Research And Implementation Of Intent Translation And Intelligent Routing Algorithm In Intent Driven Network

With the development of computer networks,various network types and network devices emerge in an endless stream.How to enable the network understand people’s expectations or intentions,generate strategy and configure has gradually attracted the attention of researchers.In this context,academia and industry have respectively proposed the concept of Intent Driven Network.Intent network can receive people’s intentions(expected network state from people),and automate network policy inference and configuration.In this paper,the main researches focus on how to make IDN understand people’s subjective thoughts and translate intent into policy and configuration.This paper proposes an intent translation algorithm based on deep learning.According to the existing research,IDN receives people’s high-level expectations of the network,and its representation can be sentences expressed in natural language.With the development of deep learning in the field of text processing,key information in sentences can be extracted by detection model.In this context,this paper builds an intent translation algorithm that can perform intent category detection and key information extraction from intent-expressing sentences.The algorithm can identify the intent category of the sentence,and label the words in the sentence.These words and labels plus the category type will be used as the effective information for the intent-driven network to generate configuration.Also,this paper constructs a sequence-to-sequence model that can use the extracted information for intermediate code conversion.It can structure the sequence of unordered extracted information and generate intermediate code sequences to facilitate further generation of the bottom layer configuration.This paper proposes an intelligent routing algorithm based on reinforcement learning.When intent involves performance expectations for traffic flows in the network,the IDN network should be able to adjust routing policies based on network status and its understanding of intent.Reinforcement learning,as an algorithm that continuously interacts with the environment and learns the optimal strategy,can deal with the ever-changing network environment.In this paper,in order to deal with the situation that traffic flows with multiple intent types compete for resources in the network,a multi-agent reinforcement learning algorithm with a double-layer structure is proposed,which can periodically adjust the bandwidth resources occupied by traffic flows with various intent types.At the same time,it can find the optimal routing path for different types of business flows under the allocated resources,and the algorithm will continuously modify the routing strategy to meet the QoS metrics expected to be achieved.