越來越多的企業或是個人使用者將他們的資料儲存到雲端儲存
器上，使得雲端儲存的效能以及花費更加被重視，雲端供應商在世界
各地建立資料中心來提供全球化的儲存設備，除此之外，若為單一副
本的資料儲存，因為其他的資料中心並不會擁有該檔案，造成來自世
界各地的存取都必須前往唯一的資料中心存取檔案，使得跨國甚至跨
洲的存取有很大的傳輸負擔。因此，勢必需要一個整合平台與多副本
儲存方式並結合這些位於世界各地的儲存器來解決全球性存取的問
題，同時，這些副本該分別存放在哪些地方才能更有效地增強效能與
減少花費也是考量的重點。在這篇論文中，我們開發了一聯邦式雲端
儲存系統，可結合現今多數的雲端儲存器，同時提出一動態副本管理
演算法，根據不同的使用狀況與行為，動態地調整副本存放的位置，
達到提高效能與降低花費的目的。

As more and more enterprises and personal users move their data to cloud storage, performance and cost consumption are getting more and more important for user. Cloud service provides global infrastructures by building data centers at different locations. In addition, single replica can not handle user’s global access behaviors
such as changing access pattern or sharing data across continents. A single platform which is capable to combine all of the cloud storage in different locations and
benefit from global infrastructure to solve the global access problem is necessary. Furthermore, where should the replica be stored to make the system has higher
performance and lower cost consumption should also be considered. In this paper, we develop a federated cloud storage system that combines multiple cloud storage
into a single platform, and propose a dynamic replica placement algorithm that can decide the best replica placement according to different workloads and user behaviors.

1 Introduction 5
2 System Design and Implementation 8
2.1 Basic Replication Unit 8
2.2 Instance Overview 9
2.2.1 Metadata Server 10
2.2.2 Access Gateway 11
2.2.3 Storage Node 12
2.3 I/O Operations Sequence Diagram 13
3 Data Management Problem Description 15
3.1 Performance Model 15
3.2 Cost Model 17
3.3 Problem Objective 18
3.4 NP-Complete Problem 18
4 Data Migration Algorithm 20
4.1 Greedy Algorithm 20
4.2 Approximation Policies 21
4.2.1 Main User Filter 21
4.2.2 Read Combination Filter 22
4.2.3 Write Combination Filter 23
5 Simulation 26
5.1 Setup 26
5.1.1 System Nodes 26
5.1.2 Workload 26
5.2 Algorithm Comparison 27
5.2.1 Read Intensive Folder Comparison 28
5.2.2 Write Intensive Folder Comparison 29
5.3 Performance Breakdown by Time 29
5.4 Cost Limited Effective 31
5.5 Filter Level 33
6 AWS Experiment 35
6.1 Setup 35
6.1.1 Data Storage and Access Gateways 35
6.1.2 Metadata Server 36
6.1.3 Workload 36
6.2 Read Intensive Workload 36
6.3 Write Intensive Workload 37
7 Related work 39
8 Conclusion 41
A AWS Benchmark 43
A.1 EC2 to S3 Access Distance 43
A.2 S3 to S3 Transfer Distance 46
A.3 Summary 48

[1] Amazon EC2 http://aws.amazon.com/ec2.
[2] Amazon S3 http://aws.amazon.com/s3.
[3] Google Drive https://www.google.com/drive.
[4] Microsoft Azure http://azure.microsoft.com.
[5] Rackspace http://www.rackspace.com.
[6] H. Abu-Libdeh, L. Princehouse, and H. Weatherspoon. Racs: a case for cloud
storage diversity. In Proceedings of the 1st ACM symposium on Cloud computing,
pages 229–240. ACM, 2010.
[7] M. S. Ardekani and D. B. Terry. A self-configurable geo-replicated cloud storage
system. In Symp. on Op. Sys. Design and Implementation (OSDI), pages 367–
381, 2014.
[8] B. Calder, J. Wang, A. Ogus, N. Nilakantan, A. Skjolsvold, S. McKelvie, Y. Xu,
S. Srivastav, J. Wu, H. Simitci, et al. Windows azure storage: a highly available
cloud storage service with strong consistency. In Proceedings of the TwentyThird
ACM Symposium on Operating Systems Principles, pages 143–157. ACM,
2011.
[9] B. Chun, D. Culler, T. Roscoe, A. Bavier, L. Peterson, M. Wawrzoniak, and
M. Bowman. Planetlab: an overlay testbed for broad-coverage services. ACM
SIGCOMM Computer Communication Review, 33(3):3–12, 2003.
[10] W.-C. Chung, P. C. Shih, K. C. Lai, K. C. Li, C. R. Lee, J.-H. Chou, C. H. Hsu,
and Y. C. Chung. Taiwan unicloud: A cloud testbed with collaborative cloud services. In Cloud Engineering (IC2E), 2014 IEEE International Conference
on, pages 107–116. IEEE, 2014.
[11] M. Karlsson, C. Karamanolis, and M. Mahalingam. A framework for evaluating
replica placement algorithms. Technical report, Citeseer, 2002.
[12] S. U. Khan and I. Ahmad. Comparison and analysis of ten static heuristicsbased
internet data replication techniques. Journal of Parallel and Distributed
Computing, 68(2):113–136, 2008.
[13] L. Richardson and S. Ruby. RESTful web services. ” O’Reilly Media, Inc.”,
2008.
[14] D. Slamanig and C. Hanser. On cloud storage and the cloud of clouds approach.
In Internet Technology And Secured Transactions, 2012 International
Conference for, pages 649–655. IEEE, 2012.
[15] V. J. Sosa-Sosa and E. M. Hernandez-Ramirez. A file storage service on a
cloud computing environment for digital libraries. Information Technology and
Libraries, 31(4):34–45, 2012.
[16] Z. Xu and L. Bhuyan. Qos-aware object replica placement in cdn networks.
2005.
[17] Z. Ye, S. Li, and X. Zhou. Gcplace: Geo-cloud based correlation aware data
replica placement. In Proceedings of the 28th Annual ACM Symposium on
Applied Computing, pages 371–376. ACM, 2013.