure, and dual machines are deployed in each data center backbone network core router and backbone network edge router device to form a dual plane. The data centers are mainly connected by high-bandwidth optical transmission network lines, and are fully interconnected between each other.

[0040] The control layer adopts an SDN controller cluster deployed in a geographically dispersed disaster recovery architecture, including network management, network control and network analysis;

[0041] The application layer comprises various network applications developed based on the controller northbound interface.

[0042] Further, the control layer further comprises:

[0043] An SRv6 policy management module for generating, issuing and managing SRv6 policies;

[0044] An intelligent path calculation module for path calculation based on multi-constraint conditions;

[0045] A QoS policy management module for formulating and executing QoS policies;

[0046] A dynamic tuning module for automatic tuning when link congestion occurs;

[0047] A data acquisition module for real-time acquisition of network state data.

[0048] Further, the SRv6 policy management module comprises:

[0049] A policy generation submodule for generating a SID list of SRv6 according to service requirements;

[0050] A policy verification submodule for verifying the syntax and semantic correctness of the policy;

[0051] A policy issuing submodule for issuing the policy to network devices through a NETCONF / YANG protocol;

[0052] A policy monitoring submodule for monitoring the execution state and effect evaluation of the policy.

[0053] Compared with the prior art, the present application has the following advantages:

[0054] 1. The present application accurately identifies and classifies service traffic through deep packet inspection and five-tuple information, and matches differentiated SRv6 tunnel models and QoS policies for the service traffic, thereby providing completely different network services for financial services of different natures such as transactions, operation and maintenance, and backup, and truly realizing network resource on-demand allocation.

[0055] 2. The present application perceives the global network state through a centralized SDN controller, and intelligently calculates paths based on multi-dimensional constraints such as delay and bandwidth, thereby avoiding local congestion. The dynamic tuning function can automatically and accurately schedule non-critical service traffic to idle paths when the link utilization rate exceeds a threshold, thereby fully exploiting network potential and avoiding waste of bandwidth resources.

[0056] 3、The dual-ring dual-plane physical architecture and the off-site disaster recovery deployment of the controller provide high reliability at the device level and the site level. In the traffic optimization process, by ensuring that the high-priority service path remains unchanged and only scheduling low-priority traffic, the network quality of key applications such as core transaction services is completely unaffected by congestion processing.

[0057] 4、The application changes the traditional network operation and maintenance mode relying on manual experience into an automatic mode based on strategy. From traffic identification, path calculation to congestion optimization, the whole process is automatically completed by the system, greatly reducing the operation and maintenance complexity, improving the fault response and processing speed, and effectively reducing the risk of business interruption caused by network problems. BRIEF DESCRIPTION OF DRAWINGS

[0058] Other features, objects and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments, made with reference to the accompanying drawings:

[0059] Figure 1 The flowchart of the embodiment of the application is shown in the figure.

[0060] Figure 2 The system architecture diagram of the embodiment of the application is shown in the figure.

[0061] Figure 3 The dual-ring dual-plane backbone network topology diagram of the embodiment of the application is shown in the figure. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical scheme and advantages of the application more clear, the application will be described in detail below with reference to the drawings and specific embodiments.

[0063] As shown in the figure, the SRv6-based financial intelligent backbone network traffic scheduling method comprises the following steps: Figure 1 Step S1, based on the five-tuple information of the deep packet inspection service flow, the ingress traffic is identified and divided into predefined service types;

[0064] Step S2, for each service type identified, a corresponding SRv6 traffic tunnel model is matched for it, the SRv6 traffic tunnel model includes a single-path multi-segment routing model for latency-sensitive services and a multi-path single-segment routing model for bandwidth-sensitive services;

[0065] Step S3, different QoS policies are allocated for different types of service flows, and the network nodes perform corresponding queue scheduling according to the QoS policies;

[0066]

[0067] ​Step S4, based on the real-time collected network state information, an optimization objective function is constructed for the SLA requirements of different service types for path calculation;

[0068] Step S5, through real-time data flow monitoring the link state of the whole network, when detecting that the link utilization rate exceeds the preset threshold, the intelligent optimization process is automatically triggered;

[0069] Step S6, the calculated optimal SRv6 strategy is encoded as a SID list, and is issued to the ingress PE device through the control protocol for execution.

