一個兼具搜尋速度與儲存效率的容錯雲端架構

Abstract

隨著雲端運算的迅速發展，網路上有越來越多的雲端服務興起，雲端儲存是其中一項熱門的服務之一，如電子信箱、網路相簿等常見的應用，雲端儲存已融入人們的日常生活中，因此，雲端的搜尋速度和容錯能力是本論文研究的重要議題。由於傳統的欄位基礎資料模式(Column-Based Data Model)在稀疏矩陣（Sparse Matrix）的情況下會有搜尋效率不佳的問題，因此我們提出一個改良式欄位基礎資料模式，再搭配由網路編碼（Network Coding）原理建立的穩健性雲端備援機制，建構出一個可以提高搜尋速度和容錯能力的雲端架構，以達到系統較佳的執行效率，並確保使用者在雲端儲存的資料完整性和資料可還原性。經由本論文實驗證明，在資料矩陣的空間使用率少於50%時，改良式欄位基礎資料模式比欄位基礎資料模式減少了41%的搜尋時間，另外在容錯需求不變的情況下，針對不同的資料欄位數和允許故障數，我們所提出的備援機制比傳統備份機制所需的儲存空間分別節省了至少33%和50%。

With the advance of the internet and cloud computing environment, there have been more and more cloud service. Cloud storage is one of the popular cloud services, such as email and web album are common to see. Cloud storage has become a part of our life. In this thesis, we focus on the issue of improving the retrieval speed and the fault tolerance of the cloud storage in this thesis. As a result, we proposed a framework to enhance the retrieval speed and the fault tolerance of the cloud storage. The framework is composed of a new modified column-based data model and a new backup mechanism by using network coding. Because the traditional column-based data model has low performance in handling sparse matrix, we designed the new modified column-based data model to reform the problem. The proposed backup mechanism improves the fault tolerance of the storage system, and it also ensures the integrity and reusability of user data in cloud storage. In the results analyses, the modified column-based data model use less 41% retrieval time than the traditional column-based data model when the space utilization rate of the data matrix is less than 50%. When the fault tolerance requirements are the same, the proposed backup mechanism saves 33% and 50% more spaces than the traditional backup mechanism according to different column bases and allowable machine faults respectively.