Fault Management In Network Function Virtualization Environment

In order to meet the growing needs of network services,more and more network middleware is deployed to the network which makes the network ossification problems serious.Network Functions Virtualization(NFV)transforms the existing network architecture from the operator's perspective by decapsulating hardware and software resources.NFV integrates various network hardware devices into industry-standard x86 servers and uses software-implemented network functions to replace the functions of proprietary devices,which can effectively enhance the development efficiency of network functions and make the deployment and management of network functions more flexible and convenient.Network Functions Virtualization has attracted wide attention in academia and industry since its inception.Although NFV can improve the efficiency of network service,it introduces reliability challenges to the network at the same time.Virtualization processes also consume additional system resources,and virtual nodes are more complex than traditional physical nodes.These factors lead to a higher probability of occurrence of faults in the Network Functions Virtualization environment.Whether fault detection and fault location can be performed quickly and accurately after the fault occurs is critical to maintain the system's long-term stable and reliable service.Fault management is an essential part of network management,and its goal is to eliminate the faults and restore the normal operation of the service as soon as possible to achieve the requirements of the Service Level Agreement signed with users.This thesis studied two core technologies of fault management for NFV: fault detection and fault location.Fault detection is responsible for capturing faults in the system,and fault location is responsible for locating the cause and location of the faults.To meet the environmental requirements of Network Functions Virtualization and overcome the shortcomings of traditional numerical detection technology,this thesis builds a fault detection model based on unsupervised machine learning.This thesis proposed an unsupervised fault detection mechanism based on model fusion,which can effectively cover the detection blind spot of traditional detection algorithm and improve the detection performance.In addition,this thesis also deals with the false alarms caused by burst characteristics,and proposes a filtering mechanism based on time window correction mechanism.In the aspect of fault location for NFV,this thesis studied the methods for locating fault and it's causes in virtual network environment,and proposes a hierarchical fault location mechanism based on correlation analysis.This mechanism locates the fault causes based on the correlation analysis of time and variation of attributes,and locates the fault location based on the correlation analysis of time,Service Function Chaining and resource sharing.By extending the dependency graph to increase the relevance introduced by Service Function Chain,the accuracy of positioning results can be improved effectively.The positioning mechanism integrates the positioning results of each level through the implementation principle of "top-down,layer-by-layer positioning",which breaks through the bottleneck of information acquisition.This thesis builds the simulation platform based on three-tier cloud service,and several different types of faults will be injected into the environment to conduct the experiment.The simulation results show that the proposed solution can effectively detect and locate faults injected into the platform.