[0070] The ingress PE device parses the five-tuple information of the input service packet through a deep packet inspection engine. The five-tuple includes: source IP address, destination IP address, source port number, destination port number, and protocol type.

[0071] In the step S1, the pre-defined service types include: operation and maintenance class, transaction service class, monitoring service class, backup service class, and other service class.

[0072] In the step S2, the single-path multi-segment routing model matches the transaction service class service and the operation and maintenance class service, and the multi-path single-segment routing model matches the monitoring service class service, the backup service class service, and the other service class service.

[0073] Among them, the single-path multi-segment routing model specifies the exact path through an ordered SID list; the multi-path single-segment routing model only specifies the destination SID, and utilizes ECMP for load sharing.

[0074] In the step S3, the QoS policy includes assigning different priority weights to different service classes, and configuring corresponding queue scheduling weights on the whole network device according to the priority weights.

[0075] The calculation formula of the queue scheduling weight is:

[0076]

[0077] Among them, Q i represents the queue scheduling weight of the service class i, Q i represents the service class of i, Q j represents the jth service class, Q represents the total number of service classes, P i represents the priority weight of the service class i.

[0078] For the SLA requirements of different service types, the corresponding optimization objective function is constructed for calculation, which specifically includes two modes of delay optimization calculation and bandwidth optimization calculation;

[0079] ​The time delay optimal calculation is used for time delay sensitive services, including transaction service type services and operation and maintenance type services, adopts a target function of minimizing end-to-end time delay paths, and the target function includes three components of link propagation time delay, node processing time delay and service queuing time delay;

[0080] The calculation formula of the time delay optimal target function is:

[0081]

[0082] wherein, denotes the total time delay of the path P, denotes the propagation time delay of the link e, denotes the processing time delay of the node v, denotes the service queuing time delay on the link e, denotes the addition symbol.

[0083] The problem can be converted into a shortest path problem on a weighted directed graph. Since all time delay components are non-negative values, Dijkstra algorithm is used for solving. The algorithm first initializes the time delay of all nodes to be infinite, then starts from the source node and gradually expands, each time selects the node with the minimum current time delay to perform relaxation operation, updates the time delay values of its neighbor nodes, until the shortest path to the destination node is found.

[0084] The bandwidth optimal calculation is used for bandwidth sensitive services, including monitoring service type services, backup service type services and other service type services, adopts a target function of finding path combinations meeting bandwidth requirements, and the bandwidth requirements of the services are met by calculating the sum of the available bandwidths of the paths.

[0085] The calculation formula of the bandwidth optimal target function and constraint condition is:

[0086]

[0087]

[0088] wherein, denotes the maximum available bandwidth of the path P, which is determined by the link with the minimum available bandwidth in the path, denotes the available bandwidth of the link e, denotes the bandwidth requirement of the service type .

[0089] A breadth-first search method is adopted. The available bandwidth of the link is used as the measurement standard of path selection, and the path with the maximum bandwidth among all paths meeting the bandwidth requirements is found. The algorithm maintains the maximum available bandwidth values of each node, and finds the optimal path by constantly updating these values.

[0090] When the link utilization exceeds a threshold of 80%, a dynamic tuning process is triggered.

[0091] The tuning process specifically includes the following steps:

[0092] All traffic flows passing through the congested link are identified to form a set of flows to be adjusted, and for each flow in the set, its traffic characteristics are extracted, including traffic class, traffic priority, bandwidth requirement, and latency requirement.

[0093] The traffic flows are sorted based on traffic priority and a priority threshold is set, and traffic flows with a priority lower than the threshold are included in a set of adjustable flows.

