Research On Prediction-based Queue Management And Frame Generation Technology In AOS

For the past few years,with the speed development of science and technology,there are more and more network services,which will inevitably lead to an increasing number of serious network situation.For the sake of achieve the needs of the various network business and supply ours with preferable taste,it is necessary to further study network congestion control technology.Self-similarity is the characteristic of network traffic itself,and it is also a common characteristic,which has been verified at present.The self-similarity of network is determined by many factors,including user status,network protocol,service type and so on.The research on congestion control becomes more difficult because of the self-similarity of network traffic,but self-similarity also provides us with a new research idea--introducing the traffic prediction module,so as to better realize the network congestion control.This paper studies congestion control based on self-similarity,establishes a model,and proposes an improved queue management algorithm and an improved frame generation algorithm.Finally,the combination of forecasting and queue management is realized;Combination of queue management and frame generation algorithm;Combination of frame generation and scheduling module.Firstly,this paper chooses wavelet neural network as the prediction model.Wavelet neural network is a new prediction model which includes wavelet analysis theory and neural network thought.By setting appropriate wavelet basis as hidden layer,the model constantly adjusts the parameters of training set to make the output of training set close to the actual expected value,then puts the adjusted parameters into the test set,finally makes overall prediction,and generates self-similar traffic after multiple runs,thus making the wavelet neural network prediction model achieve the desired effect.Secondly,the traffic generated by the model is introduced into the active queue management mechanism,and the improved queue management algorithm is used.By controlling the relationship between queue and packet loss rate,data packets are discarded properly,and finally the stable queue length is obtained.The model can dynamically control the dropping probability according to the predicted traffic,thus realizing congestion control.Finally,the data processed by queue management algorithm is introduced into the improved AOS frame generation algorithm,the comprehensive objective function is determined,and the AFSA is used for optimization to determine the optimal framing time threshold of the adaptive frame generation algorithm which is most suitable for this traffic.Finally,the above algorithms are simulated jointly,and it is found that the average delay and residual amount of the improved queue management and frame generation algorithm are obviously improved compared with those before the improvement.