本文提出一個利用模糊臨界值構成之空間優先排序器的訊務模型,並完成了其於真實網路條件下的各種效能分析,其中利用矩陣分析方法(matrix-analytic method)驗證了在高訊務流量以及低訊務流量條件下的封包遺失率。 基於既有的網路基礎架構上,本研究利用模糊臨界值構成之空間優先排序器來評估分析訊務模型的效能。經數值分析的結果顯示,在固定的臨界值設定方式下,一旦網路環境有突然大量的訊務流量湧入時,將會造成高優先權的封包遺失率上升之現象。反觀在不同的訊務流量模型的情況下,模糊臨界值構成之空間優先排序器確實能滿足高、低優先權封包遺失率的實際需求,在此以連續以及離散時間型群組馬可夫來模擬抵達過程。
In this dissertation, we develop both continuous-time and discrete-time queueing model of fuzzy threshold-based space priority buffer management and study its performance under realistic conditions. It applies a matrix-analytic approach to analyze the relevant performance measure, including the packet loss probability of high-priority traffic and the packet loss probability of low-priority traffic. Based on the proposed framework, we explore the properties of the fuzzy threshold-based space priority buffer management scheme. Numerical results reveal that the fixed threshold scheme, through its abrupt nature, causes a relatively higher low-priority packet drop. Intuitively, the fuzzy threshold adapts well to different input traffic conditions and packet loss rate requirements of high-priority packet, yielding a lower packet loss probability for low-priority packet. The input process is adopted by discrete-time batch Markovian arrival process.