摘要
由於科技日新月異，資料量日益劇增之下，雲端技術與應用近年來已經是許多企業的重點研究領域，需多知名企業如Google、Amazon、百度等，透過硬體以及軟體技術的提升，提供更優質的服務給使用者。以往的儲存容易受到空間以及硬體的價格的限制，如今Apache 軟體基金會提出的HDFS (Hadoop Distributed File System)分散式儲存系統，將過去集中儲存的模式轉變成分散式的儲存，使雲端儲存資源能夠更有效的利用及高效能、高容錯、高可靠、高可用及高擴充的特性。本論文實作一個以HDFS 為雲端儲存系統的影像監控系統，同時據此雛形系統模擬大型監控大量儲存的需求，分析檔案上傳與下載能力，比較在不同大小的檔案、資料備份數、與系統節點組態下，其傳輸能力對系統效能的影響。
並列摘要
With rapid development of technology, data volume is much bigger than before. Cloud technology and application in recent years has been a hot research topic not only for academic society but also for industries, e.g. Google, Amazon, to name a few, through ascension hardware and software technologies to provide better services to users. Traditionally, a high-end storage system with large storage capacity and fast data access is usually due to some sorts of customized design and thus cost much. Another important issues are such as scalability and faulttolerance. Hadoop Distributed File System (HDFS) provides a cost-effective solution to the needs. In this thesis study, we investigate design of video surveillance systems, especially in the realm of video data storage and access. We built a video surveillance system prototype which incorporated HDFS as the storage subsystem for video file access. Based on the prototype system, we made intensive experiments to evaluate video file performance under different system considerations such as CPU virtualization, degree of data duplication, file access pattern, and file size.