在雲端計算與雲端資源分配的議題中，虛擬化的技術一直是非常重要，它能模擬一台虛擬機器、讓多台虛擬機器放在同一台實體機器上、每台虛擬機都是獨立運作的，不同於實體機器，虛擬機器還能在不關機的情況下搬移到另一台機器上繼續執行，因為這特性的關係，雲端機器管理的方法有了更多的應用空間，其中一個就是動態資源分配。

在這篇論文中，我們提出了蟲洞搬移的技術，讓虛擬機器可以非常快速的在兩台實體機上互相搬移。藉由虛擬機交互的存在不同機器，可以適用於一些應用情況得到好處。第一，我們可以快速的做到容錯移轉(failover)，在實體機器快不能夠使用時，搬移到可用的機器上；第二，在雲端計算上，因為快速搬移及對搬移時對機器效能影響很低，反而可以減少平行處理的程式在資料傳遞的時間；第三，節省電費，讓虛擬機器在比較低電價的時區上運行，一天之中必須要來來回回好幾次。

我們修改了libvirt實作動態搬移方法，在QEMU/KVM 虛擬化平台上完成蟲洞搬移。在第一次搬移時用原本的方式作搬移，建立蟲洞機制，讓之後的搬移可以使用蟲洞。跟原本實作方法不同的地方是，蟲洞搬移只需要傳輸新被改寫過的記憶體，這樣能夠同時減少搬移的時間跟網路頻寬的使用。現在我們在研究蟲洞結合虛擬化的另一項技術 快照，快照是可以把虛擬機的運作當下包括記憶體等所有資源儲存下來。蟲洞的實驗成果，搬移時間平均可以節省4倍，而網路頻寬節省90% 在最好的情況下。

The four abilities of virtual machines (VMs): emulation, consolidation, isolation, and migra-tion, have made virtualization an important technique in cloud computing and resource provision-ing. Among them, the ability of migration that can move a VM among different physical machines without disrupting the running services on it has many applications, such as dynamic resource real-location.

In this thesis, we proposed a technique, called wormhole migration, which allows a VM to be fast migrated between two physical machines back and forth frequently. Such mechanism can be very useful in some applications, three of which were presented here. The first is the fast failover, which switches the running services to a redundant machine when the original machine fails. The second one is the communication cost reduction for the parallel programs running on cloud plat-forms. The wormhole migration can move the frequent communicating VMs together with little overhead. The third one is for saving of power rate by migrating VMs to the zones of cheaper power rate periodically.

We have implemented the wormhole migration on the hypervisor QEMU/KVM (Kernel Vir-tual Machine) by modifying the live migration mechanism of libvirt. A wormhole tunnel between two physical machines is built in the first migration, and used to transmit the VM later. Unlike the ordinary VM live migration, the wormhole migration only needs to transmit the new dirty pages, which saves the migration duration and network bandwidth. We also investigate the wormhole migration with snapshot, which can further save the computational resources for physical machines. Experiments showed that by using wormhole migration, the duration time can be reduced 4x and the network bandwidth usage can be saved by 90% for best case.

摘要 i
Abstract ii
Table of Contents iii
1. Introduction 1
2. Background and Motivations 4
2.1. Existing Migration Schemes 4
2.1.1. Pre-Copy migration 4
2.1.2. Post-Copy migration: 4
2.2. Related work 7
2.3. Applications of Wormhole Migration 7
2.3.1. Fast Failover 7
2.3.2. Communication Cost Reduction for Parallel Programs 9
2.3.3. Saving of Electricity Cost and Energy 10
3. Design and Implementation 11
3.1. Design of Wormhole Migration 11
3.2. Implementation of Wormhole Migration 14
3.2.1. Reuse Original VM 15
3.2.2. Transfer New Dirty Pages 16
3.2.3. Prefetching 18
3.2.4. Migration stealth: 19
3.2.5. Other implementation on Reusing Original VM 21
3.2.6. Other implementation on Transferring New Dirty Pages 21
4. Performance Evaluation 22
4.1. Environment Setup 22
4.2. Performance of Wormhole Migration 22
4.2.1. Total Migration Time and Transfer Pages Size 22
4.2.2. Prefetching and stealth 24
4.2.3. Saving of Electricity Cost 25
5. Conclusion 29
5.1. Future work 30
5.1.1. Wormhole migration with snapshot 30
5.1.2. Migration on Docker 31
6. References 32

