Research On Intent Parsing And Verification Method In Intent-Driven Network

The operation and maintenance of traditional networks rely heavily on the experience and skills of personnel.In order to adapt to the increasing complexity and diversity of network deployment,and promote the development of autonomous networks,Intent-driven network(IDN)emerged as the times require.IDN is a new network management paradigm,it can automatically complete a series of operations such as analysis,verification,configuration,optimization,etc.according to user business needs,achieve the expected network status,independently resolve abnormal events,and provide continuous business network guarantee and optimization capabilities.It is of great significance to carry out IDN-related research work to realize intelligent network management and operation and maintenance.This paper studies the method of intent analysis and policy verification in intent-driven network.Intent parsing is responsible for receiving and understanding user intents,and transforming high-level abstract intentions entered by users into low-level network configuration policies.In this paper,intent parsing is divided into two steps:intent recognition and intent translation.Intent recognition adopts Natural Language Understanding(NLU)technology to construct a multi-task joint model based on a pre-trained language model,which can classify the natural statements entered by users and extract finegrained business requirements.Intent translation uses the Resource Description Framework(RDF)in the Semantic Web as the information carrier,and combines the result of intention recognition with the network policy library and actual network state information to generate specific network policies.Policy verification is an important guarantee for the safe and smooth operation of the network.It is responsible for the execution verification of the policy,that is,to find out the routing path that meets the requirements of bandwidth and QoS requirements in the policy,which is essentially a QoS routing problem.This paper will study the verification methods in online mode and offline mode.Online mode processes a single intent in real time,and offline mode batch processes multiple intents that arrive within a cycle.For online mode,this paper designs a multi-constrained path(MCP)algorithm.This algorithm first uses linear coupling function to compress the network,reduce the network size,and then constructs nonlinear coupling function to calculate the optimal feasible path.For offline mode,this paper first designs a k-multiple-constrained Shortest Path(KMCSP)algorithm,which processes multiple intentions separately,finds K feasible paths satisfying QoS constraints for each intention,and then uses formal verification method to transform the problem into a satisfiability module theory(SMT)problem through mathematical modeling.Use the SMT solver to select the appropriate path for each intent,make the traffic distribution in the network more balanced,and solve the conflict and collision problem between multiple intents.