CN-122027567-A - Financial load balancing method and system with cooperation of application layer and network layer

CN122027567ACN 122027567 ACN122027567 ACN 122027567ACN-122027567-A

Abstract

The invention discloses a financial load balancing method and a financial load balancing system for cooperation of an application layer and a network layer, and belongs to the technical field of networks. The method comprises the steps of deploying an application layer load balancing node, an SDN controller, a collaborative decision-making module, a flow prediction module and a state acquisition module, establishing communication connection, acquiring state indexes of an application layer server, link state indexes of a network layer, financial transaction priorities and protocol types, whole network flow data and predicting flow change trend in real time, carrying out linkage analysis on multidimensional indexes by the collaborative decision-making module to generate synchronous application layer scheduling instructions and network layer scheduling instructions, respectively issuing the instructions to the application layer load balancing node and the SDN controller for execution, feeding back the acquired and executed indexes to the collaborative decision-making module, and dynamically adjusting the instructions according to index difference. The invention realizes index intercommunication and instruction synchronization of the application layer and the network layer, forms closed loop control, and improves service quality and operation stability of the financial SDN network.

Inventors

XU JIAN
YU YANG
ZHANG XIAOQIAN
ZHANG RONGLIANG
WANG ZHAOTAO
SHI XIAOWEI
CUI HAIBIN
GAO XIANXIAN
CHANG HONGYU
LIU XUANDONG

Assignees

山东省城市商业银行合作联盟有限公司

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (10)

