以分散式檔案系統探討雲端儲存的架構與應用－以GPFS為例

Abstract

對一個企業來說，資料是最重要的資產，近年來各種天災人禍造成的資料遺失，導致許多企業蒙受巨大的損害，許多人開始思考資料安全的問題，資料備份到哪裡才是安全的？隨著Big data時代的來臨，與雲端概念的興起，新的儲存模式不斷被提出，分散式系統高效能與高容錯的特性為資料儲存提供一個不錯的選擇。本研究透過研究分散式檔案系統的特性，提出一套具備高可靠、高可用以及具備Scale-out架構的高擴充度虛擬化儲存檔案系統，整套系統具有高度的容錯能力，即使任何一個區域的機房因為天災或任何不可抗力因素，導致區域性的停止服務，整套系統還是能夠依靠另外兩個區域的正常運作，繼續提供服務，對於一個雲端儲存服務來說，這是一個非常重要的關鍵因素。不過在本研究的測試中，可以發現整個檔案系統的效能因受限於網路的效能，不論是區域網路設備還是廣域網路路由的規劃，都是決定系統效能的關鍵，根據本研究所測試的效能，能夠提供一個小規模企業的私有雲使用，如果要運用在較大的規模，必須改善其網路環境。本研究亦架設了常見的雲端儲存服務如WebDAV、Web-Base file manager、Virtualization，來驗證此一架構的應用靈活度。

Business data is the most important asset of an enterprise. Many enterprises suffered great damage due to data loss caused by various natural disasters in recent years. People start thinking about the issue of data security and backup solution. With the era of the Big data coming and the rise of the cloud concept, a new storage model is proposed. Distributed systems, high performance and high fault tolerance features provide a good choice for data storage. This study implements a cloud storage with high reliability, high availability, high scalability, scale-out architecture, and virtualization. The entire system has a high degree of fault tolerance. No matter which one server room of a region failed due to natural disasters or any reason, resulting in regional failure. The entire system is able to rely on the normal operation of the other two regions and continue to provide services for a cloud storage service. This is very important factor for a cloud storage service. However, the performance will be limited by network, whether LAN equipment or WAN route planning is the key factor to determine the system performance. The system in this study can provide a private cloud for a small-scale enterprises. The network environment must be improved if it used in public cloud or large-scale environment. This study also implements some common cloud storage services such as WebDAV, Web-Base file manager and virtualization to verify the flexibility of this system.