雲端運算混搭服務性能分析技術之研究

Abstract

近年來，雲端運算 (Cloud Computing) 結合智慧型手機的應用快速發展。現今很多的應用結合多個雲端服務以更多樣化且方便的形式提供給使用者。在這些應用之中，即時性是相當重要的議題，例如以位置為基礎的服務 (Location-Based Services, LBSs)。在這篇論文中，我們將討論一個由多個雲端服務所構成的混搭系統 (Mashup System) 的效能分析技術。本論文提出一個利用排隊網路 (Queueing Network) 的分析技術以評估針對在不同的服務要求，如何保證到達使用者需要的服務品質。在此分析模組下，我們分別探討單一類別的要求 (Single-Class Traffic) 和兩種類別的要求 (Two-Class Traffic) ，討論在不同的服務要求速率下，如何適當的調整混搭中心 (Mashup Center) 的虛擬伺服器 (Virtual Machine, VM) 的個數以滿足使用者的服務品質需求。利用模擬的結果，我們證明所提出的方法的正確性。因此，本論文所建議之以排隊理論的分析模型可以提供一個很好的混搭多雲系統的性能分析。

In this thesis, we apply Jackson's network queueing theorem to model the service mashup cloud computing environments. The key challenge in providing new mashup mobile applications in cloud computing, such as the real-time location-based services, lies in quantifying the delay resulting from integrating multiple cloud servers. Furthermore, it is necessary to consider the effects of adjusting the number of virtual machines (VMs) on the quality of service (QoS) of mobile applications for various traffic loads. However, an effective analytical model to characterize both the effects of integrating multiple cloud servers and scalable VMs is lacked in the literature. The proposed mashup multi-cloud analytical model can calculate the service waiting time for various numbers of VMs and different arrival rates. By simulations and analysis, we show that our model can accurately predict when the waiting time performance in mashup cloud servers will increase sharply for various numbers of VMs and traffic loads. Hence, the proposed mashup multi-cloud analytical model can facilitate the resource management design in future cloud data centers.