臺灣博碩士論文加值系統

軟體定義網路（Software Defined Networking）是網路產業近年來最主要的趨勢。於生產網路環境上部署實體 SDN 網路除了十分昂貴也耗時而且也不實際，所以 SDN 研究人員或系統整合建置商在實際部署 SDN 網路前，非常需要利用 SDN 網路仿真模擬器來進行測試與評估。此外，無線區網（WLAN）已被廣泛地部署在大多數的企業或校園網路，因此網路仿真模擬器必須還要包含無線區網的模擬功能。在眾多 SDN 仿真模擬器當中，又以 OpenNet 是最適合或最符合實際情境的 SDN 仿真模擬器，因為OpenNet可以用來模擬包含 WLAN 及 SDN 網路的實體網路。
在實際部署網路之前可以將生產網路內的 WLAN 及 SDN 網路透過 OpenNet 模擬平台進行模擬。然而，一般而言，校園網路的規模非常龐大，因此要利用單一電腦的計算資源模擬校園等級的網路環境，是十分不可行的。
在先前文獻中，有提到兩個分散式的 SDN 仿真模擬器專案: Mininet Cluster Edition (Mininet CE) 和 MaxiNet。針對跨實體主機之虛擬節點間的通訊，Mininet CE 是採用 SSH 隧道的方式進行傳輸，而 MaxiNet 則是使用 GRE Tunnel 相互傳輸資料。此外，針對虛擬節點在實體主機群 (Cluster) 的放置 (Placement) 方式，Mininet CE 採用的方式是依據機器數量平均放置虛擬節點或者是使用亂數決定放置虛擬節點的主機。MaxiNet也是將虛擬主機依據實體主機 (Worker) 的數量平均分散在實體主機 (Worker)，但分群 (Partition) 時則已將網路流量盡量侷限在同一實體主機中，以減少跨實體主機的通訊，但 MaxiNet 的分佈 (Partiton) 方式並未考量實體主機和網路本身的效能。而且，無論是 Mininet CE 或 MaxiNet 皆是基於 Mininet 的仿真模擬器，兩者皆不支援 WLAN 仿真模擬。
在本論文中，我們提出了分散式 OpenNet 仿真模擬器的設計及實作方式。不同於 Mininet CE 及 MaxiNet，分散式 OpenNet 解隅 (Decouple) 了節點分群 (Partition) 演算法與放置 (Placement) 機制，透過這方式，分散式 OpenNet 使用者可以自行設計分群演算法，而不用考慮繁瑣的節點放置流程，此外，分散式 OpenNet 也提供更多跨實體主機的虛擬節點間之通訊隧道的選擇 (例如: GRE、VXLAN、SSH等隧道協定)。最後，分散式 OpenNet 也提供自動化部署的工具，以簡化和加快複雜的環境佈署與安裝過程，方便研究人員使用。

關鍵字：軟體定義網路、仿真模擬器、隧道協定、自動化工具

In recent years, Software Defined Networking (SDN) is becoming one of the major trends in networking industry. Deployment of SDN networks is both expensive and time consuming, or even impractical in production networks. Therefore, SDN network emulation is extremely desired for researchers or system integrators to test or evaluate their designs in laboratory before real deployment of SDN networks. Furthermore, wireless local area networks (WLANs) are widely deployed in most enterprise or campus production networks nowadays. Consequently, network emulation must also include WLANs emulation too. Among the SDN emulators, OpenNet is one of the most promising or realistic SDN network emulators because it can be used to emulate the target production network that consists of both WLANs and SDN networks. However, the scale of a campus-wide production network is normally very large. Therefore, it is impractical or even impossible to emulate a campus-wide production network on a single standalone computer.
Two distributed SDN emulators, namely Mininet Cluster Edition (Mininet CE) and MaxiNet, have been presented previously on the literature. Mininet CE uses ssh tunnels whereas MaxiNet uses GRE tunnels for the communication between virtual nodes on different machine. Furthermore, Mininet CE adopts either evenly-sized bins/blocks or random placement for placing nodes across cluster. MaxiNet partitions a virtual network onto workers (physical machines) of near equal weight by aiming at confine most of the emulated traffic locally to the workers. However, the partition algorithm of MaxiNet does not consider performance of the workers nor the physical networks. Furthermore, because both Mininet CE and MaxiNet are extended from base Mininet emulator, neither Mininet CE nor MaxiNet supports WLAN emulation.
In this thesis, we present the design and implementation of a distributed edition of OpenNet emulator. Unlike Mininet CE and MaxiNet, Distributed OpenNet decouples the partitioning algorithm from the node placement mechanism so that Distributed OpenNet users can focus on the design of their own partitioning algorithms without concerning about the tedious placement process. Moreover, Distributed OpenNet provides more tunnel options, namely SSH, GRE and VXLAN, for the communication between virtual nodes across physical machines. Finally, Distributed OpenNet has an automatic deployment tool to simplify and expedite the complicate deployment and environment setup procedures, and thus is more user-friendly.

