Design And Implementation Of Cross-Protocol Load Balancing Mechanism For Smart Integration Identifier Networking

With the continuous emergence of new network protocols,network systems are becoming more and more diversified,and diverse network systems promote the development of network intelligence integration.However,network integration is bound to pose a huge challenge to the traditional Internet system.On the one hand,the diversity of network protocols and terminal access methods makes it difficult to complete data transmission through a single protocol,requiring the use of cross-protocol mechanisms for data transmission.On the other hand,the coexistence of heterogeneous networks will bring huge traffic pressure and form a complex network environment,which greatly increases the difficulty of traffic load balancing,and the converged heterogeneous network faces the performance crisis of unbalanced link load.However,the traditional load balancing method cannot adapt to the heterogeneous network environment because it does not combine the cross-protocol mechanism.Therefore,combined with the reinforcement learning model,this paper designs and implements a cross-protocol load balancing mechanism based on Smart Integration Identifier Networking,which realizes the load balancing of protocol matching when the network traffic load is low,and reduces the cross-protocol processing to a certain extent.When the protocol matching node load is high,cross-protocol load balancing is performed to make full use of the forwarding resources of each forwarding node.The main works and innovations of this paper are as follows:(1)This paper improves the data packet structure design and working process of INT in-band telemetry,and reduces the overhead of node state acquisition as much as possible.Firstly,the INT in-band telemetry technology is used to collect the node status of the forwarding equipment in the whole network.Secondly,the INT header format is optimized,and the original INT Header header is optimized into an 8-bit field,and the redundant fields in the INT Metadata header are removed,and the whole collection process is completed.No control plane involvement is required.(2)This paper uses the reinforcement learning model to calculate the value of nodes to generate equilibrium decisions,and optimize the reward objective function in the process.Firstly,use the reward function and the value update function to calculate the value of each node;secondly,introduce the comparison of the protocol in the calculation of the reward value of the reinforcement learning algorithm,so that the reward value forwarded to the node matching the protocol is higher,When the load is small,priority is given to balancing to such nodes to reduce crossprotocol overhead.(3)In this paper,the flow slicing and the packet granularity scheme are combined to achieve load balancing,so that the data packets in a flow slice can be sent to each candidate path in a round-robin manner according to the value weight.And users can choose tunneling or protocol conversion for cross-protocol processing as needed.Since only one policy is required for a flow slice,the policy and flow table update overhead is reduced as much as possible,and based on flow slices,the packet is finegrained to ensure the effectiveness of cross-protocol load balancing.Finally,this paper conducts experimental analysis on the function and performance of the cross-protocol load balancing mechanism.Experimental results show that the mechanism proposed in this paper can achieve cross-protocol load balancing in heterogeneous networks.Moreover,compared with the packet granularity polling mechanism and the flow slice equalization mechanism,this mechanism has lower packet loss rate in a typical heterogeneous network environment,the average throughput is increased by 23.41% and 38.5% respectively,and the throughput change is more stable.According to the distribution of forwarding resources of nodes,the traffic can be accurately balanced to each node according to the proportion of resources.In addition,for a faulty environment,this mechanism can timely detect possible faults or interrupted nodes,and intelligently adjust the balancing strategy to reduce packet loss and unnecessary retransmission,and ensure the normal transmission of traffic in a faulty environment.