雲端運算即時轉移記憶體預測機率之改善研究

隨著雲端運算技術的成熟，近年來越來越多的雲端運算應用產生，其中雲端運算應用的系統虛擬化技術，變得相當重要。伺服器虛擬化可以提高伺服器服務效能與提供動態資源調度的特性。但伺服器會發生維修、負載過量或是毀損的情況，造成用戶出現服務中斷的狀況。針對此問題，發展出即時轉移技術，在狀況發生時，能夠持續用戶端使用，而不會感到服務中斷。此篇論文在介紹即時轉移技術中，利用預測記憶體修改機率，判別頁面是否暫停-複製轉移(Stop-and-Copy Phase)。根據實際記憶體轉移模型，我們參考Markov model 和Gilbert-Elliot Model 來實際探討預測機率的狀況，分析我們計算出來的結果，可以得到更精準的預測機率數值。根據這些數值，在未來討論設定停機時間點的問題中，可以更精準的找出來，用以減少總轉一的時間點，提供伺服器運作的效能和速度。

關鍵字

雲端運算；即時轉移記憶體預測機率

並列摘要

With cloud computing technology matures, more and more in recent years, generating cloud computing applications, including cloud computing applications, system virtualization technology has become very important. Server virtualization can improve server performance and provide dynamic resource service scheduling features. But the server will occur repair, damage or excessive case load, causing the user interruption in service conditions. To solve this problem, the development of real-time transfer of technology, in the condition occurs, the client can continue to use, and will not find the service interruption. In this paper introduces instant transfer technology, the use of predictive probability memory modify, suspend determine whether the page - copy transfer (Stop-and-Copy Phase). According to the actual memory transfer model, we refer Markov model and Gilbert-Elliot Model to predict the probability of the actual study, and to analyze our calculated results, you can get a more accurate prediction probability values. Based on these values, set down in the next time point discussed problems, can be more accurate to find out, to reduce the total transfer time of one point, to provide the performance and speed of operation of the server.