(3.238.7.202) 您好!臺灣時間:2021/03/01 23:53
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

我願授權國圖
: 
twitterline
研究生:廖政順
研究生(外文):Jean-Shun Liao
論文名稱:在分散式共享記憶體系統上支援多層次重組機制
論文名稱(外文):Supporting Multi-layer Reconfiguration Mechanism on Distributed Shared Memory System
指導教授:謝錫
指導教授(外文):Ce-Kuen Shieh
學位類別:碩士
校院名稱:國立成功大學
系所名稱:電機工程學系
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:1999
畢業學年度:87
語文別:英文
論文頁數:39
中文關鍵詞:分散式共用記憶體重組機制多層次
外文關鍵詞:Distributed Shared MemoryDSMReconfigurationMulti-layer
相關次數:
  • 被引用被引用:0
  • 點閱點閱:133
  • 評分評分:系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔系統版面圖檔
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
在分散式系統中有許多平行程式。這些程式可以被分為兩類,一種是客端應用程式而另一種是主端應用程式。 我們把分散式共用記憶體系統上的應用程式視為是客端應用程式。分散式共用記憶體系統必須提供動態重組機制去提供更多的資源給主端應用程式,而且能嘗試著增加客端應用程式的效能。我們提出一種有效的多層次重組機制。它包含了虛擬處理器重組、平行單元重組及節點重組等三個方針。假如主端應用程式的負載比較輕,則客端應用程式會刪除部分的虛擬處理器去釋放更多的資源給主端應用程式。假如主端應用程式的負載較重的話,我們繼續利用平行單元的搬移來增加客端應用程式的效能。假如主端應用程式的負載太重的話,而且搬移平行單元已經無法增加客端應用程式的效能時,客端應用程式就必須放棄那一台機器。在我們的多層次的重組機制中,我們有較好的機會去平衡主端應用程式和客端應用程式的效能。我們在Teamster上實現這個機制。Teamster是一個分散式共用記憶體系統,它使用了四台x86的對稱式多處理機。每一台機器用100Mbps的網路連接,而且作業系統是SUN Solaris 2.6的版本。客端應用程式利用我們的機制分別與五個主端應用程式一起執行。在我們的測試中,假如應用我們的多層次重組機制,主端應用程式在最好的情況下可以增加60%的效能,而客端應用程式在最差的情形下損失10%的效能。

There are many parallel programs running among distributed systems. These applications can be separated into two categories, one is guest application and the other is host application. We regard Distributed Shared Memory (DSM) applications as guest applications. The DSM system must provide the reconfiguration mechanism dynamically to release more resource for host applications and try to improve the guest applications’ performance. We introduced an efficient Multi-layer reconfiguration mechanism, which contains three policies, virtual processor reconfiguration, parallelism reconfiguration, and node reconfiguration. If the host application’s load is lower, the guest application will delete some virtual processors to release more resource for host application. If the host application’s load is heavy, we continue to use parallelism migration to improve the guest application’s performance. If the host application’s load is too heavy and parallelism migration can’t improve guest application performance, guest application will give up using the computer. We have a better chance to balance the performance between host applications and guest applications in our Multi-layer reconfiguration. We implement this mechanism on Teamster that is a DSM system running among four x86 SMP machines connected with 100 Mbps Fast Ethernet. Each node’s operating system is SUN Solaris 2.6. The guest applications adopt our mechanism and run with five host applications individually. If we adopt our Multi-layer reconfiguration mechanism, the host application performance can be improved 60% at the best case and the guest application performance only slows down 10% at the worst case in our evaluations.

CONTENTSI
ILLUSTRATIONSIII
CHAPTER 1INTRODUCTION1
1.1RECONFIGURATION1
1.2OUR PROPOSAL2
1.3RELATED WORK3
CHAPTER 2MULTI-LAYER RECONFIGURATION5
2.1MULTI-LAYER RECONFIGURATION5
2.1.1Virtual processor5
2.1.2Parallelism6
2.1.3Node7
2.2DESIGN ISSUES7
2.2.1Kernel involved and user-level7
2.2.2Asynchronous and synchronous method8
2.2.3System load monitor and keystroke9
CHAPTER 3IMPLEMENTATION10
3.1RECONFIGURATION PROCEDURES10
3.2RECONFIGURATION MECHANISM12
3.2.1Experimental Environment : Teamster12
3.2.2Keystroke13
3.2.3Memory Consistency14
3.2.4Communication17
3.2.5Synchronization19
3.2.6Thread System22
CHAPTER 4PERFORMANCE26
4.1APPLICATIONS26
4.2PERFORMANCE EVALUATION29
CHAPTER 5CONCLUSIONS AND FUTURE WORK35
5.1CONCLUSIONS35
5.2FUTURE WORK36
BIBLIOGRAPHY37