[0094] For each flow in the set of adjustable flows, a new path is recalculated, and the optimization goal is to minimize the maximum link utilization in the entire network while meeting the SLA constraints of the traffic, and the specific formula is:

[0095]

[0096] wherein, represents the maximum link utilization in the entire network after optimization, represents the utilization of link e after optimization, represents the set of adjustable traffic flows, represents the i-th flow in the set of adjustable traffic flows, represents the path of traffic flow , represents the total latency of the path of traffic flow , represents the maximum latency allowed for traffic flow , represents the available bandwidth of the path of traffic flow , represents the minimum bandwidth required for traffic flow , represents the link that is congested.

[0097] After the new path is calculated, the controller generates the corresponding SRv6 policy and issues the policy to the related PE devices for traffic switching;

[0098] The traffic switching uses a seamless switching mechanism, and the switching time includes propagation delay, processing delay, and buffering time.

[0099] The recalculation of the path for each flow in the set of adjustable flows uses a multi-stage heuristic algorithm, including a single-path search stage and a multi-path load sharing stage.

[0100] In the single-path search stage, a single replacement path is found for each flow, and all links on the new path must have a utilization lower than the threshold after the bandwidth requirement of the flow to be adjusted is added, and the calculation formula is:

[0101]

[0102] wherein, represents a single alternative path, represents the utilization of link e, represents the maximum bandwidth of link e, represents a preset threshold, which is 80%.

[0103] If the single path cannot meet the requirements, the multi-path load sharing stage is entered, the sum of the available bandwidths of multiple paths is calculated to meet the bandwidth requirement of the service, and the utilization of each path is still lower than the threshold after the load of each path is increased; K paths are calculated for sharing, and the specific formula is:

[0104]

[0105] wherein, represents the kth path of the multi-path load, represents the number of paths of the multi-path load, represents the kth path, represents the available bandwidth of the kth path.

[0106] And each path satisfies:

[0107]

[0108] wherein, represents the split ratio of the flow on the path .

[0109] wherein, the single path search stage specifically includes: constructing a cost function and calculating, selecting the path with the lowest cost value as the new path of the service flow; wherein the cost function adopts a dynamic weight mechanism, and the weights of various factors in the path cost function are automatically adjusted according to the real-time congestion degree of the network, and the specific formula is:

[0110]

[0111] wherein, represents the cost function of the path , represents a bandwidth weight coefficient, represents a utilization weight coefficient, represents the current time;

[0112] wherein, the calculation formula of the bandwidth weight coefficient and the utilization weight coefficient is:

[0113]

[0114] wherein, represents the maximum utilization of the whole network at the current time.

[0115] The controller compares the cost values of all candidate paths and selects the one with the lowest cost value as the new path for the traffic flow.

[0116] The network is heavily congested, close to or even exceeding the threshold: becomes larger, becomes smaller. The cost function focuses more on avoiding low-bandwidth bottleneck links, because it is crucial to avoid creating new congestion points at this time.

[0117] When the network is lightly loaded: the weight is relatively higher. The cost function focuses more on selecting paths with overall lower link utilization, thereby achieving better load balancing.

[0118] The multi-path load sharing phase specifically includes the following steps:

[0119] The controller generates K feasible paths to the same destination based on the network topology and constraints, forming a candidate path set;

[0120] Calculate the bottleneck bandwidth of each path, because for each path in the candidate set, its effective available bandwidth is not determined by the sum of all link bandwidths on the path, but by the link with the smallest available bandwidth on the path;

[0121] According to the available bandwidth capacity of each path, calculate the initial splitting ratio, the calculation formula is:

[0122]

[0123] wherein, represents the path the theoretical traffic proportion that should be shared;

[0124] After calculating the splitting ratio, verify whether the utilization rate constraint is met after sharing the traffic path by path and link;

[0125] If any link on any path does not meet the above constraint condition, adjustment is needed, which specifically includes: removing the path that most seriously violates the constraint from the candidate set; based on the new candidate path set, recalculate the splitting ratio of the remaining paths; repeat the verification and adjustment process until a set of paths and their splitting ratios are found that can meet all constraints.