1. The financial load balancing method for cooperation of the application layer and the network layer is characterized by comprising the following steps: Deploying a collaborative architecture to complete initialization configuration, namely deploying an application layer load balancing node, an SDN controller, a collaborative decision module, a flow prediction module and a state acquisition module, and establishing communication connection between the collaborative decision module and the application layer load balancing node, the SDN controller, the flow prediction module and the state acquisition module respectively; the multi-dimensional index real-time acquisition comprises the steps of acquiring an application layer server state index and a network layer link state index in real time through the state acquisition module, acquiring transaction priority and protocol type of a financial transaction request in real time through the application layer load balancing node, acquiring whole network flow data in real time through the flow prediction module, and predicting a flow change trend in a future preset time window by combining historical flow data; The collaborative decision module performs multidimensional linkage analysis on the acquired state index of the application layer server, the state index of the network layer link, the financial transaction priority, the dual-stack flow state and the flow prediction data, and generates a synchronous application layer scheduling instruction and a network layer scheduling instruction according to an analysis result; and the cooperative execution and closed loop feedback are that the application layer scheduling instruction is issued to the application layer load balancing node, the network layer scheduling instruction is issued to the SDN controller, the running index after the scheduling execution is acquired in real time through the state acquisition module and fed back to the cooperative decision module, and the cooperative decision module dynamically adjusts the scheduling instruction according to the index difference before and after the scheduling and issues the scheduling instruction again for execution.
2. The method for financial load balancing in coordination with an application layer and a network layer according to claim 1, wherein the multidimensional linkage analysis comprises: normalizing the collected indexes; Calculating the comprehensive health degree of the server according to the normalized state index of the application layer server; Calculating the comprehensive quality of the link according to the normalized network layer link state index; Determining a priority quantization coefficient according to the financial transaction priority; Dynamically adjusting the IPv4/IPv6 shunt proportion according to the dual stack flow state and the dual stack flow adaptation rule; according to the flow prediction data, when the predicted future flow peak exceeds a preset threshold value, the preferred strategy and the bandwidth allocation strategy of the server-link combination are adjusted in advance; And calculating the comprehensive score of each candidate server-link combination according to the comprehensive health degree of the server, the comprehensive quality of the link and the priority quantization coefficient, and selecting the server-link combination with the highest comprehensive score as an optimal scheduling target.
3. The method for balancing financial loads by cooperation of an application layer and a network layer according to claim 1, wherein the application layer server state indexes comprise CPU utilization rate, memory occupancy rate, disk IO utilization rate, application concurrency connection number, transaction response time and protocol types supported by a server, the network layer link state indexes comprise link bandwidth utilization rate, link delay, link jitter, link packet loss rate, network port occupancy rate and protocol types supported by a link, the transaction priority comprises core transaction, important transaction and common transaction, and the protocol types comprise IPv4 and IPv6.
4. The financial load balancing method of cooperation between an application layer and a network layer according to claim 1, wherein the traffic prediction module predicts a traffic peak value, a peak occurrence time and a traffic growth rate within a future preset time window based on historical traffic data and real-time acquired whole network traffic data, wherein the whole network traffic data comprises total traffic, IPv4 traffic, IPv6 traffic and various transaction request amounts, by adopting a long-short-term memory network prediction model.
5. The method for financial load balancing between an application layer and a network layer according to claim 1, wherein dynamically adjusting the IPv4/IPv6 offload ratio comprises: when the protocol suitability conflict is detected, the distribution proportion is adjusted and the protocol conversion gateway is started; when interference conflict among link stacks is detected, the split ratio is adjusted to relieve the link resource competition; and when the route concussion conflict is detected, locking the diversion ratio and prohibiting the frequent switching of the route.
6. The method for balancing financial load coordinated with an application layer and a network layer according to claim 5, wherein dynamically adjusting the IPv4/IPv6 split ratio further comprises adopting a flow-based migration mechanism to keep an original path unchanged for an established active connection, distributing a path for a newly initiated connection according to the adjusted split ratio, adopting a double-sending mechanism and a slow-starting mechanism to carry out smooth migration, simultaneously sending data packets to the original path and the new path at the initial stage of migration, and gradually increasing the flow ratio of the new path after the new path confirms normal reception.
7. The method for balancing financial load by cooperation of an application layer and a network layer according to claim 1, wherein in the cooperative execution and closed loop feedback, when an application server or a network link failure is detected, switching of an application layer standby server and switching of a network layer standby link are triggered synchronously, and after failure recovery, traffic is switched back smoothly by a slow start mode.
8. A financial load balancing system for cooperation between an application layer and a network layer, comprising: the collaborative architecture module comprises an application layer load balancing unit, an SDN control unit, a collaborative decision unit, a flow prediction unit and a state acquisition unit, wherein the collaborative decision unit is respectively in communication connection with the application layer load balancing unit, the SDN control unit, the flow prediction unit and the state acquisition unit; The initialization configuration module is integrated in the collaborative decision-making unit and is used for presetting financial transaction priority rules, load thresholds, dual-stack flow adaptation rules and interface association configuration; the state acquisition module is deployed on the application server and the network link node and is used for acquiring the state index of the application layer server and the state index of the network layer link in real time; The transaction information acquisition module is integrated with the application layer load balancing unit and is used for acquiring the transaction priority and the protocol type of the financial transaction request in real time; The flow prediction module is used for collecting the flow data of the whole network in real time and predicting the future flow change trend through the built-in prediction model; The collaborative decision-making module is used for carrying out linkage analysis on the multidimensional index to generate a synchronous application layer scheduling instruction and a synchronous network layer scheduling instruction; The instruction execution module comprises an application layer instruction execution unit and a network layer instruction execution unit, and is used for executing an application layer scheduling instruction and a network layer scheduling instruction respectively; and the closed loop feedback module is integrated in the collaborative decision unit and is used for comparing the index difference before and after scheduling and dynamically adjusting the scheduling instruction.
9. The application layer and network layer collaborative financial load balancing system according to claim 8, wherein the collaborative decision module comprises: The data preprocessing unit is used for carrying out normalization processing on all acquired indexes; The server health degree calculation unit is used for calculating the comprehensive health degree of the server according to the normalized state index of the application layer server; the link quality calculation unit is used for calculating the comprehensive link quality according to the normalized network layer link state index; A priority quantization unit for determining a priority quantization coefficient according to the priority of the financial transaction; The dual-stack cooperative scheduling unit is used for dynamically adjusting the IPv4/IPv6 split ratio according to the dual-stack flow state and the dual-stack flow adaptation rule; the traffic pre-scheduling unit is used for pre-adjusting a server-link combination optimization strategy and a bandwidth allocation strategy according to traffic prediction data when a predicted future traffic peak exceeds a preset threshold; A combination optimization unit, configured to calculate a combination composite score of each candidate server-link pair, and select a server-link combination with the highest combination composite score as an optimal scheduling target; And the instruction generation unit is used for generating synchronous application layer scheduling instructions and network layer scheduling instructions.
10. The application layer and network layer collaborative financial load balancing system according to claim 8, further comprising a collaborative fault tolerance module comprising: the fault detection unit is used for detecting the fault state of the application server or the network link; the synchronous switching unit is used for synchronously triggering the switching of the standby server of the application layer and the switching of the standby link of the network layer when the fault is detected; And the back-cut control unit is used for smoothly back-cutting the flow in a slow start mode after the fault is recovered.

