Network system and method for balancing traffic load

Abstract

The invention discloses a flow load balancing network system and method, the system comprises a user autonomous domain and an operator autonomous domain, and the user autonomous domain comprises a core layer, a convergence layer and an access layer; the core layer comprises two core routers, a network management switch and a plurality of network management servers, is connected to a network of an operator and is used for being responsible for flow forwarding of a core network; the convergence layer comprises four convergence routers and is used for converging user traffic of the access layer and transferring the user traffic to the core layer; the access layer comprises a plurality of access switches and is used for forwarding access user traffic to the convergence layer; the core router communicates with the convergence router through an ISIS and communicates with the operation router through an EBGP protocol; the access switch is configured with transparent transmission and access of a VLAN (Virtual Local Area Network) and related ports; and designing a DMtwo traffic classification method to carry out classification and load balancing on traffic. According to the method, a DMtwo flow classification method is adopted, multi-outlet network flow load balancing is achieved, and the method has the advantages of dynamic adjustment and fine management.

Description

Technical Field

[0001] The present invention relates to the field of computer network technology, and in particular to a network system and method for traffic load balancing. Background Art

[0002] With the rapid development of economy and technology, IP network has gradually evolved from the original single data network to a multi-service network integrating data, voice, image and video. Modern enterprises have increasing requirements for network services, not only requiring higher bandwidth, but also requiring excellent network performance indicators such as low latency, jitter and packet loss rate during data transmission. However, the traditional single-exit enterprise network has significant deficiencies in these aspects, making it difficult to meet the growing needs of enterprises, which in turn affects the office efficiency and business processing capabilities of enterprises.

[0003] In order to solve the bottleneck problem of single-exit networks, enterprises have begun to adopt multi-exit network systems in order to achieve traffic dispersion and load balancing through multiple network exits. However, in actual applications, multi-exit network systems often face the challenge of unbalanced traffic load. Currently, multi-exit network systems mainly rely on static routing strategies or simple load balancing algorithms to achieve load balancing. These methods have the following defects in actual applications:

[0004] 1. Static configuration is difficult to adapt to dynamic traffic: Traditional static routing configuration cannot dynamically adjust traffic distribution according to real-time traffic changes, resulting in some links being overloaded for a long time, while other links have low utilization. This not only wastes communication link resources, but may also cause serious congestion on some links, affecting overall network performance.

[0005] 2. Lack of effective redundant protection: In traditional multi-exit network systems, when a link fails, there is often a lack of effective mechanisms to quickly switch traffic to other available links, resulting in network service interruption and affecting the normal operation of the enterprise. In this case, the reliability and stability of the network are difficult to guarantee.

[0006] 3. High management complexity: Multi-exit network systems require frequent manual adjustments to routing policies, and network administrators need to continuously optimize configurations based on traffic changes and network conditions. This not only increases management complexity and maintenance costs, but also makes configuration errors more likely to occur, further affecting the stable operation of the network.

[0007] 4. The granularity of traffic classification and adjustment is relatively coarse: In the existing technology, the classification and adjustment of traffic are often based on relatively rough standards, which makes it difficult to achieve refined management. As a result, the distribution of network traffic often cannot achieve the best load balancing effect, affecting the performance and efficiency of the overall network.

[0008] Therefore, how to provide a network system and method for traffic load balancing is an urgent problem that those skilled in the art need to solve. Summary of the invention

[0009] One object of the present invention is to propose a network system and method for traffic load balancing. The present invention adopts the DMtwo traffic classification method to achieve multi-export network traffic load balancing, and has the advantages of dynamic adjustment, redundant protection and refined management. By providing an efficient, reliable and easy-to-manage traffic load balancing network system and method, it not only overcomes many defects in the prior art, but also achieves many beneficial effects, significantly improves the transmission efficiency, reliability and stability of the network, and fully meets the needs of modern enterprises for high-performance network services.

[0010] On the one hand, a network system for traffic load balancing according to an embodiment of the present invention includes a user autonomous domain and an operator autonomous domain, wherein the user autonomous domain includes a core layer, a convergence layer, and an access layer;

[0011] The core layer includes two core routers, a network management switch and several network management servers, and is connected to the operator's network to forward traffic on the core network.

