IP網路電信服務時間序列效能評估模式之研究

Performance Evaluation Modeling for Telephony Service of Time Series on IP Network

高菁穗(Jing-Suey Gau)

指導教授 ： 莊謙本 教授 ； 王宏鈴 博士

國立臺灣師範大學/科技學院/工業教育學系/碩士(2002年)

摘要

在科技進步的時代，人們對於通訊品質的要求也日益增加，然而在現今的通訊網路中常常因頻寬速率及節點上緩衝容量的問題，造成傳輸的延遲及訊務數據資料的流失，更由於使用者呈倍速的擴增，使得通訊品質造成極大的影響。因此，電信公司面對日益俱增的種種問題，一方面要考量未來的市場需求，另一方面也要考量設備增設等問題，對於未來的訊務量的大小及多樣性的服務，有著深切的關係。本文面對服務品質訊務需求，採以ARIMA的預測方式，來預估未來訊務特性及流量的大小，並以Kolmogorov-Smirnov Test之檢定方式，對於預測所得之數值進行檢定，預測出的數據方法更能提供網路管理的參考，進而提高現今多媒體網路服務的效能。
ARIMA預測模式架構為：
1. 依訊務資料種類加以分類，並建立初步使用的模型。
2. 將初步使用選定的模型，找出訊務參數。
3. 將三種訊務資料（Voice、Fax、Data）所得之數值，以ARIMA找出最佳預測模型。
對於高品質服務要求的網路而言，尋找良好的網路控制方式為網路管理者極需要的參考依據。經由ARIMA的預測方式，能提供對於未來需求的參考方向有著極佳的價值，也因ARIMA具有可靠度高且有極高的精確度，對於所求的各節點之路徑所預測出來的結果，對於網路上的需確實可做為模擬實際狀況。此點就管理者的觀點，能對於未來的訊務狀況，不但能事先了解到運轉情形，且又能在合乎成本與經濟效益的考量下，對於設備的增設與否有著極佳的參考考量。因此，在提昇服務品質上有著是有其卓實的貢獻。

關鍵字

訊務預測 ； 服務品質 ； 時間序列

並列摘要

In the age of advancing technology, people have more and more demands for the quality of communication. However, due to the problems of bandwidth speed and buffer capacity on nodes, today’s communication can cause delay of transmission and loss of signal data. Furthermore, the rapid growth of consumers’ use causes huge impact on the quality of service. Therefore, when facing a variety of increasing problems, telecommunication companies must consider not only the demand of the future market, but also the expansion of equipments, because they are closely related to the scale of business in the future and diversified services. For the demands of the quality of service, the prediction criteria of ARIMA are applied in this article. They can estimate the signal characteristics and the flow quantity of the future. Also, the Kolmogorov－Smirnov Test is applied to analyze the predicted values. These criteria of predicting values can be a reference for network administrators and enhance the performance of today’s multimedia network service. The structure of the ARIMA prediction model is:
1.To classify the signal data and establish preliminarily used models.
2.To find signal parameters by using the selected models.
3.To find the optimized prediction models of ARIMA by using the values gained from the three signal data (Voice, Fax and Data).
For the network with high quality of service, to find a proper controlling method is strongly required by network administrators. By the prediction criteria of ARIMA, the orientation of future demands can be offered. Thanks to high reliability and precision, the result predicted by the path of every node can simulate the actual situation of the demands on the network. For the signal situation in the future and from the viewpoint of administrators, the functioning condition can be acquired in advance and the decision making of purchasing new equipments can be advised by valuable references of considering the cost and economic benefit. Therefore, it really has prominent contribution of upgrading the quality of service.

並列關鍵字

ARIMA ； Times Series ； forecasting ； Signal prediction ； quality of service

參考文獻

[6] Deering S. & Hinden（1995），The internet Protocol Version 6（IPv6）Specification，RFC 1883 Dec.

[7] Deering S. & Hinden（1995），The internet Protocol Version 6（IPv6）Specification，RFC 1884 Dec.

[9] Deering S. & Hinden（1998），The internet Protocol Version 6（IPv6）Specification，RFC 2460 Dec

[10] Behrouz Forouzan（2000），TCP/IP Protocol Suite，The McGraw-Hill Companies，Inc.

[17]Chen T.M. & Oh T.H.（1999），Reliable Services in MPLS，IEEE Communication Magazine.