Description

Financial load balancing method and system with cooperation of application layer and network layer Technical Field The invention relates to the technical field of networks, in particular to a financial load balancing method and a financial load balancing system for cooperation of an application layer and a network layer. Background A Software Defined Network (SDN) achieves flexible scheduling and centralized management of network traffic by separating control from forwarding. In high reliability scenarios such as the financial industry, SDN is widely used to build efficient network infrastructure. However, as network scales expand, single controller ease becomes a performance bottleneck, and multi-controller distributed deployment becomes the mainstream. Under the architecture, how to realize load balancing among controllers is a key for guaranteeing network performance. In the prior art, a scheme for realizing load balancing of an SDN controller through prediction and switch migration is available. For example, chinese patent publication No. CN114338537B discloses a method and system for migration of SDN load balancing dual-weight switches based on prediction. According to the method, the controller load is monitored in real time through the flow collection component, the flow prediction component is utilized to predict future loads, and when the controller is predicted to be overloaded, a double-weight switch migration method is adopted to migrate the switch to a target controller with a lower load, so that the loads among the controllers are balanced. However, the above prior art mainly focuses on load balancing of the network layer controller, and its decision process depends only on network layer metrics, such as packet_in message arrival rate. This decision mechanism has significant limitations in the financial SDN environment: First, the decision dimension is single, and the cooperation between the application layer and the network layer cannot be realized. The prior art does not consider the real-time running state of the application server, such as CPU utilization rate, memory occupancy rate, transaction response time and the like. This may result in transaction requests being distributed to paths where the server load is low but the network link is congested, resulting in transaction churning or timeout, whereas good links scheduled by the network layer may also result in wasted resources due to the connected server having reached the upper performance limit. The architecture with the disjointed decision of the application layer and the network layer cannot realize the optimal configuration of the whole system resources. Second, there is a lack of traffic priority awareness and is not adaptable to IPv6 dual stack environments. On one hand, financial transactions have clear priority differences, such as core transactions like large-amount transfer and the like, have far higher requirements on time delay and success rate than common inquiry, the prior art adopts a unified strategy for all traffic and cannot reserve high-quality resources for the core transactions, and on the other hand, as IPv6 is deeply modified, a large number of dual stack traffic exists in a network, the prior art does not consider dual stack cooperative scheduling, traffic conflict and route oscillation can be caused due to problems of protocol suitability mismatch, inter-link stack interference and the like, and data packet loss and retransmission are caused. Third, the scheduling policy is stiff and lacks active prediction and dynamic adaptation capabilities. The prior art adopts 'passive response' load balancing, namely, the scheduling is triggered after overload or congestion, and the active prediction and pre-scheduling capability of the flow peak value is lacked. Meanwhile, most of scheduling strategies are static configuration, cannot be dynamically adjusted according to real-time states, and are difficult to deal with traffic burst scenes such as payday, annual end settlement and the like, so that operation and maintenance cost is high and scheduling effect is poor. In summary, how to break the decision barrier of load balancing between the application layer and the network layer, realize multidimensional collaborative awareness and scheduling, and adapt to the dual stack environment of financial service priority and IPv6, and simultaneously have the capability of active prediction and dynamic self-adaptive scheduling, so as to improve the overall service quality and operation stability of the financial SDN network, which is a technical problem to be solved in the art. Disclosure of Invention The invention aims to solve the problems that in the prior art, load balancing decision of an application layer and a network layer is disjoint, financial service priority and IPv6 dual stack environment cannot be adapted, scheduling strategies are stiff and the like, and provides a financial load balancing method and a financial l