[0012] The aggregation layer includes four aggregation routers, which are responsible for aggregating user traffic of the access layer and transferring it to the core layer;

[0013] The access layer includes a number of access switches, which are responsible for forwarding access user traffic to the aggregation layer;

[0014] The core router and the aggregation router communicate via ISIS;

[0015] The core router communicates with the operation router through the EBGP protocol on the basis of being reachable through the static routing protocol;

[0016] The aggregation router is configured with a sub-interface to terminate the VLAN transmitted from the access switch;

[0017] The access switch is configured with transparent transmission and access of VLAN and related ports;

[0018] The DMtwo traffic classification method is designed to classify and load balance traffic.

[0019] Furthermore, the network system has the following conditions:

[0020] The network system has two circuits for load balancing network traffic;

[0021] The egress circuits of the network system have redundant protection, and when any circuit fails, traffic will be switched to another circuit;

[0022] The traffic of the network system is finely adjusted according to the IP address.

[0023] Furthermore, the network management server collects the source address and destination address of the data packets of the AS in real time, and classifies them by the DMtwo method.

[0024] Furthermore, the EBGP routing strategy on the core router includes the following steps:

[0025] Divide the source address set into two sets Sou1 and Sou2 according to the DMtwo method;

[0026] Set the MED value of Sou1 and the MED value of Sou2 on core router 1, and Sou1<Sou2;

[0027] Set the MED value of Soul and the MED value of Sou2 on core router 2, and Sou1>Sou2;

[0028] According to the DMtwo algorithm, the destination address set is divided into two sets Des1 and Des2;

[0029] On core router 1, the local priority of Des1 and the local priority of Des2 are set, and Des1<Des2;

[0030] The local priority of Des1 and the local priority of Des2 are set on the core router 2, and Des1>Des2.

[0031] Furthermore, the DMtwo traffic classification method specifically includes:

[0032] Divide the packets entering the AS into two sets of approximately the same size based on the source address;

[0033] Divide the packets leaving the AS into two sets of approximately the same size based on the destination address;

[0034] Create a routing policy Policy1 for the source address set based on the MED attribute, and create a routing policy Policy2 for the source address set based on the MED attribute;

[0035] The Policy1 and the Policy2 are sent to the core router for traffic forwarding and receiving.

[0036] Furthermore, the algorithm design and implementation of the DMtwo traffic classification method specifically includes:

[0037] Assume two arrays Addr1 and Addr2, the array lengths are L1 and L2 respectively, the two arrays store the network addresses detected by the network management program, and SAddr1>=SAddr2, where SAddr1 represents the sum of the element lengths of the array Addr1, and SAddr2 represents the sum of the element lengths of the data Addr2;

[0038] When SAddr1 == SAddr2, no exchange is required and the minimum difference is 0;

[0039] When SAddr1>SAddr2, let the difference SUB=SAddr1-SAddr2, and continuously reduce the value of SUB by exchanging the elements in the two arrays;

[0040] Each time, an element Addr1[i] and Addr2[j] are selected from array Addr1 and array Addr2 respectively for exchange. The new difference SUB after the exchange cha for:

[0041] SUB cha =(SAddr1-Addr1[i]+Addr2[j])-(SAddr2-Addr1[i]+Addr2[j])=SUB-2×(Addr1[i]-Addr2[j]);

[0042] Among them, SUB cha Indicates the new difference, Addr1[i] indicates the length of the element with subscript position i in array Addr1, and Addr2[j] indicates the length of the element with subscript position j in array Addr2;

[0043] The above exchange process is repeated until no Addr1[i] and Addr2[j] that meet the conditions for exchange are found; the conditions are SUB>=2×(Addr1[i]-Addr2[j]) and Addr1[i]>Addr2[j].