[ Bershad 88 ]Brain N. Bershad, Edward D. Lazowska, and Kenry M. Levy.
PRESTO : A system for object-oriented parallel programming
Software-Practice and Experience, 18(8), August, 1988
[ Casas 94]Jeremy Casas, Ravi Konuru, Steve W. Otto, Robert Prouty, and Jonathan Walpole
Adaptive Load Migration systems for PVM
Supercomputing '94., Proceedings , Page(s): 390 —399
[ Chang 98 ]J. B. Chang, Y.J. Tsai, C.K. Shieh, and P.C. Chung
An Efficient Thread Architecture for a Disrtibuted Shared Memory on Symmetric Multiprocessor Clusters
Parallel and Distributed Systems, Page(s): 816-824
[ Itzkovitz 98]Ayal Itzkovitz, Assaf Schuster, Lea Shalev
Thread migration and its applications in distributed shared memory system
The Journal of Systems and Software 42 (1998) 71-87
[ Lai 94 ]An Chow Lai
Design and Implementation of Release Consistency Protocol on Cohesion
Master thesis, Department of Electrical Engineering, National Cheng Kung University, R.O.C, 1994.
[ Lai 97 ]An-Chow Lai; Ce-Kuen Shieh; Yih-Tzye Kok
Load balancing in distributed shared memory systems
Performance, Computing, and Communications Conference, 1997. IPCCC 1997., IEEE International , Page(s): 152 -158
[ Li ]Li Kai
Shared Virtual Memory on Loosely Coupled Multiprocessors.
Ph.D Thesis, Yale Univ, 1986.
[ Litzkow 88]Michael J. Litzkow, Miron Livny, and Matt W. Mutka
Condor — A Hunter of Idle Workstations
Distributed Computing Systems, 1988., 8th International Conference on , Page(s): 104 —111
[ Liu 97 ]Wu-Hwa Liu
Adding Run-Time Scalability to Cohesion
Master thesis, Department of Electrical Engineering, National Cheng Kung University, R.O.C, 1997.
[ Morin 97 ]Christine Morin and Isabelle Puaut
A survey of Recoverable Distributed Shared Virtual Memory Systems
Parallel and Distributed Systems, IEEE Transactions on Volume: 8 9 , Page(s): 959 —969
[ Ryu 98 ]Kyung Dong Ryu, Jeffrey K. Hollingsworth
Linger Longer: Fine-Grain Cycle Stealing for Networks of Workstations
SC’98. Nov. 1998, Orlando.
[ Ryu 99]Kyung Dong Ryu, Jeffrey K. Hollingsworth, Peter J. Keleher.
Mechanisms and Policies for Supporting Fine-Grained Cycle Stealing
1999 ACM International Conference on Supercomputing (ICS’99), Rhodes, Greece, June, 1999.
[ Shieh 95 ]Ce-Kuen Shieh, An-Chow Lai, Jyh-Chang Ueng, Tyng-Yue Liang, Tzu-Chiang, Su-Cheong Mac.
Cohesion : An Efficient Distributed Shared Memory System Supporting Multiple Memory Consistency Models
The First Aizu International Symposium on Parallel algorithms/Architecture Synthesis, Aizu-WaKa-matsu, Fukushima, Japan, March, 1995.
[ Speight 97 ]Evan Speight and John K. Bennett
Brazos : A Third Generation DSM system
Proceedings of the First USENIX Windows NT Workshop, August, 1997
[ Theimer 88]Marvin M. Theimer, Keith A. Lantz
Finding Idle Machines in a Workstation-based Distributed System
Software Engineering, IEEE Transactions on Volume: 15 11 , Page(s): 1444 —1458
[Thitikamol 97 ]Kritchalach Thitikamol and Pete Keleher
Multi-threading and Remote Latency in Software DSMs
Distributed Computing Systems, 1997., Proceedings of the 17th International Conference on , Page(s): 296 -304
[ Thitikamol 99]Kritchalach Thitikamol and Pete Keleher
Thread Migration and Communication Minimization in DSM systems
Proceedings of the IEEE Volume: 87 3 , Page(s): 487 -497
[ Tsai 97 ]Yu-Jia Tsai
An Efficient Thread architecture for a Distributed Shared Memory on Symmetric Multiprocessor Cluster
Master thesis, Department of Electrical Engineering, National Cheng Kung University, R.O.C, 1997.
[ Yang 97 ]Kuo-Cheng Yang
Adding Symmetric Multiprocessor Supporting on C-thru
Master thesis, Department of Electrical Engineering, National Cheng Kung University, R.O.C, 1997.

QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
系統版面圖檔 系統版面圖檔