（1.國防科技大學計算機學院，湖南長沙 410073；2. 長沙大學計算機系，湖南長沙 410003）
　　摘要：高品質的視頻業務用戶體驗同時要求網絡的低延遲和高速率，并且需要網絡在較長一段時間內保持穩定的性能。通過對某省電信網絡樣本容量為45000的視頻連接分析，當前的傳輸網絡不能完全滿足高質量視頻傳輸的要求。針對上述問題，本文設計了一個基于從全局角度獲取的客戶端和網絡狀態進行用戶體驗評估的機制，用以確定視頻流量的分發策略（傳輸路徑、數據提供服務器和啟動傳輸速率），以優化在不可靠的傳輸網上進行的視頻傳輸。分析結果顯示，該分發策略可有效降低視頻業務的緩沖播放比和啟動延遲，在高負載和負載波動劇烈的情況下，仍能有效發揮作用。
　　關鍵詞：評估反饋；視頻；流量分發
A Video Traffic Distribution Strategy Based on Evaluation and Feedback
LUO Yu 1，ZHUO Lin2, QIAO Qiao 1，LUO Wei1
 (1.College of Computer, National University of Defense Technology, Changsha Hunan 410073,China;
2. Department of Computer, Changsha University, Changsha Hunan 410003,China)
　　　Abstract: Both low-latency and high-speed is required for the high-quality video services user experience, and at the same time, the stable network performance is required within a longer period of time. Based on the analysis of a sample from one province telecommunications network of which contains 45,000 video links, we argue that the current transmission network can not fully meet the requirements of high-quality video transmission. In response to these problems, this paper completed the following work：a user experience evaluation mechanism based on the global perspective of client and network status is designed to determine the distribution of video traffic strategy(transmission path, data server and transfer rate). So the video transmission through an unreliable transport network is optimized. The analysis showed that the distribution could reduce the ratio of buffer and play and start-up delay, even in the case of high load and wild load fluctuating, it can play an effective role.
　　　Key Words: Evaluation and Feedback; Video; traffic distribution

參考文獻
[1] Global Internet Phenomena 1H 2012 Report. Sandvine Intelligent Broadband Networks.
[2] http://blogs.cisco.com/sp/comments/cisco\_visual\_networking\_index\_forecast\_annual\_update/. Cisco forecast.
[3] F. Dobrian, V. Sekar, A. Awan, I. Stoica, D. A. Joseph, A. Ganjam, J. Zhan, and H. Zhang. Understanding the impact of video quality on user engagement. In Proc. SIGCOMM, 2011.
[4] MacQueen, J. B. Some Methods for classification and Analysis of Multivar-iate Observations. Proceedings of 5th Berkeley Symposium on Mathematic-al Statistics and Probability. University of California Press. Page:281-297.
[5] Amorim, R. C.; Mirkin, B (2012). Minkowski metric, feature weighting and anomalous cluster initializing in K-Means clustering. Pattern Recognition 45 (3): 1061-1075.
　　　
作者簡介：
羅彧（1983- ），男，江西省吉安市人，碩士研究生，主要研究方向為無線傳感器網絡；
卓琳（1975- ），女，副教授，主要研究方向為網絡應用開發；
喬橋（1982- ），男，貴州省貴陽人，碩士研究生，主要研究方向為計算機應用技術；
羅偉（1983- ），男，海南省�？谑腥�，碩士研究生，主要研究方向為計算機網絡技術。