[0044] On the other hand, a method for traffic load balancing according to an embodiment of the present invention includes the following steps:

[0045] S1. Create a network model to determine the specific device configuration and connection relationship of the core layer, aggregation layer and access layer;

[0046] S2, deploy network protocols, configure the EBGP protocol between the core router and the aggregation router and the ISIS protocol between the aggregation router and the core router;

[0047] S3, real-time monitoring of traffic, using the network management server to collect the source address and destination address of the data packet;

[0048] S4. Generate two source address sets and two destination address sets using the DMtwo algorithm;

[0049] S5, generating routing policies Policy1 and Policy2 according to the source address set, and generating corresponding routing policies according to the destination address set;

[0050] S6, sending the Policy1 and the Policy2 to the core router to forward and receive traffic;

[0051] S7, set the MED value and local priority value on core router 1 and core router 2 to achieve load balancing of traffic on the two egress circuits;

[0052] S8. Refresh the routing policy regularly, regenerate and issue a new routing policy every 30 minutes or when congestion is detected on the egress circuit;

[0053] S9. Adjust traffic classification and routing strategies according to network operation status;

[0054] S10. During the operation of the network system, continuously monitor the traffic and network performance, and perform manual adjustments and optimizations when necessary.

[0055] Further, the S10 specifically includes:

[0056] S101. Use network performance monitoring tools to detect network parameters in real time, including network packet loss rate, jitter and delay parameters;

[0057] S102, when abnormal parameters are detected, automatically generating an alarm and triggering a flow optimization program;

[0058] S103. Analyze historical and real-time data through machine learning algorithms to predict traffic peaks and make adjustments in advance;

[0059] S104. Generate network performance reports regularly, evaluate the effectiveness of traffic load balancing strategies, and make adjustments and improvements based on the evaluation results.

[0060] The beneficial effects of the present invention are:

[0061] (1) Dynamically adjust traffic distribution: The network system of the present invention can dynamically generate and adjust traffic classification strategies based on real-time collected network data through the DMtwo traffic classification method. This can ensure that traffic is evenly distributed among multiple outlets, avoid the problem of long-term overload of some links while other links are idle, improve the utilization of network resources, and avoid the waste of communication link resources.

[0062] (2) Improving the redundant protection capability of the network: The present invention designs a network system with redundant protection function. When any circuit fails, the system can quickly switch the traffic to another circuit to ensure the continuity and stability of network services. This not only improves the reliability of the network, but also reduces the business interruption caused by network failures, ensuring the normal operation of the enterprise.

[0063] (3) Simplified management and maintenance: Through the automated traffic classification and routing policy generation mechanism, the present invention reduces the complexity and error risk of manual configuration. Network administrators no longer need to frequently adjust routing policies manually, thereby greatly improving the efficiency and accuracy of network management and reducing maintenance costs.

[0064] (4) Refined traffic management: Using the DMtwo algorithm, the present invention can finely classify and adjust the traffic entering and leaving the autonomous system, achieving fine-grained control of network traffic. This can not only improve the effect of load balancing, but also make fine adjustments based on IP addresses, ensuring that the network system has greater flexibility and adaptability.

[0065] (5) Improve network performance: Through reasonable traffic classification and load balancing strategies, the present invention significantly reduces the delay, jitter, packet loss rate and other undesirable phenomena in the network transmission process. The transmission efficiency of the network is improved, thereby improving the overall network performance and meeting the needs of modern enterprises for high-performance network services.

[0066] (6) Enhanced network security: Through multi-exit design and redundant protection mechanism, the present invention not only improves the reliability of the network, but also enhances the security of the network. Even if a link is attacked or fails, the system can quickly switch to other links to ensure the continuity of network services and reduce security risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0068] Figure 1 A system topology diagram of a network system and method for traffic load balancing proposed by the present invention;

[0069] Figure 2 A network system operation flow chart of a network system and method for traffic load balancing proposed by the present invention;

[0070] Figure 3 A network traffic diagram of a network system and method for traffic load balancing proposed by the present invention;

[0071] Figure 4 This is a diagram showing the effect of inbound and outbound traffic load balancing of a network system and method for traffic load balancing proposed by the present invention. DETAILED DESCRIPTION

[0072] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0073] refer to Figure 1-4 , On the one hand, a network system for traffic load balancing according to an embodiment of the present invention includes a user autonomous domain and an operator autonomous domain, wherein the user autonomous domain includes a core layer, a convergence layer, and an access layer;

[0074] The core layer includes two core routers, a network management switch and several network management servers, and is connected to the operator's network to forward traffic on the core network.