[0126] The final output is a set of (path, splitting ratio) pairs, such as:

[0127]

[0128] The controller encodes this result into an SRv6 policy and sends it to the ingress PE device. The PE device will then distribute the service flow packets across multiple paths for transmission using mechanisms such as hashing, based on this ratio.

[0129] like Figure 2 As shown, the SRv6-based intelligent financial backbone network traffic scheduling system is implemented based on the SRv6-based intelligent financial backbone network traffic scheduling method, and includes three layers: network layer, control layer and application layer.

[0130] The network layer adopts a dual-ring dual-plane topology. Each data center has two backbone core routers and two backbone edge routers deployed to form a dual plane. High-bandwidth optical transmission network lines are mainly used between data centers, and data centers are fully interconnected.

[0131] The control layer employs an off-site disaster recovery architecture-deployed SDN controller cluster, which includes network management, network control, and network analysis.

[0132] The application layer includes various network applications developed based on the controller's northbound interface.

[0133] Thanks to its dual-ring, dual-plane topology, the third-generation backbone network not only has traditional board-level, device-level, and link-level redundancy capabilities, but also ring network redundancy and multi-plane protection capabilities. The SDN controller also adopts a multi-cluster redundancy scheme in different locations, making the high-availability architecture more complete and ensuring network robustness.

[0134] The control layer also includes:

[0135] The SRv6 policy management module is used for the generation, distribution, and management of SRv6 policies.

[0136] The intelligent path calculation module is used for path calculation based on multiple constraints.

[0137] The QoS policy management module is used for the formulation and execution of QoS policies;

[0138] The dynamic optimization module is used for automatic optimization when the link is congested.

[0139] The data acquisition module is used for real-time acquisition of network status data.

[0140] The SRv6 policy management module includes:

[0141] The strategy generation submodule is used to generate a list of SRv6 SIDs based on business requirements;

[0142] The policy validation submodule is used to validate the syntactic and semantic correctness of policies;

[0143] A policy issuing sub-module issues policies to network devices through a NETCONF / YANG protocol;

[0144] A policy monitoring sub-module is configured to monitor the execution state and effect evaluation of the policies.

[0145] As shown in Figure 3 the production backbone network (dual plane) design idea is consistent with the non-production backbone network (single plane), and the difference mainly lies in the dual plane and NCE multi-live architecture. In the dual plane architecture, the P-PE devices in each data center are deployed with dual machines to form a dual plane, which has higher availability and more flexible traffic scheduling. The NCE (SDN controller) is deployed in a geographically dispersed disaster recovery architecture, with the NCE master site deployed in Shiqiang, the standby site deployed in Ya'an, and the arbitration server deployed in Luzhou, ensuring high availability of the NCE. The CEs and PEs in each site are fully interconnected, and the PEs are also fully interconnected. The two-way adjustment of EBGP available paths between CEs and PEs is 2. Other basic routing design, tunnel design, traffic scheduling design, and QoS design are consistent with the non-production backbone network.

[0146] According to the overall architecture design, a three-layer structure design of core layer, aggregation layer, and access layer is adopted, and each layer adopts a dual plane architecture.

[0147] Core layer: composed of DC-P backbone routers in Shiqiang, Xixin, Luzhou, and Ya'an, providing high-speed forwarding and traffic scheduling between data centers;

[0148] Aggregation layer: composed of DC-PE and branch aggregation PE nodes, serving as the access and aggregation point of backbone network services;

[0149] Access layer: composed of data center core switches and branch in-line area switches, serving as the data center and branch service side.

[0150] The non-production backbone network has the same architecture as the production backbone network, with reduced device and link resources, and single plane deployment.

[0151] Traffic control mainly reflects in SRv6 tunnel design and QoS design.

