軟體定義網路 (Software-defined networking, SDN) 是一種全新型態的網路架構。透過將傳統網路轉送封包的硬體 (forwarding hardware) 與控制封包轉送的決策單元分離，形成額外的控制層。控制層能夠以程式控制，大幅簡化網路事件回應規則與行為模式的管理，使得硬體與通訊協定的更新更加容易。控制層實體實現方式即為控制器 (controller)，使用者與開發人員可透過北向的API與對控制器下達命令，控制器則透過南向的API來跟硬體溝通。
然而隨著現今大眾對於網路的速率、延遲都有越來越高的要求，越來越多的終端開始存取網路，單一控制器已難以同時處裡如此多的流量。因此，多控制器的SDN架構就被提出，以此提高網路的總負載量、可靠性及擴展性。在論文中，我們採用分散式多控制器系統，並提出了一個在不超過每台控制器的自定義負載量的情況下，最小化每個交換器（switch）與對應控制器的來回通訊延遲 (round-trip time, RTT)。
我們的方法是透過將系統負載與RTT等參數轉換成多重背包問題 (Multiple Knapsacks Problem, MKP)。由於MKP本身是一個NP-hard的問題，所以我們嘗試著提出一個快速的優化演算法，使得每次獲得數據後能找出一組以負載平衡為首要目標的狀態下最小化RTT的解。實驗結果顯示，我們的演算法在正常狀態下能夠找到比其他比較方法都低的RTT值；且在控制器過載的情況下會優先將其調整為正常狀態，然後重新優化RTT。

Software-defined networking (SDN), is a new type of network architecture. By decoupling control packet delivering unit form data packet forwarding hardware in the traditional network, the control unit (control plane) can be directly programmable and dramatically simplifies not only the rules of event reply but the management of network behavior. Under SDN, an update of hardware and protocols are much easier. The implementation of control layer unit is controller. Developers and administers can send orders to controller through the northbound API, and controller will communicate to the network hardware through the southbound API.
As the requirement of high speed and low delay in modern network, a single controller system is hard to handle such situation. Therefore, the multiple controller architecture is proposed to enhance network capacity, reliability and scalability. In this thesis, we adopt a distributed multi-controller system. Under the premise that each controller’s load will not exceed a pre-defined threshold, the round-trip time (RTT) between controllers and switches will be minimized.
We transform the above question into a multiple knapsack problem (MKP). Since MKP is a NP-hard problem, we proposed an optimization algorithm to find a minimum RTT solution under the constraint of controller load balance in a faster way.
Simulation results show that our method is able to find a better RTT value. When any controller is overloaded, our method also provides a fast algorithm to turn it into a feasible solution and then optimizes it again.

Table of Contents
Acknowledgement................................................iii
Abstract........................................................iv
Table of Contents...............................................vi
List of Figure................................................viii
Chapter 1. Introduction......................................1
Chapter 2. Related Work......................................4
2.1. Software-Defined Networking.............................4
2.2. Centralized Multi-Controllers System....................5
2.3. Distributed Multi-Controllers System....................6
2.4. Multi-Controllers System Optimization...................6
2.5. The Choices of Controller Platforms.....................7
2.6. In-Band Mode and Out-of-Band Mode.......................8
Chapter 3. System Model......................................10
Chapter 4. The Load Balancing Procedure......................12
4.1. The Goal of Our Studies.................................12
4.2. Load Collection Module..................................12
4.3. Load Balancing Module...................................14
4.4. Switch Migration Module.................................15
Chapter 5. Proposed Methods in Load Balancing Module.........18
5.1. Problem Transformation..................................19
5.2. Fast Overload Handle Algorithm..........................22
5.3. DFS-based BnB Pruning Algorithm.........................23
Chapter 6. Experimental Evaluation...........................28
6.1. Experiment Environment..................................28
6.2. Compared Algorithms.....................................33
6.3. Performance Evaluation..................................36
Chapter 7. Conclusion........................................41
Reference.......................................................42

