Automatic Adaptation Scheme Of Yang Models Based On Configuration Tree Similarity

With the rapid development of Internet technology,the scale of the network has expanded rapidly,and the pressure for the configuration and management of network equipment has also increased.The traditional SNMP(Simple Network Management Protocol)can't meet the requirements of management for large-scale network configuration.Therefore,the IETF set up a working group and proposed a more powerful and flexible network configuration protocol,NETCONF.NETCONF messages are usually formatted and defined using YANG as a modeling language in the content layer.However,there is no uniform standard when various vendors use the network configuration model defined by YANG.When the YANG models of different vendors describe the same device function,the defined structure and content often differ from each other.Therefore,when the local device uses the YANG model of the third party,it needs to be translated by the translation middleware to complete the mapping of the configuration syntax between the third party model and the local model.The existing solutions is represented by Juniper's translation script and Cisco's translation plug-in.They all have the disadvantages of high coupling with third-party models,high learning cost,and inflexibility.This thesis proposes a new solution for fulfilling the requirement of uniformly configuring and managing devices from various vendors in complex network environment.The solution extracts the information of the YANG models of both parties and calculates the similarity.The solution calculation results are used to automatically generate the mapping between the two models.The solution has the characteristics of high flexibility and low labor costs.The main work of the thesis are as follows:(1)According to the requirements similarity calculation algorithm and the characteristics of YANG Model,this thesis proposed the concept of configuration tree.The construction process of the configuration tree includes the merger of modules,the expansion of reference statements,the extraction of special structures,and the flattening of the configuration tree structure.Under the premise of ensuring high effectiveness for algorithm processing,the useful information is retained in the YANG model.(2)Combining the features of the configuration tree,this thesis designs a method for calculating the similarity of configuration tree(CTS).Two techniques are provided to calculate the semantic similarity of node attributes when calculating the similarity of configuration tree node attributes.These technologies are WordNet and Word2 Vec.After the similarity between words is obtained,the similarity between short texts is computed by using WMD algorithm.So the attribute similarity of each node can be gotten.Then the algorithm uses similarity of attributes to calculate the similar of each nodes.Based on this,the path information is combined to calculate the similarity between the paths of the configuration tree.Finally,the algorithm computes the similarity of the two configuration trees.(3)Aiming at the scenario of adaptation for YANG models,this thesis designs a mapping rule generation algorithm based on optimal matching of nodes(MRG).Based on the result of similarity calculation between configuration tree,the algorithm use depth-first traversal to traverse the configuration tree to be matched and generates a candidate node set through global optimization and local optimization.Finally the algorithm traces its parent node up to the root node in the candidate node set.Finally,the node with the highest path similarity is selected from the candidate node set as the optimal matching and a mapping rule is generated.Based on this,the vocabulary of mutex is constructed and the alternative nodes are screened according to the vocabulary,which improves the performance of the algorithm.(4)This thesis implements a prototype system for YANG model automatic adaptation based on similarity of configuration tree.The main functions such as configuration tree construction,similarity calculation,and mapping rule generation have been completed.The results of the experiments about function and performance show that the YANG model automatic adaptation solution proposed in this thesis can accurately find out some mapping relationships from YANG models,and the running time is controlled at the minute level,which is far less than the time required by the traditional solution.