基于奇异谱方法的网络行为分析.PDF

第32 卷第6 期 东南大学学报( 自然科学版) Vol132 No16 2002 年11 月 JOURNAL OF SOUTHEAST UNIVERSITY (Natural Science E ition) Nov. 2002 吴 桦 丁 伟 ( , 210096) : 运用数字信号处理中的奇异谱分析方法, 讨论了如何 取网络流量的周期和趋势特征, 并 结合最大熵谱方法对吞吐量趋势做出预测. 将该方法运用于中国教育和科研网CERNET 华东 ( 北) 地区网的主干网, 分析了若干天的网络流量行为特征, 分析结果与使用情况相吻合. 对主干 的流量进行了预测, 并将预测结果与实际情况做了比较. 定义了评价预测准确性的公式, 将预测 的准确性与国际上具有代表性的成果进行多方面的比较, 结果说明该方法准确性较好. : 网络测量; 网络行为; 性能分析; 奇异谱分析 : TP393. 07 : A : 1001- 0505( 2002) 06-0889-06 Study of network behavior based on singular-spectrum analysis Wu Hua Ding Wei ( Department of Computer Science an Engineering, Southeast University, Nanjing 210096, China) Abstract: A metho borrowe from igital signal processing is applie to i entify an retrieve low-fre- quency variability an tren components from history metrics. Combine with maximum entropy metho , pre ictions of network throughput can be ma e. This arithmetic was applie to the backbone throughput of China E ucation an Research Network Eastern China( North) Regional network, an the resultswere con- sistent with the practice. Pre ictions of the backbone throughput were ma e an the results were compare with the practice. A function which can evaluate the veracity of pre iction results is efine , the veracity of this arithmetic is compare with the results of other international research communities. It is shown that this arithmetic has a better veracity. Key words: network measurement; network behavior; performance analysis; singular-spectrum analysis , analysis) . , . , , ,