Research On Multi-service Isolation Evaluation And Control Technology For Large Scale Packet Transport Network

With the increasing popularity of 5G and the trend of fixed-mobile convergence(FMC),transport network will face new challenges.On the one hand,the explosive growth of traffic will lead to network congestion and increased network latency.On the other hand,IPTV,AR and other real-time applications require the network to provide millisecond-level latency,which gives users a good user experience.Therefore,the transport network needs to bear the dramatic increase in traffic while ensuring the quality of service(QoS).In spite of the fact that there is still no consensus among researchers regarding how the transport network in 5G should be designed,one of the possible solutions is using PTN(Packet Transport Network).PTN ensures effective isolation between different channels through the LSP and the PWE3(Pseudo Wire Emulation Edge to Edge)mechanism.When entering the PTN network,an LSP is established according to the destination IP or VLAN and PW is established according to the interface type of the service or its QoS requirements so that different types of services are transmitted in different PWs of the LSP respectively.The LSP/PWE3 mechanism is designed to avoid the interaction between different services.However,whether the isolation mechanism meets the requirements still needs to be verified.Therefore,a strategy is needed to evaluate the service isolation effect.In this paper,we propose a method to evaluate service isolation based on mutual information,which is evaluated by calculating the average mutual information between performance indicators(throughput,delay,packet loss rate,jitter,etc.).The greater the correlation,the smaller the isolation between services and vice versa.The experimental results show that the service correlation between different LSPs is smaller,that is,the service isolation degree is larger and the service isolation effect is better.The service correlation between different PWs in the same LSP is bigger,that is,the service isolation degree is smaller,and the service isolation effect is worse.The result is consistent with the actual situation,thus verifying the correctness of the model.Considering that the mutual information-based evaluation method is only a simple calculation of performance indicators and ignores revealing the essential reason of the service correlation.Thus,the paper further proposes the service correlation model and reveal the mechanism of the burst spreading process.Numerical results show that the model is in consistence with the reality,that is to say,this model has theoretical guidance for large-scale packet transport network deployment.Finally,this paper proposes a service isolation control strategy based on community structure.Firstly,the community information of different services is obtained according to the community classification algorithm.Then,services with high importance are isolated.Process goes on until services are not affected by the burst.Through the control strategy,small-scale services isolation can guarantee the vast majority of services from the impact of the burst.