[0075] The aggregation layer includes four aggregation routers, which are responsible for aggregating user traffic of the access layer and transferring it to the core layer;

[0076] The access layer includes a number of access switches, which are responsible for forwarding access user traffic to the aggregation layer;

[0077] The core router and the aggregation router communicate via ISIS;

[0078] The core router communicates with the operation router through the EBGP protocol on the basis of being reachable through the static routing protocol;

[0079] The aggregation router is configured with a sub-interface to terminate the VLAN transmitted from the access switch;

[0080] The access switch is configured with transparent transmission and access of VLAN and related ports;

[0081] The DMtwo traffic classification method is designed to classify and load balance traffic.

[0082] In this implementation, the network system meets the following conditions:

[0083] The network system has two circuits for load balancing network traffic;

[0084] The egress circuits of the network system have redundant protection, and when any circuit fails, traffic will be switched to another circuit;

[0085] The traffic of the network system is finely adjusted according to the IP address.

[0086] In this implementation manner, the network management server collects the source address and destination address of the data packets of the AS in real time, and classifies them by using the DMtwo method.

[0087] In this implementation manner, the EBGP routing strategy on the core router includes the following steps:

[0088] Divide the source address set into two sets Sou1 and Sou2 according to the DMtwo method;

[0089] Set the MED value of Sou1 and the MED value of Sou2 on core router 1, and Sou1<Sou2;

[0090] Set the MED value of Sou1 and the MED value of Sou2 on core router 2, and Sou1>Sou2;

[0091] According to the DMtwo algorithm, the destination address set is divided into two sets Des1 and Des2;

[0092] On core router 1, the local priority of Des1 and the local priority of Des2 are set, and Des1<Des2;

[0093] The local priority of Des1 and the local priority of Des2 are set on the core router 2, and Des1>Des2.

[0094] In this implementation, the DMtwo traffic classification method specifically includes:

[0095] Divide the packets entering the AS into two sets of approximately the same size based on the source address;

[0096] Divide the packets leaving the AS into two sets of approximately the same size based on the destination address;

[0097] Create a routing policy Policy1 for the source address set based on the MED attribute, and create a routing policy Policy2 for the source address set based on the MED attribute;

[0098] The Policy1 and the Policy2 are sent to the core router for traffic forwarding and receiving.

[0099] In this implementation, the algorithm design and implementation of the DMtwo traffic classification method specifically includes:

[0100] Assume two arrays Addrl and Addr2, the array lengths are L1 and L2 respectively, the two arrays store the network addresses detected by the network management program, and SAddr1>=SAddr2, where SAddr1 represents the sum of the element lengths of the array Addr1, and SAddr2 represents the sum of the element lengths of the data Addr2;

[0101] When SAddr1 == SAddr2, no exchange is required and the minimum difference is 0;

[0102] When SAddr1>SAddr2, let the difference SUB=SAddr1-SAddr2, and continuously reduce the value of SUB by exchanging the elements in the two arrays;

[0103] Each time, an element Addr1[i] and Addr2[j] are selected from array Addr1 and array Addr2 respectively for exchange. The new difference SUB after the exchange cha for:

[0104] SUB cha =(SAddr1-Addr1[i]+Addr2[j])-(SAddr2-Addr1[i]+Addr2[j])=SUB-2×(Addr1[i]-Addr2[j]);

[0105] Among them, SUB cha Indicates the new difference, Addr1[i] indicates the length of the element with subscript position i in array Addr1, and Addr2[j] indicates the length of the element with subscript position j in array Addr2;

[0106] The above exchange process is repeated until no Addr1[i] and Addr2[j] that meet the conditions for exchange are found; the conditions are SUB>=2×(Addr1[i]-Addr2[j]) and Addr1[i]>Addr2[j].

[0107] On the other hand, a method for traffic load balancing according to an embodiment of the present invention includes the following steps:

[0108] S1. Create a network model to determine the specific device configuration and connection relationship of the core layer, aggregation layer and access layer;

[0109] S2, deploy network protocols, configure the EBGP protocol between the core router and the aggregation router and the ISIS protocol between the aggregation router and the core router;

[0110] S3, real-time monitoring of traffic, using the network management server to collect the source address and destination address of the data packet;

[0111] S4. Generate two source address sets and two destination address sets using the DMtwo algorithm;

[0112] S5, generating routing policies Policy1 and Policy2 according to the source address set, and generating corresponding routing policies according to the destination address set;

[0113] S6, sending the Policy1 and the Policy2 to the core router to forward and receive traffic;

[0114] S7, set the MED value and local priority value on core router 1 and core router 2 to achieve load balancing of traffic on the two egress circuits;

[0115] S8. Refresh the routing policy regularly, regenerate and issue a new routing policy every 30 minutes or when congestion is detected on the egress circuit;

[0116] S9. Adjust traffic classification and routing strategies according to network operation status;

[0117] S10. During the operation of the network system, continuously monitor the traffic and network performance, and perform manual adjustments and optimizations when necessary.

