非同步傳輸模式網路的呼叫允入控制

Abstract

在本論文中, 我們主要是探討接受不同種資訊源的 ATM 網路在呼叫建置 階段所做的流量控制。 即所謂的 ``呼叫允入控制"。在這個題目下, 又 分成兩部份來討論： 確的滿溢機率分析而得的 數"與``由等效頻寬方法 得到的連結個數"兩者數值的比較結果,我們 P種資訊源在緩衝器無限大的 佇列系統的頻寬分配方法來做呼叫允入控制,峏顝翕薿e忍度大但對封包丟 失機率要求嚴格的資訊來源。在第二部份,蘆漕t統模型中的馬可夫資訊源 提出一最佳量化的方法。根據文獻 [13]對於開關式資訊源提出的即時計 算法則稍加修改, 拓展至 且考慮由文獻[13]的計算方法可能造成的量化 損失,我們提出一對資訊源量化的方法, 以期將量化損失減至最小, 達到 較好的鏈結頻寬使用率。 In this thesis, we mainly investigate the traffic control performed at the call setup phase in an ATM network loaded with heterogeneous sources. And this is what so called "Call Control". Under this topic, it is still divided into two categories:first, based on the numerical comparison results between the allowable number of connections from the exact analysis and that from the effective bandwidth method, here, we propose a new bandwidth allocation method for the call admission control to heterogeneous on-off sources in an infinite-buffer queueing system, which the network system accommodating the delay-tolerable but almost loss-free required sources. In the second part, we propose an optimum quantization scheme for the Markov-modulated sources in a bufferless fluid- flow model of an ATM network. Following the real-time computation algorithm for on-off sources proposed in \cite{duann}, with little modification, we extend the application to a more general source model, Markov-modulated sources. Furthermore, considering the quantization loss that might be caused under the computation algorithm in \cite{duann}, we suggest a policy for the sources to minimize the loss and achieve better utilization of link capacity.