通訊網路使用網路聚合（Network Convergence）的策略，以此提供一個開放的服務平台，ALL-IP網路，因ALL-IP網路是以封包交換網路承載各類服務[1]，所以不容易保證服務品質。一個經常被採用的方式是使用資源保留（Resource Reservation）方式以保證服務品質，但服務可能會因失約或未依約沒有充分運用所保留之資源而導致資源閒置、降低網路資源使用率。為了要提高資源使用率，營運者必須要有解決的方法，以提高整體資源使用的效能。 本研究提出一個有效使用資源的概念—超額允入（Over-Admission），超額允入的意義是讓允入控制機制允入比所擁有的資源量還要多的訊務量，以增加資源使用的效能。但是超額允入訊務之後，Admission Controller可能會允入過多訊務造成溢流（Overflow）。這會使系統效能下降，所以網路管理者必須要找出合適的允入量以提高系統整體效能。 本研究建立分析模型係以供計算最佳允入量，並以模擬進行模型驗證與效能評估，經由實驗結果可以驗證本分析模型相當準確而超額允入的確可以大幅提高資源使用率。 / The telecommunication industry is moving toward a converged network, which uses a single global IP based packet-switched network, which uses a single global IP based packet-switched network to carry all types of network services, the quality of service cannot be fully guaranteed [1]. One of the popular methods to guarantee the quality of service is to use resource reservation. This kind of methods, however, might cause resource underutilization because the resource requesters may not show up or, even if does, it does not fully utilize the reserved resources. In order to enhance the efficiency of resource utilization, this study proposed an efficient resource utilization concept, that is, over-admission. Over-admission, in which the admission controller may admit more traffic volume than what it can allowed. Even though over-admission is effective, it has its own problems: the admission controller may admit too excessive traffic to impair the system performance. In order to maintain system performance, it is important for the network administrators to find out the appropriate admitted amount that does not cause excessive overflow. In this thesis, an analysis model is constructed to estimate the optimal admitted traffic amount under various traffic conditions. Then, simulation is conducted to verify this model and to evaluate the performance of our proposed method. The results reveal that this analysis model can improve the efficiency of resource utilization.

本論文對IEEE 802.16 協調分散式之網狀網路提出一允入控制之演算法。在此類網路中，控制子訊框交換各站台之排程訊息，並預留資料子訊框之時槽作為實際資料傳輸之用。我們利用令牌桶機制來控制網路訊流之流量特徵，如此可簡單的估計各訊流所需之頻寬。我們使用了所提出的頻寬估計方法，並一起考慮各訊流之跳接數與延遲時間之需求，提出的允入控制演算法能夠保證即時性串流之延遲時間需求，且可避免低等級訊流發生飢餓情形。模擬結果顯示，所提出的允入控制方法可以有效的把超過延遲時間需求之即時性訊流封包數目降低，並且低等級訊流在網路負載大時仍然可以存取頻道。 / We propose a routing metric (SWEB: Shortest-Widest Efficient Bandwidth) and an admission control (TAC: Token bucket-based Admission Control) algorithm under IEEE 802.16 coordinated, distributed mesh networks. In such network architectures, all scheduling messages are exchanged in the control subframes to reserve the timeslots in data subframes for the actual data transmissions. The token bucket mechanism is utilized to control the traffic pattern for easily estimating the bandwidth of a connection. We apply the bandwidth estimation and take the hop count and delay requirements into consideration. TAC is designed to guarantee the delay requirements of the real-time traffic flows, and avoid the starvation of the low priority ones. Simulation results show that TAC algorithm can effectively reduce the number of real-time packets that exceed the delay requirements and low priority flows still can access the channel when the network is heavily-loaded.

為了進一步支援無線網路上的服務品質(QoS)，IEEE 802.11e通訊協定目前正在制訂中。其提供兩種媒介存取方式，一為基礎的進階分散存取(EDCA)，另一種是建構在進階分散存取下的混合控制存取(HCCA)。802.11e中，藉由不同的訊框間隔(IFS)與競爭視窗(CW)相關參數的設定以區隔不同的存取等級(AC)。為了在802.11e網路下進一步的發展有效的服務品質管理機制，我們提出了一個數學模型以分析在進階分散存取網路下所使用的頻寬與媒體存取層所延遲的時間。在這個模型下，每個站台可以支援多個不同等級的資料流(聲音/影音與資料)。透過這個數學模型，允入控制與資源管理可以很容易實現，也可以支援不同應用程式的不同需求。 / For supporting Quality of Service (QoS) for wireless networks, the IEEE 802.11 Task Group E currently defines enhancements to the IEEE 802.11 MAC, called 802.11e. The IEEE 802.11e provides two mechanisms for the support of applications with QoS requirements, namely, Enhanced Distributed Channel Access (EDCA) and HCF Controlled Channel Access (HCCA). EDCA mechanism defines four access categories (ACs) that provide support for the delivery of traffic with user priorities. Different AC uses various Inter-Frame Space (IFS) and Contention Window (CW) parameters. In order to further develop efficient QoS management for the IEEE 802.11e networks, we propose an analytical model to evaluate throughput and MAC delay of the basic access method of the IEEE 802.11e - EDCA. Our Markov chains consider the situation of different multimedia (voice/video and data) traffic flows in a mobile station. This is improved from previous work, and closer to real usage scenarios. The correctness of our analysis has been validated via simulation results. Throughout our model, call admission control (CAC) and resource management can be easily applied, and thus QoS for hybrid requirements is supported.