在本文中，首先我們提出了一個移除原本CDPthread中在Execution Engine數量上限制的方法，也就是讓Execution Engine 動態啟動而不是靜態啟動。接著我們又提出了一個改善工作元間彼此工作量平衡的方法，並且將此技術帶入CDPthread系統，藉以改善CDPthread的效能。我們所提出的方法會持續追蹤每個節點的工作量來了解接下來的工作必需要交給那一個節點。更精確的來說，我們也會依據每個節點的核心數量，按照比例分配工作。從最後的實驗結果來看，跟原本靜態啟動Execution Engine 的CDPthread比較，我們損失了一些效能。而在有負載平衡的新CDPthread 中發現效能比沒有負載平衡的新CDPthread提昇了大約25%到48%的效能。而新的CDPthread現在可以處理任何線程數量的需求。

In this thesis, we first propose a modified version of the CDPthread to eliminate the restriction on the number of execution engines supported—by dynamically instead of statically allocating the execution engines to a process. Then, we describe a method to balance the workload among nodes under the control of the modified CDPthread to improve its performance. The proposed method keeps track of the workload of each node and decides to which node the next job is to be assigned. More precisely, the number of jobs assigned to each node is proportional to, but not limited to, the number of cores in each node. Our experimental results show that with a small loss of performance compared to the original CDPthread, which uses a static method for allocating the execution engines to a process, the modified CDPthread with load balancing outperforms the modified CDPthread without load balancing by about 25 to 45 percent in terms of the computation time. Moreover, the modified CDPthread can now handle as many threads as necessary.

Chapter 1 Introduction 1
1.1 DSM system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Organization of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Chapter 2 Related Work 5
2.1 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Memory Consistency . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Thread Migration and Load Balancing . . . . . . . . . . . . . . . . . 6
2.2 Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Chapter 3 Distributed Shared Memory and Load Balancing 8
3.1 The Architecture of CDPthread . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.1.1 Control Server . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.1.2 Execution Engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.3 Loader . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.1.4 The Flow of CDPthread . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Waker and Load Balancing for CDPthread . . . . . . . . . . . . . . . . . . . 12
3.2.1 Restriction on the Number of Execution Engines . . . . . . . . . . . 12
3.2.2 The Flow of CDPthread with Waker . . . . . . . . . . . . . . . . . . 16
3.2.3 Implementation of Load Balancing . . . . . . . . . . . . . . . . . . 18
3.2.4 The Flow of Load Balancing for CDPthread . . . . . . . . . . . . . . 21
Chapter 4 Simulation Results 25
4.1 Environment Setting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.2 Restriction of Original CDPthread . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
4.3.1 Results of Executing 1000 Threads . . . . . . . . . . . . . . . . . . 26
4.3.2 Results of Accessing Shared Data . . . . . . . . . . . . . . . . . . . 27
4.3.3 Results of Matrix Multiplication . . . . . . . . . . . . . . . . . . . . 28
4.3.4 Results of Load Balancing Test . . . . . . . . . . . . . . . . . . . . 30
Chapter 5 Conclusion and Future Work 33
5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
Appendix A Details of Experimental Results 35
Bibliography 45

[1] C. Amza, A. Cox, S. Dwarkadas, P. Keleher, H. Lu, R. Rajamony, W. Yu, and W. Zwaenepoel, “Treadmarks: shared memory computing on networks of workstations,”
Computer, vol. 29, no. 2, pp. 18–28, 1996.
[2] L. Iftode and J. Singh, “Shared virtual memory: progress and challenges,” vol. 87, no. 3, pp. 498–507, 1999.
[3] J. Protic, M. Tomasevic, and V. Milutinovic, “Distributed shared memory: concepts and systems,” IEEE Parallel & Distributed Technology: Systems & Applications, vol. 4,
no. 2, pp. 63–71, 1996.
[4] K. Thitikamol and P. Keleher, “Thread migration and communication minimization in dsm systems,” vol. 87, no. 3, pp. 487–497, 1999.
[5] G. Antoniu and L. Bouge, “Dsm-pm2: a portable implementation platform for multithreaded dsm consistency protocols,” in Proc. 15th International Parallel and Distributed Processing Symposium, pp. 1080–1080, 23–27 April 2001.
[6] T.-Y. Liang, J.-C. Ueng, C.-K. Shieh, D.-Y. Chuang, and J.-Q. Lee, “Distinguishing sharing types to minimize communication in software distributed shared memory systems,” Journal of Systems and Software, vol. 55, pp. 73–85, December 2000.
[7] A.-C. Lai, C.-K. Shieh, and Y.-T. Kok, “Load balancing in distributed shared memory systems,” in Proc. IEEE International Performance, Computing, and Communications Conference IPCCC 1997cdis, pp. 152–158, 1997.
[8] Y. Zhuang, C.-K. Shieh, T.-Y. Liang, J.-Q. Lee, and L.-M. Tseng, “A group-based load balance scheme for software distributed shared memory systems,” in Proc. First IEEE/ACM International Symposium on Cluster Computing and the Grid, pp. 371–378, 2001.
[9] G.-F. Tseng, “CDPthread: A POSIX-Thread Based Distributed Computing Environment,” Master’s thesis, National Sun Yat-sen University, 2009.
[10] K. Li and P. Hudak, “Memory coherence in shared virtual memory systems,” ACM Transactions on Computer Systems (TOCS), vol. 7, pp. 321–359, November 1989.
[11] J. B. Carter, “Design of the munin distributed shared memory system,” Journal of Parallel and Distributed Computing, vol. 29, pp. 219–227, September 1995.
[12] E. Speight and J. Bennett, “Brazos: A third generation dsm system,” in Proc. of the USENIX Windows/NT Workshop, pp. 95––106, August 1997.
[13] A. S. A. Itzkovitz and L. Shalev, “Thread migration and its application in distributed shared memory systems,” Journal of Systems and Software, vol. 42, pp. 71–87, July
1998.
[14] P. Keleher, “The relative importance of concurrent writers and weak consistency models,” in Proc. 16th International Conference on Distributed Computing Systems, pp. 91–98, 27–30 May 1996.
[15] K. Thitikamol and P. Keleher, “Thread migration and load balancing in non-dedicated environments,” in Proc. 14th International Parallel and Distributed Processing Symposium IPDPS 2000, pp. 583–588, 2000.
[16] L. S. A Schuster, “Using remote access histories for thread scheduling in distributed shared memory systems,” Proceedings of the 12th International Symposium on Distributed Computer, pp. 347–362, 1998.
[17] J.-C. Ueng, C.-K. Shieh, W.-H. Liu, and C.-C. Lin, “Proteus: a runtime reconfigurable distributed shared memory system,” in Proc. Eighth International Symposium on High Performance Distributed Computing, pp. 347–348, 1999.
[18] W. Shi and Z. Tang, “Dynamic computation scheduling for load balancing in homebased software dsms,” in Proc. Fourth InternationalSymposium on Parallel Architectures, Algorithms, and Networks (I-SPAN ’99), pp. 248–253, 1999.