近幾年來通訊技術的發展有顯著進步,通訊系統工程師如何利用新的管理策略,支援高速網路的服務品質要求是十分重要的工作。在有限的網路資源中,透過限制服務資源的方式使網路能提供更好的服務品質是一個很好的處理方式。因此,本研究針對這個方式提出一種新式的訊務管理策略,稱為動態化訊務管理策略。依照訊務突發狀況的不同分別使用二種不同的服務模式,突發性(尖鋒資料量�平均資料量)重--優先權管理模式,一般突發性--智慧型服務模式。 優先權管理模式的架構: (1)依訊務資料種類建立優先權運算矩陣,並透過馬可夫遞迴分析法預測訊務變動情況。 (2)將三種訊務(voice、fax、data)依預測比例混合模擬訊務參數。 (3)以基因演算法找尋最佳路由與路由參數最佳分數。 智慧型服務管理模式的架構: (1)使用模糊時間序列預測訊務變動量。 (2)依預測比例混合模擬訊務參數。 (3)比較多媒體訊務在不同服務技術傳送的參數變化。 除此之外,為提昇網路性能,由研究結果可見動態化管理策略可以支援骨幹網路設計、增強DNHR/RTNR的路由策略並且支援企業VPN建構使用。 關鍵字:多媒體訊務、基因演算法、模糊時間序列、服務品質
The development for communication techniques has made remarkable progress in recent years. It becomes important for communication system engineer who uses new management strategy to beef up the source of service quality of high-speed network. Through using limited source of service to let network have a more quality of service is a good way. Therefore, this study pointing on this subject to provide a new traffic management strategy, naming Dynamic Traffic Management Strategy. According to different burst situations two kinds of service module. We made: heavy-burst (peak data rate/average data rate) using Priority Management Module. common-burst using Intelligence Service Module. The structure of priority management module includes: (1) Classify the incoming data and setting up Markov's recursive process to predict the changes in the flowing data. (2) Using three traffic service messages (voice, fax and data) of previous data to predict next service parameters. (3) Using genetic algorithm to find the best-fit parameters and best-fit route. And the structure of Intelligence Service Management module includes: (1) Using Fuzzy time series to determine the amount of traffic flows. (2) Using predictive ratio to simulate traffic parameters. (3) To compares the different service techniques to know the change in multimedia traffic. Besides, to upgrade the network service, the Dynamic Traffic Management Strategy was designed to support the network improvement, to enhance DNHR/RTNR routing strategy﹐ and to support the use of business VPN. Keywords:Multimedia、Genetic Algorithm、Fuzzy time series、QoS