資料網格使得分散在不同區域上的計算及儲存資源(同質或異質)，可以達到分享、選擇、以及相互的溝通的應用。尤其是需要分析大量且密集資料的科學實驗，諸如高能物理、生物資訊的運用、以及氣象的模擬等，透過應用資料網格的都獲得良好的問題解決方式。近幾年來，許多研究學者持續地研究資料網格中副本機制的應用。副本機制可以讓一份資料產生多份的副本並且分散儲存於不同的地方，使得世界各地的研究學者可以容易取得所需的資料，縮短資料取得時間，提高資料分析的效率。由於網路的頻寬狀態屬於變動，使用者所需的資料會隨著時間變動，資料存取點亦需隨之變更。在本論文中，提出了一個基於動態管理服務與貝式網路下的一個副本維護策略。資料副本會動態的調整至適當的位置以供使用者使用，並避免存取次數與資料量隨著節點的增多而造成過多負擔。 Data grids are a very important and useful technique for solving problems created by the large amounts of data scientific experiments and simulations produce. Data replication, a technique much discussed by data grid researchers in past years, creates multiple copies of files and stores them in convenient locations to shorten file access times. In this thesis, we propose a dynamic replication maintenance service for maintaining data in grid environments. Replicas are adjusted to appropriate locations using Bayesian Networks (BN) due to the continuous change of the network. The maintenance strategy we propose is called Implicit Dynamic Maintenance Service with Bayesian Network (IDMSBN). If the probability value exceeds the user–defined threshold, the IDMSBN can duplicate the same file to the next best site without replication, in order to reduce the network bandwidth when the user requests a file.
