| 摘要: | 隨著雲端運算技術的蓬勃發展，軟體作為服務(SaaS)為雲端運算三大服務之一，資料可藉由雲端軟體儲存至雲端環境。如何讓儲存在雲端的資料能有效傳輸、索引與搜尋，為一重要研究議題。本研究以個人雲端(personal could)為例探討雲端軟體資料儲存與傳輸之問題，並且提出有效率的解決方法。
本研究提出個人雲端儲存管理模型機制，在本機制中有個人訊息擷取、訊息儲存處理及分散式索引管理等三個模組。個人訊息擷取模組之功能為從社群網站上擷取個人訊息(post)，其內容包含文字、圖片及連結等；為便於瀏覽與降低儲存空間成本，訊息儲存處理模組將擷取下來的訊息以倒傳遞類神經網路演算法判斷應否上傳到雲端儲存空間，若該則訊息應上傳則將該則訊息的JSON格式轉換成XML與BJSON格式後以訊息ID建立索引主鍵再傳送並儲存至分散式索引管理模組；分散式索引管理模組係將訊息儲存處理模組處理過的個人訊息依圖片、文字及連結做檔案類型的分類，再將訊息同步儲存到雲端環境設備中，亦可透過訊息ID可搜尋到該訊息。
本研究以Dropbox雲端儲存軟體為例進行實作，以索引方式進行檔案大小縮減並指向實體檔案進行同步儲存備份。本模型實作係以Windows 7為開發平台，Visual Studio 2010 C#為開發語言，分散式資料庫為本地端儲存資料庫;經本模型實作結果圖片檔案減少50%、文字檔案減少70%、圖片與文字混和檔案減少61%的儲存空間。
本研究具自動判別使用者檔案應否上傳，可避免過度的檔案同步儲存、並以訊息ID自動建立索引主鍵上傳並儲存至雲端，可節省雲端儲存空間、降低雲端儲存成本以及提升傳輸速度。
With the rapid development of cloud computing, Software as a Service (SaaS) is one of the cloud models as a service and platform for data storage. Therefore, it is a critical issue that how to effectively send and search data in cloud computing infrastructure. To develop an effective solution, this study focuses on the data storage and delivery in personal cloud.
There are three modules composed in our Personal Storage Management Model. The first one called personal post capture module, which function is to collect the personal post including characters, pictures and links etc, we develop a post process storage module to decide whether the system should upload the captured post to cloud storage with the algorithm of Back-Propagation Neural Network, BPN.
If the post should be uploaded, the module will transfer the JSON format to both XML and BJSON format, to build the index primary post ID and send it to save in the distribution index management module. The third module called distribution index management module which is to sort the type of file of processed personal post according to pictures, characters and links in the post process storage module. Then the post will be saved synchronously in cloud computing infrastructure and searched through post ID.
We use the Dropbox, cloud computing software, to do the implementation which resizing the file by index and synchronously saving the physical file as a backup file. Our model use Windows 7 as a development platform, also using Visual Studio 2010 C# as a development language, as well as using distribution database as a local storage database. As a result, we find that our model can do 50% space saving for picture files, 70% space saving for character files, and 61% space saving for the mix of picture and character files.
Our model can do automatic interpretation to decide whether the file should be uploaded or not, and to build the index primary key based on post ID to upload and save in cloud. Finally, our model can save much storage space, reduce storage cost and make uploading effectively. |
