以熱門程度為基礎在雲端運算環境中建構可自適應檔案複製機制

Popularity-based Adaptive Replica Mechanism underlying Cloud Computing Environment

陳聖中(Sheng-Zhong Chen)

指導教授 ： 王淑卿

朝陽科技大學/資訊學院/資訊管理系/碩士(2013年)

摘要

隨著時代的演進，電腦科技技術的成長。對於預算不足無法建構超級電腦卻又需要大量資料運算的使用者有了「雲端運算」新的選擇。然而，當越來越多人使用雲端運算，而使用者將其個人的相關資訊儲存在雲端運算環境時，雲端運算系統的可用度與可靠度就成為提供雲端服務時必須考慮的重要因素之一。在雲端運算中為了維護資料檔案的可用性，因此可透過複製備份的方式，增加資料的副本。由於在雲端運算中的實際應用中，副本的儲存如果不一致則使用者就無法得到可靠的服務。換言之，透過副本的一致性才能提供可靠的服務給使用者。使用不同的副本一致性策略，將產生不同的影響。如同步的策略是強壯的一致性策略，對分散式系統的效能要求太高；而非同步的策略是虛弱的一致性策略，在更新檔案的速度較慢。然而，在雲端快速變化的環境中，不同的一致性並不能反應雲端服務的使用狀況。
因此在本研究中使用以熱門程度為基礎，在雲端運算環境中建構『可自適應檔案複製機制(PARM)』，在PARM中分為兩方面去執行檔案複製機制，分別是以原子衰退值調整文件複製策略(ARAFR)與以熱門程度的浮動門檻值進行雲端運算中檔案一致性的快速調節策略(RCASP)。透過ARAFR決定副本的數目，再依照RCASP來快速調節副本的一致性。其中ARAFR可以提高雲端運算環境中的可用性，而RCASP可以確保雲端運算環境中的可靠性。整體而言，本研究所提出的PARM透過檔案熱門程度的評估，可以快速將較熱門的檔案更新至各個資料副本，而較冷門的檔案也不會占據更新時的效能，更進而可以達到雲端環境中的資源平衡，並且依據使用者的熱門程度去達到快速回應熱門檢索。

關鍵字

雲端運算 ； 原子衰退 ； 副本一致性策略 ； 副本複製策略 ； 資料副本

並列摘要

Nowadays, the computing technology was growing up. There are many related applications that require large amounts of data computing. However, Cloud computing is a new choice. In addition, when more and more users use the services provided by Cloud Computing, the personal information of users are stored in the Cloud Computing. Then, the reliability and availability of Cloud computing is an important issue. However, to increase the number of replica in the Cloud computing can provide high availability of services. As Cloud computing applications, user will not be able to get reliable service when the replica is inconsistent. In other words, a reliable service is provided to the user through the consistency of the replica. Using different replica consistency strategy will produce different effects. Such as synchronous consistency is a strong consistency, the performance requirements of the distributed system are high. In addition, the asynchronous consistency is a weak consistency strategy, it can slow the speed of the files updated.
In this study, a Popularity-based Adaptive Replica Mechanism (PARM) is proposed underlying Cloud computing environment. There are two phases of PARM. The first phase is Atomic Recession based Adjustable File Replication (ARAFR), and the second phase is Rapid Consistency Adjustment Strategy based on Popularity (RCASP). According to the popularity, the ARAFR is used to determine the number of replica. Then, the RCASP can used to adjust the consistency of replica rapidly. However, ARAFR can ensure the availability and RCASP can ensure the reliability in the Cloud Computing environment. However, by using PARM, the replica can be updated rapidly depend on the popularity. In addition, the colder replica will not to occupy the performance in the update time.

並列關鍵字

Cloud computing ； Replica strategy ； Replica consistency ； Atomic recession ； Data replica

參考文獻

[3] K. Birman, G. Chockler and R. Renesse(2009), “Toward A Cloud Computing Research Agenda,” ACM SIGACT News, Vol. 40, No. 2, pp. 68-80.

[4] R. Buyya, C.S. Yeo and S. Venugopal(2009), “Market-oriented Cloud Computing: Vision, Hype, and Reality for Delivering IT Services as Computing Utilities,” Proceedings of the 10th IEEE International Conference on High Performance Computing and Communications, pp. 5-13.

[5] B. Cai, C. Xie, and G. Zhu(2007), “EDRFS: An Effective Distributed Replication File System for Small-File and Data-Intensive Application,” Proceedings of the 2007 2nd International Conference on Communication Systems Software and Middleware, pp. 1-7.

[8] R.L. Grossman(2009), “The Case for Cloud Computing,” IT Professional, Vol. 11, No. 2, pp. 23-27.

[9] R.L. Grossman, Y. Gu, M. Sabala and W. Zhang(2009), “Compute and Storage Clouds Using Wide Area High Performance Networks,” Proceedings of the Future Generation Computer Systems, Vol. 25, No. 2, pp. 179-183.