[0118] In this implementation manner, the S10 specifically includes:

[0119] S101. Use network performance monitoring tools to detect network parameters in real time, including network packet loss rate, jitter and delay parameters;

[0120] S102, when abnormal parameters are detected, automatically generating an alarm and triggering a flow optimization program;

[0121] S103. Analyze historical and real-time data through machine learning algorithms to predict traffic peaks and make adjustments in advance;

[0122] S104. Generate network performance reports regularly, evaluate the effectiveness of traffic load balancing strategies, and make adjustments and improvements based on the evaluation results.

[0123] Embodiment 1:

[0124] In order to verify the feasibility of the present invention in implementation, the present invention is applied to a certain network system.

[0125] The router in this network system is H3C 3620, the switch is H3C 3100, and the network management server is DELL 2950. The network topology is shown in the figure below. Figure 1 The basic idea of ​​the router sending EBGP routing policy is:

[0126] (1) Divide the source address set into two sets Sou1 and Sou2 according to the DMtwo algorithm;

[0127] (2) Set the MED value of Sou1 to 100 and the MED value of Sou2 to 1000 on core router 1; set the MED value of Sou1 to 1000 and the MED value of Sou2 to 100 on core router 2. In this way, the inbound traffic to Sou1 flows in from core router 1, and the inbound traffic to Sou2 flows in from core router 2.

[0128] (3) Divide the destination address set into two sets Des1 and Des2 according to the DMtwo algorithm;

[0129] (4) On core router 1, set the local priority value of Des1 to 1000 and the local priority value of Des2 to 100; on core router 2, set the local priority value of Des1 to 100 and the local priority value of Des2 to 1000; in this way, the outbound traffic accessing Des1 flows out from core router 1, and the outbound traffic accessing Des2 flows out from core router 2.

[0130] (5) Refresh and send routing policies every 30 minutes or when there is a congestion alarm on the egress circuit.

[0131] The specific network traffic diagram is as follows Figure 3 As shown (Sou is input, Des is output).

[0132] After the network model was built and the diversion method was issued, the network parameters such as packet loss, jitter and delay were normal. The maximum packet loss rate in one hour was 0.03%, the maximum jitter was 5 times / hour, and the maximum access delay was 75 milliseconds. The network traffic also basically achieved load balancing. Figure 4 shown.

[0133] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A network system for traffic load balancing, comprising a user autonomous domain and an operator autonomous domain, characterized in that: The user autonomous domain includes a core layer, a convergence layer and an access layer; The core layer includes two core routers, a network management switch and several network management servers, and is connected to the operator's network to forward traffic on the core network. The aggregation layer includes four aggregation routers, which are responsible for aggregating user traffic of the access layer and transferring it to the core layer; The access layer includes a number of access switches, which are responsible for forwarding access user traffic to the aggregation layer; The core router and the aggregation router communicate via ISIS; The core router communicates with the operation router through the EBGP protocol on the basis of being reachable through the static routing protocol; The aggregation router is configured with a sub-interface to terminate the VLAN transmitted from the access switch; The access switch is configured with transparent transmission and access of VLAN and related ports; The DMtwo traffic classification method is designed to classify and load balance traffic.

2. A network system for traffic load balancing according to claim 1, characterized in that: The network system has the following conditions: The network system has two circuits for load balancing network traffic; The egress circuits of the network system have redundant protection, and when any circuit fails, traffic will be switched to another circuit; The traffic of the network system is finely adjusted according to the IP address.

3. A network system for traffic load balancing according to claim 1, characterized in that: The network management server collects the source address and destination address of the data packets of the AS in real time, and classifies them by the DMtwo method.

