隨著雲端應用(Cloud Application)的興起，眾多企業，如知名的如Google、 Facebook等公司，紛紛投入以軟體定義網路(Software Defined Networking, SDN)為基礎資料中心(Data Center)的研究，以SDN控制器(controller)來管理資料中心裡的網路設備，以增加資料中心處理大量資料的靈活性以及處理能力。此外，由於資料流的流量模式(Traffic Pattern)逐漸由客戶伺服器間(Client-Server)的南北向流量(North-South Traffic)轉變為大量伺服器間(Server-Server)的東西向流量(East-West Traffic)，網路拓撲(Network Topology)架構有逐漸扁平化的趨勢。常見的扁平化網路拓撲架構有胖樹(Fat tree)及葉脊(Leaf spine)架構等。
在本論文中，我們根據建構方式、擴展性及階層數等性質來比較SDN網路中常見的網路拓撲架構:胖樹架構、葉脊架構、水母型(Jellyfish)架構、BCube架構、DCell架構以及FiConn架構。為了探討東西向流量的增加是否影響網路拓撲的效能，本文從三層式架構與兩層式架構各挑選一個代表，分別為Google資料中心使用的胖樹架構以及當今最大社群網站Facebook使用的二層葉脊架構，以SDN模擬器Mininet模擬出core switch 為4的胖樹架構以及spine switch為4且leaf switch為16的葉脊架構，透過網路效能測試工具D-ITG(Distributed Internet Traffic Generator)在拓撲中產生東西向的流量並統計吞吐量(throughput)、網路延遲(delay)以及網路時間抖動(jitter)等指標來進行比較，最後我們發現針對東西向的流量模式，兩層式葉脊架構具有較好的表現。

With the development of cloud applications, well-known companies, such as Google and Facebook, have focused on the data center based on SDN (Software Defined Networking) to manage their network devices by SDN controllers for processing large amounts of data. In addition, traffic in data center have gradually changed from the client-server (i.e., north-south) traffic pattern to the server-server (i.e., east-west) traffic pattern, which calls for flatten network topologies, such as the Google Fat Tree and the Facebook Leaf Spine.
In this study, we compare several common network topologies: Fat Tree, Leaf Spine, Jellyfish, BCube, DCell and FiConn, in the aspects of the construction approach, the number of levels, and extensibility. We also do simulation experiments for two well-known topologies, the Fat Tree and the Leaf Spine, for comparing their performance in terms of throughput, latency, and jitter. The Mininet simulator is used to virtually construct the Fat Tree topology with 4 core switches and the Leave Spine topology with 4 spine switches and 16 leaf switches. Furthermore, the tool D-ITG (Distributed Internet Traffic Generator) is used to generate traffic of the constant-bit-rate pattern and the Poisson distribution pattern for simulations. The simulation results show that the Leaf Spine topology is superior to the Fat Tree topology.

[1]Open Networking Foundation, “OpenFlow Switch Specification version 1.5.0,” December 19,2014.
[2]B. Nunes, M.Mendonca, X. Nguyen, K. Obraczk, and T. Turletti, “A survey of software-defined networking: Past, present, and future of programmable networks,” to appear in IEEE Communications Surveys &Tutorials, 2014..
[3] S. Jain, et. al., “B4, Experience with a Globally-Deploy-Deployed Software Defined WAN,” CAN SIGCOMM conference,2013.
[4]N. McKeown, T Anderson, H. Balarishnan, G. Parulkar, L. Peterson, J. Rexford, S. Shenker, and J. Turner, “OpenFlow: Enabling Innovation in Campus Networks,” in ACM SIGCOMM Computer Communication, 2008.
[5] IETF ForCES charter. “ForCES Charter,” 2010. https://datatracker.ietf.org/wg/forces/charter/
[6]A. Andreyev, “Introducing data center fabric, the next-generation
Facebook data center network,” 2014. [Online]. Available:
https://code.facebook.com/posts/360346274145943
[7]M. F. Bari, R. Boutaba, R. Esteves, L. Z. Granville, M. Podlesny, M. G.
