通常來說，在雲端上磁碟區的效能與大小相關。磁碟區的大小決定磁碟區的基準效能。然而雲端服務提供者如亞馬遜，微軟等為了應對不同時間資料存取頻率的不平等，提出了 I/O 額度的概念。I/O 額度代表當需求超出基準效能時，磁碟區可用來爆發大量 I/O 的可用頻寬。當需求低於磁碟機的基準效能時，I/O 額度就會累積。磁碟區愈大，基準效能層愈高，累積 I/O 額度的速度愈快。當累積的 I/O 額度愈多，需要更多效能時，磁碟區能爆量超過基準效能的時間愈長，表現也愈佳。

為了保持系統運行在有 I/O 額度的狀態，之前有其他人提出了可以透過複製資料的方法，並比較成本和收益證明可行性。然而純粹的資料複製會導致程式中斷，並且消耗掉 I/O 額度，為解決這兩個問題，我們提出了動態遷移，藉由快照和記憶體暫存，以及延遲暫存轉存，讓中斷時間最小化，並保持中斷其間的頻寬。我們將動態遷移的方法應用在天氣預測系統 WRF 上，和沒有 I/O 額度的方法相比，加快了運行時間 332 %

Burst storage, offered by cloud providers such as Amazon AWS and Microsoft Azure, could boost its bandwidth for a short period of time. The time length is controlled by burst credits. Upon running out of burst credits, the performance will drop to baseline.

To keep burst credits available, previous work have used direct data copy, and prove that this method is cost-efficient. However, direct data copy will introduce application interrupt and consume credits. To tackle above problems, this paper introduce live migration. Live migration replace the old volume with the new one without ceasing the running application, thus mitigate downtime. We achieve live migration by snapshot, memory buffer and delay dumping. Applying this system on real world application WRF improve its running time by 332%.

摘要 . . . . . . . . . . .. . . . . . . . . . .1
Abstract . . . . . . . . . . .. . . . . . . . .2
Acknowledgements . . . . . . . . . . . . . . . .0
1 Introduction . . . . . . . . . . .. .. . . . .4
2 Motivation. . . . . . . . . . .. . . . . . . . 7
3 System Design . . . . . .. . . . . . . . .. . 10
3.1 Architecture . . . . . . . . . . . . . . . 10
3.2 Burst Manager . . . . . . . . . . . . . . . 12
3.3 Interface . . . . .. . . . . . . . . . . . . 13
3.4 Dumping . . . . . . . . . . . . . . . . . . 15
4 Experiment . . . . . .. . . . . . . .. . . .. . 18
4.1 Performance Comparison with Direct Copy . . . 18
4.2 Workload’s effect . . .. . .. . .. . .. . . . 20
4.3 Parameters’ effect . . . . . . . . . . . . . . 21
4.4 Real-world application . . . . . . . . . . . .. 23
5 Related Work . . . . . .. . . . . . . . .. . .. . 25
6 Conclusion. . . . . .. . . . . . .. . . . . . . 27
References. . . . . .. . . . . . . .. . . . . ... . 28