4. A network system for traffic load balancing according to claim 1, characterized in that: The EBGP routing strategy on the core router includes the following steps: Divide the source address set into two sets Sou1 and Sou2 according to the DMtwo method; Set the MED value of Sou1 and the MED value of Sou2 on core router 1, and Sou1 <Sou2; Set the MED value of Sou1 and the MED value of Sou2 on core router 2, and Sou1>Sou2; According to the DMtwo algorithm, the destination address set is divided into two sets Des1 and Des2; Set the local priority of Des1 and the local priority of Des2 on core router 1, and Des1 <Des2; The local priority of Des1 and the local priority of Des2 are set on core router 2, and Des1>Des2.

5. A network system for traffic load balancing according to claim 1, characterized in that: The DMtwo traffic classification method specifically includes: Divide the packets entering the AS into two sets of approximately the same size based on the source address; Divide the packets leaving the AS into two sets of approximately the same size based on the destination address; Create a routing policy Policy1 for the source address set based on the MED attribute, and create a routing policy Policy2 for the source address set based on the MED attribute; The Policy1 and the Policy2 are sent to the core router for traffic forwarding and receiving.

6. A network system for traffic load balancing according to claim 1, characterized in that: The algorithm design and implementation of the DMtwo traffic classification method specifically include: Assume two arrays Addr1 and Addr2, the array lengths are L1 and L2 respectively, the two arrays store the network addresses detected by the network management program, and SAddr1>=SAddr2, where SAddr1 represents the sum of the element lengths of the array Addr1, and SAddr2 represents the sum of the element lengths of the data Addr2; When SAddr1 == SAddr2, no exchange is required and the minimum difference is 0; When SAddr1>SAddr2, let the difference SUB=SAddr1-SAddr2, and continuously reduce the value of SUB by exchanging the elements in the two arrays; Each time, an element Addr1[i] and Addt2[j] are selected from the array Addr1 and the array Addr2 respectively for exchange. The new difference SUB after the exchange cha for: SUB cha =(SAddr1-Addr1[i]+Addr2[j])-(SAddr2-Addr1[i]+Addr2[j])=SUB-2×(Addr1[i]-Addr2[j]); Among them, SUB cha Indicates the new difference, Addr1[i] indicates the length of the element with subscript position i in array Addr1, and Addr2[j] indicates the length of the element with subscript position j in array Addr2; The above exchange process is repeated until no Addr1[i] and Addr2[j] that meet the conditions for exchange are found; the conditions are SUB>=2×(Addr1[i]-Addr2[j]) and Addr1[i]>Addr2[j].

7. A method for traffic load balancing, applied to the network system as claimed in any one of claims 1 to 6, characterized in that: The steps include: S1. Create a network model to determine the specific device configuration and connection relationship of the core layer, aggregation layer and access layer; S2, deploy network protocols, configure the EBGP protocol between the core router and the aggregation router and the ISIS protocol between the aggregation router and the core router; S3, real-time monitoring of traffic, using the network management server to collect the source address and destination address of the data packet; S4. Generate two source address sets and two destination address sets using the DMtwo algorithm; S5, generating routing policies Policy1 and Policy2 according to the source address set, and generating corresponding routing policies according to the destination address set; S6, sending the Policy1 and the Policy2 to the core router to forward and receive traffic; S7, set the MED value and local priority value on core router 1 and core router 2 to achieve load balancing of traffic on the two egress circuits; S8. Refresh the routing policy regularly, regenerate and issue a new routing policy every 30 minutes or when congestion is detected on the egress circuit; S9. Adjust traffic classification and routing strategies according to network operation status; S10. During the operation of the network system, continuously monitor the traffic and network performance, and perform manual adjustments and optimizations when necessary.

8. A method for traffic load balancing according to claim 7, characterized in that: The S10 specifically includes: S101. Use network performance monitoring tools to detect network parameters in real time, including network packet loss rate, jitter and delay parameters; S102, when abnormal parameters are detected, automatically generating an alarm and triggering a flow optimization program; S103. Analyze historical and real-time data through machine learning algorithms to predict traffic peaks and make adjustments in advance; S104. Generate network performance reports regularly, evaluate the effectiveness of traffic load balancing strategies, and make adjustments and improvements based on the evaluation results.

Images