| 摘要: | 本計畫預計發展一個具容錯及保密能力之網路儲存系統，讓使用者不必將資料放置於自己的電腦上，隨時隨地只要透過網際網路連線就能夠存取自己的資料，充份達到資料可攜性的要求。而為了避免重要資料遺失，我們使用以編碼為基礎的資料備份(coding-based replication)技術，利用奇偶檢查(parity check)的概念來達到容錯的要求。因此，即使有部份資料遺失，使用者在大部分情況下仍能夠回復原始資料。不同於以資料副本為基礎的備份(replica-based replication)技術，以編碼為基礎的資料備份技術擁有較低的備份成本與較高的容錯能力。另外，由於編碼技術的使用，資料也可以平行地在不同位置存取，達到加速存取及負載平衡的效果。另一方面，為了保證資料的隱密性，避免資料被他人不當存取，我們採用位元區塊位元重排(permutation)的概念，配合奇偶檢查編碼，將資料打散分成不同的條狀(strip)儲存於不同的位置，再加上資料存取者的驗證(authentication)功能，大大增加資料的保密性。相較於傳統先使用加密技術將資料加密之後再儲存於網路儲存系統中的作法，我們發展的技術擁有較低的計算成本，但是仍然擁有一定程度的資料保密性。本計畫發展的方法可以應用在一般的網路儲存系統上，如儲存雲端系統(storage cloud)、資料格網(data grid)及同儕式(peer-to-peer, P2P)儲存系統上。由於在同儕式系統中的節點會頻繁的加入或離開節點，因此最不容易達到容錯功能；另外，由於同儕式系統缺乏集中式的管理設備，因此也最不容易達到資料保密的要求。因此，本計畫預計以同儕式儲存系統為例，說明所發展技術的執行細節。綜合而言，本計畫預計達成以下目標： 完成以編碼為基礎的資料備份(coding-based replication)技術，以一維及二維奇偶檢查位元的編碼概念來達到資料容錯目標。 完成位元區塊位元重排(permutation)及資料存取者驗證的功能，防止資料被不當使用者存取，達到資料保密的目標。 效能評估: 我們預計進行分析及執行模擬實驗以完成效能評估，來證實所發展的方法具有不錯的容錯性及保密能力以及較低的運作成本。 This project is intended to develop a fault-tolerance and secret networked storage system. With the system, users do not need to store their data locally. Instead, users can store the data in remote sites via Internet connection. Data thus have a very high degree of portability. The developed scheme achieves fault-tolerance by applying the coding-based replication technique. By appending few additional parity check bits, the original content can be recovered when some partial content gets lost. Unlike traditional replica-based replication schemes, the coding-based replication scheme has a lower replication cost and higher degree of fault-tolerance. Furthermore, by the coding-based replication scheme, data can be stored in different locations in parallel, accelerating the access process and bringing about load balancing. To prevent unauthorized users from accessing the stored data, data bits are divided into blocks and permuted and distributed into strips before storing into different locations. An access authentication scheme is further integrated into the stored data to make the data more secret. Comparing with traditional schemes that encrypt data before storing them, the developed scheme has lower computation overheads, but still has fair secrecy. The developed scheme can be applied to networked storage systems, such as storage clouds, data grids, and peer-to-peer (P2P) storage systems. However, it is harder for P2P storage systems to achieve fault-tolerance, since they have the highest churn rate. Furthermore, it is also harder for P2P storage systems to keep data secret, since they have no centralized control unit for authentication. The project therefore plans to take P2P storage systems as the test bed for the developed scheme. In summary, the project is intended to achieve the following goals. We will develop the coding-based replication scheme that applies the 1D and 2D parity bit check concepts for the purpose of fault-tolerance. We will develop the permutation scheme of block data bits and the data access authentication scheme for the purpose of preventing unauthorized users from accessing the stored data. We will do simulation experiments for performance evaluation to demonstrate that the developed scheme has high degree of fault-tolerance and secrecy and low cost. 研究期間：9908 ~ 10007 |