Keywords: Software Defined Network (SDN), Emulator, Tunnel Protocols, IT Automated Tool

1. Introduction - p1
1.1 Motivation - p1
1.2 Objective - p2
1.3 Synopsis - p2
2. Background - p4
2.1 Software-Defined Network (SDN) - p4
2.1.1 SDN Concept - p4
2.1.2 SDN Architecture - p5
2.1.3 OpenFlow - p7
2.2 Mininet - p7
2.2.1 Mininet Design - p8
2.2.2 Mininet Components - p9
2.2.3 Mininet Cluster Edition - p11
2.3 OpenvSwitch - p12
2.3.1 OpenvSwitch Architecture - p14
2.3.2 OpenvSwitch Native Tunneling - p15
2.4 OpenNet - p16
2.4.1 OpenNet Overview - p17
2.5 Ansible - p18
2.5.1 Ansible Concepts - p21
2.5.2 Ansible Deployment Workflow - p22
3. Related Work - p24
3.1 MaxiNet: Distributed Network Emulation - p24
3.2 Mininet Cluster Edition - p24
4. Distributed OpenNet - p27
4.1 Design Concept - p27
4.2 Overview of Distributed OpenNet - p27
4.2.1 OpenNet Slave - p27
4.2.2 OpenNet Master - p28
4.3 Feature of Distributed OpenNet - p28
4.3.1 Pluggable Tunnel Module - p29
4.3.2 Pluggable Partition & Placement Module - p30
4.3.3 Automated Installer - p30
5. System Implementation - p31
5.1 Pluggable Tunnel Module - p31
5.1.1 Build Topology by Mininet - p31
5.1.2 Build Topology by Distributed OpenNet - p32
5.2 Decoupled Placement Mechanism - p34
5.3 Automated Installer - p34
6. Experimental Evaluation - p36
6.1 Environment - p36
6.2 Performance Comparison of Various Tunnels - p36
6.3 Simulate simplied network topology of EC, NCTU - p37
7. Conclusion and Future Work - p38
7.1 Conclusion - p38
7.2 Future Work - p38
8. Reference -p40

[1] O. N. Foundation, “Software-de ned networking: The new norm for networks,” ONF
White Paper, 2012.
[2] B. Lantz, B. Heller, and N. McKeown, “A network in a laptop: rapid prototyping for software-de ned networks,” in Proceedings of the 9th ACM SIGCOMM Workshop on Hot Topics in Networks. ACM, 2010, p. 19.
[3] “Mininet,” https://github.com/mininet/mininet.
[4] “Openvswitch,” http://openvswitch.org.
[5] “Openvwrt,” http://github.com/pichunag/openvwrt.
[6] B. Pfa , J. Pettit, T. Koponen, E. J. Jackson, A. Zhou, J. Rajahalme, J. Gross, A. Wang, J. Stringer, P. Shelar et al., “The design and implementation of open vswitch,” in 12th USENIX Symposium on Networked Systems Design and Implemen- tation, 2015.
[7] M.-C. Chan, C. Chen, J.-X. Huang, T. Kuo, L.-H. Yen, and C.-C. Tseng, “Opennet: A simulator for software-de ned wireless local area network,” in Wireless Communica- tions and Networking Conference (WCNC), 2014 IEEE. IEEE, 2014, pp. 3332–3336.