具有限緩衝器和批次布松到達的寬頻整合服務網路系統之分析研究

Abstract

在本論文中，我們分析了具有限緩衝器和批次布松到達的寬頻整合服務網路的系統效能。依照服務特性的不同，進入網路的顧客被區分成兩種型式：即時性和非即時性資料。批次成任意正整數分佈是導致於封包切割。因此，這樣的分佈可適當的描述從窄頻至寬頻各種不同速率的服務。網路提供兩組有限的緩衝器，以調節即時性和非即時性資料；有多個服務時槽以同步運作；及以先到先服務的原則傳送封包。 本論文討論了三個相關的主題： ˙在第一個主題中，我們假設兩種型態的批次布松到達程序是互相獨 立的，且即時性比非即時性服務有較高的優先優。我們導出了幾個重要的系統參數，包括被阻絕的機率，封包的延遲和系統輸出。 ˙在第二個主題中，我們提出一種頻寬配置的策略以替代原先的優先策略。此種頻寬配置策略對即時性和非即時服務的系統效能的消長影響，有詳細的討論。最佳的頻寬配置參數也有深入的討論。 ˙在第三個主題中，除了即時性服務的到達程序被修正成MMBPP (Markov-Modulated Batch Poisson Process)外，其他的假設和第一個主題相同。MMBPP程序是用來描述即時性服務速率變化的時間關聯性。藉著和第一個主題的結果比較，MMBPP程序如何影響系統效能，有詳細的探討。 每個主題除了理論推導之外，並以模擬的方式來驗證分析的正確性。

In this thesis, we analyze an integrated services system with finite buffers and batch Poisson arrivals in broadband networks. According to the characristics of services, customers arriving at the system are classified into two types: real-time (type-1) and nonreal-time (type-2). The arbitrarily positivevalued batch distributions of the message size are due to packetizations and can properly describe the multirate services from NB (narrow-band) to WB wide-band). Two separate buffers are provided to accommodate each type of traffics. The system operates synchronously with multiple servers, accepts messages in the complete rejection strategy and serves each type of customers in the first-come-first-serve (FCFS) discipline. Three related topics are studied in this thesis. ˙ In chapter 2, both of the arrival processes are assumed independent batch Poisson. For the time-critical requirement of the type-1 traffic, we straightforward let the type-1 traffic have higher priority over type-2. Several system measures are derived, including the blocking probabilities, the packet delay and the throughput. ˙ In chapter 3, in order to provide some protection to the type-2 traffic, a bandwidth allocation strategy with cutoff parameter (c2, α) is proposed instead of the priority strategy. The tradeoff effects of the cutoff parameter on system measures are illustrated. And the optimal (c2, α) parameters are discussed. ˙ In chapter 4, the type-1 arrival process is modified to the Markov-Modulated batch Poisson process (MMBPP) to describe the inherent variation of the arrival rate of the type-1 traffic. While the type-2 process is still assumed batch Poisson. The priority schedule favoring the type-1 traffic is also used. How the correlated type-1 process influ- ences the system measures is discussed by compared with those results under the batch Poisson mentioned in chapter 2. In the preceding analyses, a two-dimensional or even three-dimensional Markov chain must be computed. This usually results in computational in-tractability. This problem is successfully solved by the decomposition method or approximation. Simulations are also conducted to verify the validity of the analyses.