Rabbani, Q. Zhang and M. F. Zhani, “Data Center Network
Virtualization: A Survey,”IEEE Communications Surveys & Tutorials,
vol. 15, no. 2, pp. 909-928, 2013.
[8]International Data Corporation, “The New Need for Speed in the
Datacenter,” 2015.
[9]S. Shah and N. Guenov, “Multicore Processing: Virtualization and Data
Center,”Freescale, 2012.
[10]B. Bland, “Titan - Early experience with the Titan system at Oak Ridge
National Laboratory,”in 2012 SC Companion: High Performance
Computing, Networking, Storage and Analysis, 2012.
[11]M. Yokokawa, F. Shoji, A. Uno, M. Kurokawa and T. Watanabe, “The K
computer: Japanese next-generation supercomputer development
project,” in 2011 International Symposium on Low Power Electronics
and Design (ISLPED), 2011.
[12]AL-FARES M, LOUKISSAS A, VAHDAT A. “A Scalable, Commodity Data Center Network Architecture,” ACM SIGCOMM Computer Communication Review, 2008, 38(4): 63-74.
[13] GREENBERG A , HAMILTON J R, JAIN N, et al. “VL2: A scalable and flexible data center network,” ACM SIGCOMM Computer Communication Review, 2009, 39(4): 51-62.
[14]C. Guo, H. Wu, K. Tan, L. Shi, Y. Zhang, and S. Lu, “Dcell: a
scalable and fault-tolerant network structure for data centers,” ACM
SIGCOMM Computer Communication Review, vol. 38, no. 4, pp. 75–
86, 2008.
[15]C. Guo, G. Lu, D. Li, H. Wu, X. Zhang, Y. Shi, C. Tian, Y. Zhang,
and S. Lu, “Bcube: a high performance, server-centric network architecture
for modular data centers,” ACM SIGCOMM Computer
Communication Review, vol. 39, no. 4, pp. 63–74, 2009.
[16]R. L. S. de Oliveira, A. A. Shinoda, C. M. Schweitzer, and L. Rodrigues
Prete, “Using mininet for emulation and prototyping softwaredefined
networks,” in Communications and Computing (COLCOM),
2014 IEEE Colombian Conference on. IEEE, 2014, pp. 1–6.
[17]RATNASAMY, S., AND MCCANNE, S. “Inference of Multicast Routing Threes and Bottleneck Bandwidths using End-to-end Measurements,” Proceedings of INFOCOM ‘99 (March 1999).
[18] M. Alizadeh and T. Edsall, “On the data path performance of leaf-spine datacenter fabrics,” in HotInterconnects 2013, pp. 71–74
[19]A. Singla, C.-Y. Hong, L. Popa, and P. B. Godfrey. Jellyfish: Networking
data centers randomly. In NSDI ’12. USENIX, 2012.
[20]Li D，Guo CX，Wu HT，Tan K，Zhang YG，Lu SW “FiConn：Using backup port for server interconnection in data centers,”In：Proc．
ofthe INFOCOM 2009．Rio de Janeiro：IEEE，2009．2276—2285．
[21]Cnblog, “fat-tree network,” from “http://www.cnblogs.com/lfzark/p/4269497.html”
[22] Jiang Dongyi, “Data Center Evolution and Security Challenges
,”from “http://slidesplayer.com/slide/11437822/”
[23]Cnii, fromhttp://www.cnii.com.cn/technology/2015-04/03/content_1556883.htm
[24]政府資料開放平台:教育部學術網路(TANet)各縣市網路流量即時狀況。2015年11月26日，取自https://data.gov.tw/dataset/24374。
[25]Streiffer C, Chen H, Benson T, et al. “DeepConf: Automating Data Center NetworkTopologies Management with Machine Learning,” arXiv:1712.03890v1[cs.NI]