雲端運算環境之彈性系統架構效能分析

Performance analysis of flexible system architecture in cloud computing environment

郭文洲(Wen-Chou Kuo)

指導教授 ： 林志浩

中原大學/商學院/資訊管理研究所/碩士(2017年)

https://doi.org/10.6840/cycu201700808

摘要

近年來，大數據分析需求的成長與大受歡迎的雲端計算，加上許多網路應用程式被運用於商業用途，如社交網路、 影音串流媒體服務和企業資料中心，資料中心網路的優勢是巨大的存儲空間，低計算延遲、低佈署成本和高可用性，但是對於大量的使用者及數據傳輸，也提高延遲網路擁塞和網路瓶頸，是雲計算中的一些問題，從集中式架構到分散式架構，邊緣計算可以將負載轉移到邊緣主機，可以減少應用程式反應時間並提高了整體使用率，而且建置所需費用可能會使得企業面臨抉擇，是要選擇有著不中斷服務提供的雲端資源，或是選擇自行架設效能相稍低於大型主機，較為低階版本的伺服器主機，或是採用最基本的Laptop來架設系統環境，但隨著資料成長快速，資料中心已處在過載環境下，持續的擴展資料中心處理規模，也消耗了大量的資源，為此在本文中，我們分別測試雲端計算、霧計算與露計算，針對Hadoop執行wordcount所需時間，來決定建議客戶採用哪種建置方式，以符合客戶需求並減少建置成本的規劃，對於現在普遍處於資訊過載的數據中心是相當重要的議題。

關鍵字

雲端計算 ； 系統架構 ； 效能分析 ； Hadoop

並列摘要

In recent years, the growth in needs of big data analytics and cloud computing services promoted numerous commercial applications, such as social networks, audio and video streaming services, and enterprise data centers. Advantages of data center networks are its enormous storage space, low delay, low deployment cost, and high availability. However, a large amount of users and data transmission deteriorates transmission delay, network congestion, and performance bottlenecks in cloud computing environment. Edge computing architecture transfers computational loading from centralized servers to a distributed edge hosts in order to reduce application response time and increase overall usage. With rapid growth in amount of data, there are several choices for companies to development their system architecture, which can be cloud resources, their own medium servers, or basic personal computers. To solve the problem, we evaluated and compared the performance among cloud, fog, and dew computing environments in this thesis. According to the required time for performing a set of wordcount testing on Hadoop, the experimental results recommend several architectures to satisfy users’ performance requirements and cost efficiency. This is a fairly important topic for companies to deal with their big-data computation.

並列關鍵字

Cloud computing ； System architecture ； Performance analysis ； Hadoop

參考文獻

[1] Amir, Taherkordi and Frank, Eliassen, “Data-Centric IoT Services Provisioning

[3] Xuyi Wei, ” Based on VMware technology's Campus network cloud platform

technology research.” Computer Science and Electronics Engineering

IEEE International Conference on, pp. 170-175, December. 2016.

[5] Zheng Li et al. “Performance Overhead Comparison between Hypervisor and