Research On Video Application Performance Optimization Technology Based On Network Long-Term Feature

WebRTC(Web Real-Time Communications)is a technology that supports web browsers to conduct real-time voice conversations or video conversations.The GCC(Google Congestion Control)algorithm is a congestion control algorithm customized by WebRTC,which is responsible for adjusting the sending bit rate in time according to the changes in the network congestion state during the video playback process to ensure smooth and clear video playback.The native design of WebRTC is universal,and it can provide qualified video services for most network scenarios by using short-term prediction of network status.Through the performance test of WebRTC in DCI(Data Center Internet),it is found that the GCC algorithm has a misjudgment of the network congestion trend..Serious misjudgment will lead to a decrease in the transmission bit rate and a short time of video resolution.frequent drops within the video,which adversely affects the clarity of the video playback.This is because the GCC algorithm only relies on the short-term prediction of the network delay to judge the network congestion state,and does not save any historical network state,that is,it provides a stateless service.The short-term prediction uses a sliding time window to predict the network congestion trend,which is easily affected by network delay jitter.Combining the longterm network characteristics can reduce the error of the delay jitter on the prediction results.The optimization idea of this paper is to use the long-term characteristics of the network to alleviate the misjudgment of the network congestion state by the GCC algorithm in the public cloud network DCI.Long-term network characteristics refer to the study of the probability distribution characteristics of network traffic or network performance indicators such as network delay and packet loss,which can maintain stable network characteristics for a long time.The distribution characteristics combine the long-term characteristics of the network with the original short-term prediction of the GCC algorithm to maintain a certain historical network state.By reducing the misjudgment of the network congestion state by the GCC algorithm,it avoids unnecessary sharpness degradation and improves the quality of WebRTC video playback.This paper proposes an optimization scheme of GCC algorithm based on outlier filtering and dynamic descent rate(OFDD-GCC).The research content can be divided into two parts:(1)Aiming at the high misjudgment rate of the network congestion trend in the short-term prediction of the GCC algorithm,a long-term network characteristics analysis method(ALCN)is proposed.Through the process of packet group processing,clustering,fitting and other processes on the network data of paths between different public cloud nodes,the long-term characteristics of the network are obtained through analysis.(2)In order to reduce the misjudgment rate of the GCC algorithm for network congestion,the Outlier Filter Module(OFM)is added by using the network packet group delay characteristics in the long-term characteristics of the network.In order to alleviate the negative impact of misjudgment on the transmit bit rate,a Transmit Bitrate Drop Controller(TBDC)is added to calculate the continuous occurrence of congestion based on the distribution characteristics of the time interval and duration of network congestion events.The probability of implementing a dynamic rate adjustment strategy.At the expense of the time and computational cost to analyze the longterm characteristics of the network offline,this solution does not increase the time complexity of the GCC algorithm in video communication,and optimizes the smoothness and clarity of WebRTC video playback in public cloud networks.