[0152] In terms of SRv6 tunnel design, based on the consideration of delay priority, bandwidth priority, cross-plane scheduling and other service scenarios, two types of traffic tunnel models, "single path multi-segment routing" and "multi-path and single-segment routing", are designed, and the service traffic is divided into five categories: operation and maintenance, transaction business, monitoring business, backup business, and other business, which are classified into the matching traffic tunnel model.

[0153] In the QoS design aspect, mainly focus on the priority re-marking when the core network external traffic is injected and the QoS queue scheduling when the traffic is on the wide area link, inherit the service importance definition of the existing network, based on the consistency principle, different types of services adopt different characteristic value marking, facilitate the fine QoS policy adjustment.

[0154] The examples described in the present application are only used to describe the preferred embodiments of the present application, and do not limit the concept and scope of the present application. Without departing from the design idea of the present application, various modifications and improvements of the technical solutions of the present application made by the engineering technicians in the art shall fall within the protection scope of the present application.

Claims

1. A method for SRv6-based financial intelligent backbone network traffic scheduling, characterized in that, The method comprises the following steps: Step S1, identifying and classifying the ingress traffic into predefined service types based on the five-tuple information of the deep packet inspection service flow; Step S2, matching a corresponding SRv6 traffic tunnel model for each identified service type, wherein the SRv6 traffic tunnel model comprises a single-path multi-segment routing model for latency-sensitive services and a multi-path single-segment routing model for bandwidth-sensitive services; Step S3, assigning different QoS policies to different categories of service flows, and performing corresponding queue scheduling at network nodes according to the QoS policies; Step S4, constructing a corresponding optimization objective function for path calculation based on real-time collected network state information and SLA requirements of different service types; Step S5, monitoring the link state of the whole network through real-time data flow, and automatically triggering an intelligent optimization process when detecting that the link utilization rate exceeds a preset threshold; Step S6, encoding the calculated optimal SRv6 strategy into a SID list and issuing it to the ingress PE device for execution through a control protocol; In step S4, the corresponding optimization objective function is constructed for calculation according to the SLA requirements of different service types, specifically including two modes of latency optimization calculation and bandwidth optimization calculation; The latency optimization calculation is used for latency-sensitive services, including transaction service type services and operation and maintenance type services, and adopts an objective function of minimizing the end-to-end latency path, which includes three components of link propagation delay, node processing delay and service queuing delay; The bandwidth optimization calculation is used for bandwidth-sensitive services, including monitoring service type services, backup service type services and other service type services, and adopts an objective function of finding a path combination that meets the bandwidth requirement, and the sum of the available bandwidths of the paths is calculated to meet the bandwidth requirement of the service; The calculation formula of the latency optimization objective function is: ； wherein, denotes the total latency of path P, denotes the propagation latency of link e, denotes the processing latency of node v, denotes the traffic the queuing latency on link e, denotes the additive symbol; The calculation formula of the bandwidth optimization objective function and the constraint condition is: ； ； wherein denotes the maximum available bandwidth of the path P, determined by the link with the smallest available bandwidth in the path, denotes the available bandwidth of the link e, denotes the bandwidth requirement of the traffic type .

2. The method of claim 1, wherein, In step S1, the predefined service types include operation and maintenance type, transaction service type, monitoring service type, backup service type and other service type.

3. The method of claim 2, wherein, In step S2, the single-path multi-segment routing model matches transaction service type services and operation and maintenance type services, and the multi-path single-segment routing model matches monitoring service type services, backup service type services and other service type services; The single-path multi-segment routing model specifies the exact path through an ordered SID list; the multi-path single-segment routing model only specifies the destination SID and uses ECMP for load sharing.

4. The method of claim 3, wherein, In step S3, the QoS policy includes assigning different priority weights to different service categories, and configuring corresponding queue scheduling weights at network devices according to the priority weights.

