透過您的圖書館登入
IP:3.21.248.119
  • 學位論文

以共享記憶體架構為基礎之分散式行動計算環境之研究

The Research of Distributed Mobile Computing Based on Shared Memory Architecture

指導教授 : 梁文耀

摘要


本論文提出一個實作於行動裝置上之分散式行動計算環境-D2MCE (Dynamic Distributed Mobile Computing Environment)。D2MCE利用分散式共享記憶體(Distributed Shared Memory, DSM)為系統基礎架構。由於本系統採用Mobile Ad-hoc Network (MANet)為其通訊環境,基於節點可移動之特性,所構成的分散式行動運算組態有可能隨時改變,亦即工作節點(Node)可能於任何時刻進入或離開計算工作群組的傳輸範圍,因此本系統提供一個能於運算期間以動態方式加入或離開計算工作群組之機制,並於離線時將各管理權責轉移,例如共享資料管理者、群組協調者、同步協調者與管理者等各項權責。此外,系統中提出三項運算最佳化方式,藉以減少訊息傳輸量,包括寫入標記、修改共享資料之差異版本控制與動態共享資料管理者。1.利用寫入標記機制減少共享資料回寫至資料管理者(Home)的次數。2.透過版本的判別,僅更新不同版本間修改的資料內容,以降低工作節點間網路傳輸資料量。3.移動Home至存取頻率最高之節點,以降低遠端資料存取次數。由測試結果顯示,加入最佳化方法後此系統能有效提升整體運算效能。

並列摘要


D2MCE (Dynamic Distributed Mobile Computing Environment), a system based on DSM (Distributed Shared Memory) and implemented on mobile devices, is proposed in my thesis. This system adopts Mobile Ad-hoc Network (MANet) as its network architecture. Due to the feature of mobile nodes, a distributed mobile computing group constructed by the nodes can possibly change any time, that is, nodes can join or leave the transmission scope of the computing group. For the reason, this system provides a mechanism allowing users dynamically join or leave a computing group during the computing period, and transferring the authority of management such as Home, Group Coordinator, Synchronization Coordinator and Manager while nodes are leaving. Besides, there are three optimized methods offered in D2MCE to reduce the number and the amount of messages, including Writing-labeling, Diff Version Control and Moving Home. First, the mechanism of writing-labeling can decrease times that shared data are written back to Home. Second, by judging the difference of versions, nodes can merely update the modified content among different versions to reduce the amount of messages. Third, the method of Moving Home can decrease times of dealing remote shared data by moving Home to the node with the most frequency of access. From the testing result, it is found that adopting optimized methods can effectively improve performance.

並列關鍵字

D2MCE DSM MANet Diff Version Control Moving Home

參考文獻


[1] T. Abe, and S. Okamoto, “A Moving Home-based Software DSM System”, Proceedings of the IEEE Pacific Rim Conference on Communications, Computers and signal, Vol. 1, Aug. 2003, Page(s): 17 - 20.
[3] A. L. Cox, E. D. Lara, C. Hu, and W. Zwaenepoel, “A Performance Comparison of Homeless and Home-based Lazy Release Consistency Protocols in Software Shared Memory”, Proceedings of the 5th International Symposium On High Performance Computer Architecture, Jan. 1999, Page(s): 279 - 283.
[4] L. Dagum and R. Menon, “OpenMP: An Industry Standard API for Shared-Memory Programming”, Proceedings of the IEEE Computational Science and Engineering, Vol. 5, Issue 1, Jan. 1998, Page(s): 46 - 55.
[7] A. Geist, and V. S. Sunderam, “Network Based Concurrent Computing on the PVM System”, Journal of Concurrency: Practice and Experience, Vol. 4, Issue 4, Jun. 1992, Page(s): 293 - 311.
[8] K. Gharachorloo, D. Lenoski, J. Laudon, P. Gibbons, A. Gupta, and J. Hennessy, “Memory Consistency and Event Ordering in Scalable Shared-Memory Multiprocessor”, Proceedings of the 17th Ann. International Symposium Computer Architecture, May 1990, Page(s): 15 - 26.

延伸閱讀