網路服務和應用程序的意外和動態負載浪湧往往會淹沒服務提供商和降低計算資源的利用率，雲端設備概念透過虛擬化跟彈性的概念減輕了這個問題。但是為了確保服務的品質和資源的使用率，我們仍然需要更有智能的自動化控制系統來管理。
在此篇論文裡，我們發展了一個自動化的機制給雜湊式的儲存系統，並且提出了縮放策略跟負載平衡演算法來達到SLA 以及降低系統熱點。我們在內部的叢集電腦裡進行實驗，以評估我們的實作跟證明我們解決方案的有效性跟性能。

The unexpected and dynamic load surges from web-based services and applications often overwhelm service providers and lower computation resource utilization. Cloud infrastructure mitigates the problem by developing technologies based on the concept of virtualization and elasticity. But to insure the service quality and resource utilization, we still require a more intelligent automation control system to manage the system. In this thesis, we develop an automated mechanism for hash-based storage, and propose scaling policies and load-balancing algorithms to satisfy the SLA agreement and minimize system hot-spot. We conduct experiments to evaluate our implementation on a in-house cluster, and demonstrate the effectiveness and performance of our solutions.

1 Introduction 4
2 Control System 7
2.1 Hash-based Storage 7
2.2 Controller 8
3 Load Balancing Algorithm 11
3.1 Problem formulation 11
3.2 Workload-Based Strategy 12
3.3 Proportional-Based Strategy 13
4 Virtual Node Solution 16
4.1 Scaling of Virtual Node 17
4.2 Load-Balance of Virtual Node 18
5 Implementation 19
6 Experiments 21
6.1 Setup 21
6.2 Evaluation Results 23
6.2.1 Physical node solution 23
6.2.2 Virtual node solution 30
7 Related Work 37
8 Conclusion 39

[1] S. Adler. The Slashdot eect: an analysis of three Internet publications. 38:2,
1999.
[2] Animoto's Facebook scale-up. http://blog.rightcale.com/2008/04/23/animoto-
facebool-scale-up.
[3] AWS:, auto scaling. available: http://aws.amazon.com/autoscaling/.
[4] J. C. Y. Chou, T.-Y. Huang, K.-L. Huang, and T.-Y. Chen. Scallop: A scalable
and load-balanced peer-to-peer lookup protocol. IEEE Transactions on Parallel
and Distribued Systems, 17(5):419{433, May 2006.
[5] collectd: http://collectd.org/.
[6] J. Dean and S. Ghemawat. Mapreduce: simplied data processing on large
clusters. Communications of the ACM, 51(1):107{113, Jan. 2008.
[7] G. DeCandia, D. Hastorun, M. Jampani, G. Kakulapati, A. Lakshman,
A. Pilchin, S. Sivasubramanian, P. Vosshall, and W. Vogels. Dynamo: ama-
zon's highly available key-value store. In Symposium on Operating Systems
Principles, pages 205{220, Oct. 2007.
[8] Faban: http://faban.org.
[9] S. Ghemawat, H. Gobio, and S.-T. Leung. The google le system. In Sympo-
sium on Operating Systems Principles, pages 29{43, Oct. 2003.
[10] Glusterfs: http://www.gluster.org.
[11] A. Kamra. Yaksha: A self-tuning controller for managing the performance of
3-tiered web sites. pages 47{56, 2004.
[12] D. R. Karger and M. Ruhl. Simple ecient load balancing algorithms for
peer-to-peer systems. In Proceedings of ACM symposium on Parallelism in
Algorithms and Architectures, pages 36{43, 2004.
[13] A. Kimball, S. Michels-Slettvet, and C. Bisciglia. Cluster computing for web-
scale data processing. In Proceedings of the 39th SIGCSE technical symposium
on Computer science education, pages 116{120, 2008.
[14] Kernel-based Virtual Machine (KVM). http://www.linux-
kvm.org/page/Main Page.
[15] H. C. Lim, S. Babu, and J. S. Chase. Automated control for elastic storage. In
IEEE International Conference on Autonomic Computing, pages 1{10, 2010.
[16] T. J. M. Arlitt. A workload characterization study of the 1998 world cup web
site. 14:30{37, 2000.
[17] C. Maltzahn, E. Molina-Estolano, A. Khurana, A. J. Nelson, S. A. Brandt, and
S. Weil. Ceph as a scalable alternative to the Hadoop Distributed File System.
;login: USENIX Magazine, pages 38{49, Aug. 2010.
[18] P. Padala, K.-Y. Hou, K. G. Shin, X. Zhu, M. Uysal, Z. Wang, S. Singhal, and
A. Merchant. Automated control of multiple virtualized resources. In Proceed-
ings of the 4th ACM European conference on Computer systems, EuroSys '09,
pages 13{26, 2009.
[19] H. Shen and C.-Z. Xu. Locality-aware and churn-resilient load-balancing al-
gorithms in structured peer-to-peer networks. IEEE Trans. Parallel Distrib.
Syst., 18(6):849{862, Jun. 2007.
[20] K. Shvachko, H. Kuang, S. Radia, and R. Chansler. The hadoop distributed
le system. In Symposium on Mass Storage Systems and Technologies, pages
1{10, May 2010.
[21] K. V. SHVACHKO. HDFS scalability: The limits to growth. ;login: USENIX
Magazine, pages 6{16, Apr. 2010.
[22] I. Stoica, R. Morris, D. Liben-Nowell, D. Karger, M. Kaashoek, F. Dabek, and
H. Balakrishnan. Chord: a scalable peer-to-peer lookup protocol for internet
applications. IEEE/ACM ToN, 11(1):17{32, Feb. 2003.
[23] Swift: OpenStack Object Store, http://swift.openstack.org/.
[24] Xen:, http://www.xen.org/.
[25] Y. Zhu and Y. Hu. Ecient, proximity-aware load balancing for dht-based p2p
systems. IEEE Transactions on Parallel and Distribued Systems, 16(4):349{
361, Apr. 2005.