5. The method of claim 4, wherein, In step S5, the optimization process specifically includes: Identifying all service flows passing through congested links to form a set of flows to be adjusted, extracting the service characteristics of each flow in the set of flows to be adjusted, including service category, service priority, bandwidth requirement and latency requirement; Sorting based on service priority and setting a priority threshold, and including service flows with a priority lower than the threshold in the set of adjustable flows; Re-calculate the path for each flow in the adjustable flow set, and the optimization target is to minimize the maximum link utilization rate in the whole network while meeting the service SLA constraints; After calculating the new path, the controller generates the corresponding SRv6 policy and issues the policy to the related PE device for traffic switching; The traffic switching adopts a seamless switching mechanism, and the switching time includes propagation delay, processing delay and buffering time.

6. The method of claim 5, wherein, The re-calculation of the path for each flow in the adjustable flow set adopts a multi-stage heuristic algorithm, including a single-path search stage and a multi-path load sharing stage; In the single-path search stage, a single replacement path is found for each flow, and the utilization rate of all links on the new path is still below the threshold value after increasing the bandwidth demand of the flow to be adjusted, and the calculation formula is: ； wherein, represents a single alternative path, represents the utilization of link e, represents a traffic flow the minimum bandwidth required, represents the maximum bandwidth of link e, represents a preset threshold, which is 80%. If the single path cannot meet the requirements, the multi-path load sharing stage is entered, and the sum of the available bandwidths of multiple paths is calculated to meet the bandwidth demand of the service, and the utilization rate of each path is still below the threshold value after the load is increased, and the specific formula is: ； wherein, represents the kth path of the multipath load, represents the number of paths of the multipath load, represents the kth path, represents the available bandwidth of the kth path, and each path satisfies: ； wherein represents the proportion of flow split on the path ; The single-path search stage specifically includes: constructing a cost function and calculating, selecting the path with the lowest cost value as the new path of the service flow; wherein the cost function adopts a dynamic weight mechanism to automatically adjust the weight of each factor in the path cost function according to the real-time congestion degree of the network; The specific formula of the cost function is: ； wherein, a cost function of a path, a bandwidth weight coefficient, a utilization weight coefficient, a current time;​ The calculation formula of the bandwidth weight coefficient and the utilization rate weight coefficient is: ； wherein, represents the maximum utilization of the entire network at the current time; The multi-path load sharing stage specifically includes the following steps: The controller generates K feasible paths leading to the same destination based on the network topology to form a candidate path set; Calculate the bottleneck bandwidth of each path; According to the available bandwidth capacity of each path, calculate the initial shunt ratio, and the calculation formula is: ； wherein representing a path theoretically due share of flow rate; Verify whether the utilization rate constraint is met under the initial shunt ratio path by path and link by link; If any link on any path does not meet the utilization rate constraint, adjustment is needed until a set of paths and their shunt ratios are found to meet all the constraints.

7. The SRv6-based financial intelligent backbone network traffic scheduling system is implemented based on the SRv6-based financial intelligent backbone network traffic scheduling method as claimed in claims 1-6, characterized in that, It includes: Three levels of network layer, control layer and application layer; The network layer adopts a double-ring double-plane topology structure, and the core router and edge router devices of the backbone network of each data center are deployed with double machines to form a double plane, and the main use of the network line between the data centers is a large bandwidth optical transmission network, and the data centers are fully interconnected; The control layer adopts a SDN controller cluster deployed in a geographically dispersed disaster recovery architecture, including network management, network control and network analysis; The application layer includes various network applications developed based on the northbound interface of the controller.

8. The system of claim 7, wherein, The control layer further includes: An SRv6 policy management module for generating, issuing and managing SRv6 policies; An intelligent path calculation module for path calculation based on multiple constraints; A QoS policy management module for formulating and executing QoS policies; A dynamic tuning module for automatic tuning when the link is congested; A data acquisition module for real-time acquisition of network state data.

9. The system of claim 8, wherein, The SRv6 policy management module includes: A policy generation submodule for generating a SID list of SRv6 according to service requirements; The policy verification submodule is configured to verify the syntax and semantic correctness of the policy. The policy issuing submodule is configured to issue the policy to the network device through the NETCONF / YANG protocol. The policy monitoring submodule is configured to monitor the execution state and effect evaluation of the policy.

Images