流量模型基於分數綜合自還原移動平均過程

TRAFFIC MODELING BASED ON FRACTIONAL ARIMA PROCESS

尤家祥(Chia-Hsiang Yu)

指導教授 ： 洪英超

國立中央大學/理學院/統計研究所/碩士(2006年)

摘要

有別於過去電路交換網路的系統，現今的電信網路大都是改成了封包交換網路之網路系統。過去十多年來已有許多的文獻發現了封包交換網路系統中的封包傳輸的量並不符合一個Poisson過程取而代之的是，現在的封包資料顯現出兩個重要的特徵： self-similar與long-range-dependence。這也代表著我們有必要採用一個新的模型以保留封包資料的特徵。在這份研究中，我們主要是採用一個稱之為fractional ARIMA(FARIMA)的模型，此模型主要是利用傳統時間序列模型的概念並加以衍生。採用FARIMA模型最主要的好處就是它能在同一時間捕捉到資料中的long-range-dependence與short-range-dependence的特性。

在論文的最後，將以真實的網路資料來做分析。
本篇研究中的資料主要是由美國伊利諾州Naperville的Lucent Technologies的區域網路所收集得到。我們將針對乙太網路中各種不同的網路服務型態的封包資料(如：http與ftp的封包資料)配適為一FARIMA模型，並利用一些統計方法來檢查資料的long-range-dependence特性以及FARIMA模型中的參數估計。

關鍵字

長範圍相關 ； 分數綜合自還原移動平均 ； 封包

並列摘要

Departing from the circuit-switched scheme, the transition of internet traffic has been
designed to be packet switched. Over the last decade, fairly rich studies have shown that the transition of internet packets is not Poisson like. Instead, these packets reveal to hav two important features: self-similarity and long-range-dependence. This suggests that a new modeling technique for the internet traffice is necessary. By utilizing the ideas from traditional time series models, in this study we
consider a method called fractional ARIMA(FARIMA). The main benefit we gain by using FARIMA models is that it captures well both the long-range-dependence and the short-range-dependence of the data under consideration.

For application purposed, we also construct the FARIMA models for various types of Ethernet traffic (e.g. http and ftp data) gathered on a LAN at Lucent Technologies in NAperville, IL.
Statistical techniques are used to detect the long-range-dependence property and alsoto estimate the parameters in the proposed FARIMA model.

並列關鍵字

long-range-dependence ； Fractional ARIMA ； packet

參考文獻

of long memory time series models. Journal of Times Series Vol. 4, No. 4, 221-237.

[3] Granger C. W. J. and Joyeux R. (1980). An introduction to long-range time series models and fractional differencing. Journal of Time Series 1, 15-30.

[4] Grossglauser M. and Jean-Chrysostome Bolot J. C. (1999). On the relevance of long-range dependence in network traffic. IEEE/ACM Transactions on Networking Vol. 7 (5), 629-640.

[7] Lo A. W.(1991). Long-term memory in stock market prices. Econometrica 59, 1279-1313.

[8] Mandelbrot B. B.(1975). Limit theorems on the self-normalized range for weakly and strongly dependent processes. Zeitschrift fur Wahrscheinlichkeitshteorie und verwandte Gebiete 31, 271-285.