Research On Cause Analysis And Solution Design Of Cascading Failures In NFV Environment

With the development of the Internet,the demand for upper-class service is becoming more and more diversified.The traditional underlying network function is limited by the exclusive hardware and can not be quickly updated.By applying IT virtualization technology into the traditional network,Network Function Virtualization NFV separate the software function from the traditional network hardware,making the network function programmable and dynamically scalable,greatly enhancing the network updating speed.However,there are large differences bewteen the reliability requirements of traditional telecommunication networks and IT networks.Cascading failures occurred frequently in IT network systems,to satisfy the high reliability requirements of the telecommunication network,the reliability problems of the network in the NFV environment should be paid attention to.In this thesis,the mode of failure propagation in the NFV environment is analyzed from the perspective of SFC(Service Function Chain).Experiments show that there are two different failure propagation modes in the NFV environment:(1)When the service function modules that shared by different SFCs get stuck,the failure will spread across a large number of different SFCs.(2)When the performance of different service function modules in the same SFC do not match,the failure will spread along the bottleneck module of the service function chain to the upstream module.In order to prevent the large-scale failure propagation caused by the failure of shared resources,the thesis try to strengthen the protection of key resources such as shared resources to improve the reliability of the system.Firstly,a three-stated multilayer dependent model is established for the multi-layer virtualization features of NFV,and the relationship(dependency,backup,load balancing)among the modules of different levels is abstracted.Then,in order to optimize the system reliability,a backup resource allocation algorithm is proposed.Simulation results show that the algorithm can obtain the optimal or approximate optimal solution.In summary,it provides the idea for the failure isolation from the backup resource allocation angle.For the scale-up storm caused by the mismatch of different service functions' performance in the same SFC,a linkage scale up algorithm is proposed.The main idea of the algorithm is to predict the impact of the scale-up of the bottleneck module on the downstream module,then scale up the affected modules and bottleneck module at the same time,so as to avoid the multiple scale-up.To be concrete,the algorithm uses neural network to establish the relationship between the scale up amount and the input and output rate of the data stream.Then the influence of the scale-up of the upstream node on the downstream nodes can be quantified,thus the downstream nodes can scaleup earlier and the failures are prevented.The simulation results show that the linkage scale-up algorithm can avoid the secondary mismatch problem that the single point independent scale-up may cause,which saves the total time of system scale-up and reduces the average service response delay.Finally,in order to apply the algorithms,on the basis of current NFV architecture,the failure isolation architecture is proposed.The architecture is designed to satisfy the algorithm input and output requirements,adding the monitor module and the algorithm deployment module,giving the running examples,enhancing the practical value of the algorithms as a reference.