Research On Cloud-Native Virtual Network Function Orchestration

The demand for network service functions in the data center is increasing sharply.The NFV technology that uses software and hardware decoupling to build software-based network functions on a common hardware platform has developed rapidly.Under the traditional VM based architecture,NFV platforms have resource utilization,flexibility and performance shortage,and cannot effectively meet the flexible and efficient requirements of NFV.At the same time,cloud-native technology based on containerized technology has developed rapidly.This thesis has studied cloud-native based NFV from the perspective of containerized VNF orchestration requirements.Platform design,protocol mechanism,and orchestration algorithm for network function orchestration.First,this thesis studies the overall design of NFV platform with cloud-native technology as the core and containerized VNF building model.The NFVI based on Kubernetes was proposed,as well as the implementation mechanism for each component in the NFV MANO.And the effectiveness of the platform was proved by implementing a prototype system.In order to overcome the problem that the traditional VNFD mechanism was not applicable in the cloud-native environment,this thesis investigated the protocol design of the cloud-native VNFD definition mechanism,completed the VNFD construction in a templated manner,and standardized the containerized SFC definition mechanism to provide a basis for the automatic arrangement of containerized VNF.The traditional SFC scheduling algorithm was usually based on virtual machine platform,which has high network latency and insufficient resource utilization in containerized environment.In this thesis,we built a scheduling model for cloud-native NFV platform,analyzed the local optimal properties under flat topology,and proposed the Latency Optimized Placement algorithm(LOP).LOP used multi-stage decision to reduce the number of cross-hosts of SFC.Through simulation experiments,the results showed that the LOP algorithm can effectively reduce network latency.