Research On The Management Technology Of Satellite Network Queue Which Is Adapted To Self-Similar Traffic

The satellite network can be integrated with the ground network to relieve the pressure of the ground network with its wide coverage,large communication capacity.Since satellite network runs a large number of businesses,including services generated on satellites,other satellite services and ground network services,satellite network generates and sends a large amount of multimedia data,and its number changes drastically.It is because the satellite network runs a variety of businesses,the satellite network exhibits characteristics of large capacity,high burst and self-similarity on the network aggregation node.The characteristic of self-similarity will lead satellite network traffic to the occurrence of network congestion,which makes a lot of data lost and may make the network paralyzed.On the network aggregation node,the queue management algorithm can reduce the transmission rate through informing the source end of packet loss in advance,hence,to alleviate the congestion problem of the network.Based on the characteristics of self-similarity of satellite network traffic,this paper first proposed a prediction model of satellite network traffic combination based on Fruit Fly Optimization algorithm-extreme Learning Machine(FOA-ELM).The model uses EMD algorithm to decompose high-frequency and non-linear network traffic into different components.Then,the high-frequency components are predicted by FOA-ELM model and the other low-frequency components are predicted by ELM model,according to the characteristics of different components.Finally,results of different predictions are combined to land the final prediction result.This paper proposes a decomposed FOA-ELM combined prediction model which is superior to other models in the accuracy and speed of prediction and can make a prediction on satellite network traffic effectively and efficiently.The predicted traffic result at the next moment is applied to the Adaptive Random Early Detection(ARED)algorithm based on the prediction result of the FOA-ELM combined prediction model for satellite network traffic.And the prediction result is used to calculate the ARED.The maximum packet loss rate of the ARED algorithm is adjusted according to the average queue dynamics which are obtained from the prediction result.Thus,the algorithm of satellite network queue management which is based on predicted ARED is proposed.This algorithm can stabilize the routing queue length and reduce the packet los0 s rate.The research results of this paper can be used to solve the problem of effective transmission of self-similar traffic in satellite networks,so as to provide technical support for alleviating the congestion of satellite network and optimally allocating resources in the event of traffic emergencies.