[1] H. Park, G. R. Ganger, and G. Amvrosiadis, “More iops for less: Exploiting burstable
storageinpublicclouds,”in12thUSENIXWorkshoponHotTopicsinCloudComputing
(HotCloud 20), USENIX Association, 2020.
[2] I. Amazon Web Services, “Reactive systems on aws,” tech. rep., 2021.
[3] B.Calder, J.Wang, A.Ogus, N.Nilakantan, A.Skjolsvold, S.McKelvie, Y.Xu, S.Sri-
vastav, J. Wu, H. Simitci, J. Haridas, C. Uddaraju, H. Khatri, A. Edwards, V. Be-
dekar, S. Mainali, R. Abbasi, A. Agarwal, M. F. u. Haq, M. I. u. Haq, D. Bhardwaj,
S. Dayanand, A. Adusumilli, M. McNett, S. Sankaran, K. Manivannan, and L. Rigas,
“Windows azure storage: A highly available cloud storage service with strong con-
sistency,” in Proceedings of the Twenty-Third ACM Symposium on Operating Systems
Principles, SOSP ’11, (New York, NY, USA), p. 143–157, Association for Computing
Machinery, 2011.
[4] J. Spillner, J. Müller, and A. Schill, “Creating optimal cloud storage systems,” Future
Generation Computer Systems, vol. 29, no. 4, pp. 1062–1072, 2013.
[5] I.Drago,E.Bocchi,M.Mellia,H.Slatman,andA.Pras,“Benchmarkingpersonalcloud
storage,”inProceedings of the 2013 Conference on Internet Measurement Conference,
IMC ’13, (New York, NY, USA), p. 205–212, Association for Computing Machinery,
2013.
[6] R. Nishtala, H. Fugal, S. Grimm, M. Kwiatkowski, H. Lee, H. C. Li, R. McElroy,
M. Paleczny, D. Peek, P. Saab, D. Stafford, T. Tung, and V. Venkataramani, “Scaling
memcache at facebook,” in 10th USENIX Symposium on Networked Systems Design
and Implementation (NSDI 13), (Lombard, IL), pp. 385–398, USENIX Association,
Apr. 2013.
[7] M. Paksula, “Persisting objects in redis key-value database,” University of Helsinki,
Department of Computer Science, p. 27, 2010.
[8] J.Varia,S.Mathew,et al.,“Overviewofamazonwebservices,”Amazon Web Services,
vol. 105, 2014.
[9] C.Clark,K.Fraser,S.Hand,J.G.Hansen,E.Jul,C.Limpach,I.Pratt,andA.Warfield,
“Live migration of virtual machines,” in Proceedings of the 2nd conference on Sympo-
sium on Networked Systems Design & Implementation-Volume 2, pp. 273–286, 2005.
[10] W. Voorsluys, J. Broberg, S. Venugopal, and R. Buyya, “Cost of virtual machine live
migration in clouds: A performance evaluation,” in IEEE International Conference on
Cloud Computing, pp. 254–265, Springer, 2009.
[11] M. R. Hines, U. Deshpande, and K. Gopalan, “Post-copy live migration of virtual ma-
chines,” ACM SIGOPS operating systems review, vol. 43, no. 3, pp. 14–26, 2009.
[12] Y. Wu and M. Zhao, “Performance modeling of virtual machine live migration,” in
2011 IEEE 4th International Conference on Cloud Computing, pp. 492–499, 2011.
[13] D. Breitgand, G. Kutiel, and D. Raz, “{Cost-Aware}live migration of services in the
cloud,” in Workshop on Hot Topics in Management of Internet, Cloud, and Enterprise
Networks and Services (Hot-ICE 11), 2011.
[14] K. Haselhorst, M. Schmidt, R. Schwarzkopf, N. Fallenbeck, and B. Freisleben, “Effi-
cient storage synchronization for live migration in cloud infrastructures,” in 2011 19th
International Euromicro Conference on Parallel, Distributed and Network-Based Pro-
cessing, pp. 511–518, IEEE, 2011.
[15] I. Amazon Web Services, “Cloud bursting eda with fsx for netapp ontap,” tech. rep.,
2021.
[16] I. Amazon Web Services, “Storage best practices for data and analytics applications,”
tech. rep., 2021.
[17] A. S. Deepak Mohan, “Azure disk storage idc white paper,” tech. rep., 2020.
[18] N.Liu,J.Cope,P.Carns,C.Carothers,etal.,“Ontheroleofburstbuffersinleadership-
class storage systems,” in IEEE 28th Symposium on Mass Storage Systems and Tech-
nologies (MSST), IEEE, 2012.
[19] F. Tessier, M. Martinasso, M. Chesi, M. Klein, and M. Gila, “Dynamic provisioning
of storage resources: A case study with burst buffers,” in IEEE International Parallel
and Distributed Processing Symposium Workshops (IPDPSW), IEEE, 2020.
[20] S. Ifrah, Backing Up and Restoring Your Containers and Hosts on Amazon AWS,
pp. 319–337. Berkeley, CA: Apress, 2019.
[21] C. Li, P. Shilane, F. Douglis, H. Shim, S. Smaldone, and G. Wallace, “Nitro: A
Capacity-Optimized SSD cache for primary storage,” in 2014 USENIX Annual Tech-
nical Conference (USENIX ATC 14), (Philadelphia, PA), pp. 501–512, USENIX Asso-
ciation, June 2014.
[22] F. Li, Y. Lu, Z. Wu, and J. Shu, “Ascache: An approximate ssd cache for error-tolerant
applications,” inProceedings of the 56th Annual Design Automation Conference 2019,
DAC ’19, (New York, NY, USA), Association for Computing Machinery, 2019.
[23] A. Strunk, “Costs of virtual machine live migration: A survey,” in 2012 IEEE Eighth
World Congress on Services, pp. 323–329, IEEE, 2012.
[24] J.Li,H.Chen,Y.Chen,Z.Lin,B.Vucetic,andL.Hanzo,“Pricingandresourcealloca-
tion via game theory for a small-cell video caching system,” IEEE Journal on Selected
Areas in Communications, vol. 34, no. 8, pp. 2115–2129, 2016.
[25] J. Liu, Y. Chai, X. Qin, and Y. Xiao, “Plc-cache: Endurable ssd cache for
deduplication-based primary storage,” in 2014 30th Symposium on Mass Storage Sys-
tems and Technologies (MSST), pp. 1–12, IEEE, 2014