UCS: 一個雲端硬碟整合系統

隨著雲端硬碟的種類越來越多，使用者管理不同種類的雲端硬碟帳號開始變得繁複且吃重。雖然目前在市面上有很多雲端硬碟整合系統，但它們皆有以下的問題：〈一〉每個帳號在整合的系統內仍然是分開的，雖然省去了需要重複登入多組帳號的麻煩，但在使用者要上傳一個檔案的時候，系統並沒有辦法自動替使用者決定該檔案要上傳至哪個雲端硬碟帳號，尤其在所有帳號都快要滿的狀況，這個問題對使用者來說會非常麻煩〈二〉承一，當想要上傳的檔案放不進特定的帳號時，使用者必須自己手動將該帳號內的檔案移動到其他雲端硬碟帳號。若是其他帳號也都快滿了，這個工作對使用者來說負擔非常大〈三〉承一，因為每個帳號在使用者介面中被區隔開，使用者並不能完全按照自己的邏輯去對檔案樹進行管理。〈四〉承一，當一個檔案太大，無法放入當前任何帳號之時，有時候所有帳號加起來的整體空間是足以容納該檔案的，但是在當前之整合系統中並不會處理此類型的問題〈五〉上傳或者移動已經存於某雲端硬碟帳號中之檔案時，當前的系統皆是先將檔案內容上傳到整合系統，而整合系統再將資料上傳至雲端硬碟，如此增加了不必要的時間與成本上的浪費。於是我們提出了一個能夠解決上述問題的雲端硬碟整合系統：〈一〉提供單一個使用者能夠完全依照自己邏輯編輯的檔案系統，不用擔心檔案實際上放在哪個雲端硬碟帳號〈二〉檔案上傳時，系統將替使用者進行排程，決定檔案要放入何帳號〈三〉帳號空間不足時，系統會自動排程將該帳號之部分檔案移動到其他帳號，以容納當前上傳的檔案。必要時部分被移動之檔案可能被分割成數個小檔案後再分別被移動到不同的帳號。〈四〉上傳之檔案，其資料將從使用者瀏覽器直接傳至雲端硬碟帳號。已經在雲端硬碟中檔案之移動，系統會盡可能的將其工作分配給使用者之瀏覽器，以降低伺服器之工作量。對於客戶端瀏覽器因諸多原因無法執行的工作，伺服器才會介入，並以伺服器與客戶端平行化的方式進行檔案搬移工作。

關鍵字

雲端硬碟；網路服務

並列摘要

There are more and more cloud storage services in recent years, these services enable the users to: (1) Upload/Download Files through their browsers from anywhere in the world. (2) Multi-user editing in the browser for certain file types. (3) Single-user editing in the browser. (4) File Preview. For instance, most image files can be previewed on cloud storage providers. (5) Sharing file tree with other users and managing it together. Although the web interfaces of cloud storage providers are quite convenient for users with only one account, it becomes a very tedious task to manage multiple accounts that may be from different providers. This is where cloud storage integration services come in, they provide a single interface to manage multiple cloud storage accounts. However, the story doesn’t end here, there are several issues that these cloud storage integration services do not address. (1) A user still needs to decide which account to upload to, the services do not recommend which account to use. When the accounts are nearly full, it becomes a demanding task to track which account still has enough free quota. Especially when the user has other factors to consider, such as web edit/preview interface, sharing…etc. (2) When an account doesn’t have enough space for a file, user needs to move files around accounts to make space. (3) Each account is treated as a distinct folder by the integration services, it is impossible to have a project folder that contains both a Google document file hosted by the Google Drive and a Microsoft Word file hosted by the OneDrive. (4) When all accounts are nearly full and a user wants to upload a file, the fragmentation problem occurs. The integration services do not chunk the file into pieces to fit the file across multiple accounts. (5) Uploading and moving files across accounts are always performed by the servers. This might become a bottleneck when there are a lot of users uploading files at the same time. Although these integrators could utilize dynamic cloud resources to maintain their QoS, it still results in unnecessary cost. In this thesis, we propose a new architecture for integrating cloud storage services. It includes the following features. (1) A single logical view for files among multiple accounts. The users can put their files wherever they want. (2) A user’s logical view can be shared by other users. (3) Whenever a user uploads a file, our system will decide which account should store the file automatically. If the chosen account does not have enough space for the file, the system will move files into other accounts to make space for the file. (4) The system may chunk some files into pieces and distribute those pieces across multiple accounts when it is necessary. (5) When uploading a file, the file data is sent to the corresponding storage provider directly. Our system does not cache the file data. When moving files among accounts or storage providers, the works are distributed and performed by both our server and the client in parallel.