[1] Bellard, Fabrice. "QEMU, a Fast and Portable Dynamic Translator." USENIX Annual Tech-nical Conference, FREENIX Track. 2005.
[2] Kivity, Avi, et al. "kvm: the Linux virtual machine monitor." Proceedings of the Linux Sym-posium. Vol. 1. 2007.
[3] C. Clark, K. Fraser, S. Hand, J. Hansen, E. Jul, C. Limpach, I. Pratt, and A. Warfiled, “Live Migration of Virtual Machines”, in Proceedings of the 2nd Conference on Symposium on Networked Systems Design & Implementation, vol. 2, pp. 286, 2005
[4] M. Nelson, B. Lim, and G. Hutchins, “Fast transparent migration for virtual machines”, in Proceedings of the Annual Conference on USENIX Annual Technical Conference, pp. 25, 2005Other migration research
[5] Yin, Feiran, Weidong Liu, and Jiaxing Song. "Live Virtual Machine Migration with Opti-mized Three-Stage Memory Copy." Future Information Technology. Springer Berlin Hei-delberg, 2014. 69-75.
[6] Kejiang Ye, Xiaohong Jiang, Dawei Huang, Jianhai Chen, and Bei Wang, “Live migration of Multiple Virtual Machines with Resource Reservation in Cloud Computing Environ-ments” , in Proceedings of 2011 IEEE 4th International Conference On Cloud Computing, pp. 267-274, 2011
[7] Thakur, Susheel, Arvind Kalia, and Jawahar Thakur. "Performance Analysis of Server Con-solidation Algorithms in Virtualized Cloud Environment with Idle VMs."International Journal 2.3 (2014).
[8] Jeong, Jinkyu, et al. "Analysis of virtual machine live-migration as a method for pow-er-capping." The Journal of Supercomputing 66.3 (2013): 1629-1655.
[9] Tchana, Alain, Laurent Broto, and Daniel Hagimont. "Approaches to cloud computing fault tolerance." Computer, Information and Telecommunication Systems (CITS), 2012 Interna-tional Conference on. IEEE, 2012.
[10] Yin, Feiran, Weidong Liu, and Jiaxing Song. "Live Virtual Machine Migration with Opti-mized Three-Stage Memory Copy." Future Information Technology. Springer Berlin Hei-delberg, 2014. 69-75.
[11] Patel, Pankesh, Ajith H. Ranabahu, and Amit P. Sheth. "Service level agreement in cloud computing." (2009).
[12] Elnozahy, Elmootazbellah N., and James S. Plank. "Checkpointing for peta-scale systems: A look into the future of practical rollback-recovery." Dependable and Secure Computing, IEEE Transactions on 1.2 (2004): 97-108.
[13] Bila, Nilton, et al. "Jettison: efficient idle desktop consolidation with partial VM migra-tion." Proceedings of the 7th ACM european conference on Computer Systems. ACM, 2012.
[14] Kim, Jihun, et al. "Guide-copy: fast and silent migration of virtual machine for datacen-ters." Proceedings of SC13: International Conference for High Performance Computing, Networking, Storage and Analysis. ACM, 2013.
[15] Jo, Changyeon, et al. "Efficient live migration of virtual machines using shared stor-age." ACM SIGPLAN Notices. Vol. 48. No. 7. ACM, 2013.
[16] Liu, Haikun, et al. "Live migration of virtual machine based on full system trace and re-play." Proceedings of the 18th ACM international symposium on High performance distrib-uted computing. ACM, 2009.
[17] Park, Eunbyung, Bernhard Egger, and Jaejin Lee. "Fast and space-efficient virtual machine checkpointing." ACM SIGPLAN Notices. Vol. 46. No. 7. ACM, 2011.
[18] Jin, Hai, et al. "Live virtual machine migration with adaptive, memory compression."Cluster Computing and Workshops, 2009. CLUSTER'09. IEEE International Conference on. IEEE, 2009.
[19] Deshpande, Umesh, Xiaoshuang Wang, and Kartik Gopalan. "Live gang migration of vir-tual machines." Proceedings of the 20th international symposium on High performance dis-tributed computing. ACM, 2011.
[20] Yuanyuan Zhou, Peter M. Chen, and Kai Li. 1999. Fast cluster failover using virtual memory-mapped communication. In Proceedings of the 13th international conference on Supercomputing (ICS '99). ACM, New York, NY, USA, 373-382.
[21] Alex C. Snoeren, David G. Andersen, and Hari Balakrishnan, Fine-Grained Failover Using Connection Migration, USITS 2001, pp 221-232.
[22] Vogels, W. Dumitriu, D. ; Birman, K. ; Gamache, R. ; Massa, M. ; Short, R. ; Vert, J. ; Bar-rera, J. ; Gray, J., The design and architecture of the Microsoft Cluster Service-a practical approach to high-availability and scalability, Fault-Tolerant Computing, 1998. Digest of Papers. Twenty-Eighth Annual International Symposium on, June 1998, pp422-431
[23] I-Hsin Chung , Che-Rung Lee ; Jiazheng Zhou ; Yeh-Ching Chung, Scalable Communica-tion-Aware Task Mapping Algorithms for Interconnected Multicore Systems, High Perfor-mance Computing and Communications (HPCC), 2011 IEEE 13th International Confer-ence on, Sep 2011, pp759-764
[24] I-Hsin Chung, Che-Rung Lee, Jiazheng Zhou, Yeh-Ching Chung, Hierarchical Mapping for HPC Applications. Parallel Processing Letters 21(3): 279-299 (2011)
[25] Kien Le, Ricardo Bianchini, Jingru Zhang, Yogesh Jaluria, Jiandong Meng, and Thu D. Nguyen. 2011. Reducing electricity cost through virtual machine placement in high perfor-mance computing clouds. In Proceedings of 2011 International Conference for High Per-formance Computing, Networking, Storage and Analysis (SC '11). ACM, New York, NY, USA, , Article 22 , 12 pages.
[26] Iperf (https://iperf.fr/)
[27] Murphy, Richard C., et al. "Introducing the graph 500." Cray User’s Group (CUG) (2010).
[28] "Definition of: memory footprint". PC Magazine. August 2012.
[29] Martyn Honeyford, "Reduce your Linux memory footprint". IBM developerWorks. August 2012.
[30] Libvirt (http://libvirt.org/index.html )
[31] Docker (https://www.docker.com/)