[1] D. Kreutz, F. M. V. Ramos, P. E. Veríssimo, C. E. Rothenberg, S. Azodolmolky and S. Uhlig, “Software-Defined Networking: A Comprehensive Survey,” in Proceedings of the IEEE, vol. 103, no. 1, pp. 14-76, Jan. 2015.
[2] O. Blial, M. Ben Mamoun and R. Benaini, “An Overview on SDN Architectures with Multiple Controllers”, in Journal of Computer Networks and Communications, vol. 2016, Apr. 2016.
[3] Dan Levin, Andreas Wundsam, Brandon Heller, Nikhil Handigol, and Anja Feldmann. “Logically centralized? State distribution trade-offs in software defined networks.” In Proceedings of the 1st ACM International Workshop on Hot Topics in Software Defined Networks. Pages 1-6. ACM, 2012.
[4] “NOX/POX,” https://github.com/noxrepo/, [Online; accessed July 15, 2019].
[5] “Ryu,” https://osrg.github.io/ryu/, [Online; accessed July 15, 2019].
[6] “Floodlight,” http://www.projectfloodlight.org/floodlight/, [Online; accessed July 15, 2019].
[7] “OpenDaylight,” https://www.opendaylight.org/, [Online; accessed July 15, 2019].
[8] “ONOS,” https://onosproject.org/, [Online; accessed July 15, 2019].
[9] Y. Hu, W. Wang, X. Gong, X. Que and S. Cheng, "BalanceFlow: Controller load balancing for OpenFlow networks," International Conference on Cloud Computing and Intelligence Systems, Hangzhou, pp. 780-785, IEEE, 2012.
[10] A. Dixit, F. Hao, S. Mukherjee, T. V. Lakshman and R. R. Kompella, "ElastiCon; an elastic distributed SDN controller," 2014 ACM/IEEE Symposium on Architectures for Networking and Communications Systems (ANCS), Marina del Rey, CA, pp. 17-27, 2014.
[11] Y. Zhou et al., "A Load Balancing Strategy of SDN Controller Based on Distributed Decision," International Conference on Trust, Security and Privacy in Computing and Communications, Beijing, pp. 851-856, IEEE, 2014.
[12] B. Heller, R. Sherwood, and N. Mckeown. “The controller placement problem.” Proceedings of the first workshop on Hot topics in software defined networks, NY, USA, pp.7-12, ACM, 2012.
[13] G. Yao, J. Bi, Y. Li and L. Guo, "On the Capacitated Controller Placement Problem in Software Defined Networks," in IEEE Communications Letters, vol. 18, no. 8, pp. 1339-1342, Aug. 2014.
[14] A. Sallahi and M. St-Hilaire, "Optimal Model for the Controller Placement Problem in Software Defined Networks," in IEEE Communications Letters, vol. 19, no. 1, pp. 30-33, Jan. 2015.
[15] Jinke Yu, Ying Wang, Keke Pei, Shujuan Zhang and Jiacong Li, "A load balancing mechanism for multiple SDN controllers based on load informing strategy," Asia-Pacific Network Operations and Management Symposium (APNOMS), Kanazawa, pp. 1-4, 2016.
[16] H. Chen, G. Cheng and Z. Wang, "A game-theoretic approach to elastic control in software-defined networking," in China Communications, vol. 13, no. 5, pp. 103-109, May 2016.
[17] C. Liang, R. Kawashima and H. Matsuo, "Scalable and Crash-Tolerant Load Balancing Based on Switch Migration for Multiple Open Flow Controllers," International Symposium on Computing and Networking, Shizuoka, pp. 171-177, 2014.
[18] “JGroups,” http://www.jgroups.org/, [Online; accessed July 15, 2019].
[19] “OpenFlow switch specification Ver. 1.3.4” https://www.opennetworking.org/software-defined-standards/specifications/, [Online; accessed July 15, 2019].
[20] D. B. Shmoys and E. Tardos. An approximation algorithm for the generalized assignment problem. Mathematical Programming A, 62:461-474, 1993.
[21] “Cbench,” https://github.com/mininet/oflops/tree/master/cbench, [Online; accessed July 15, 2019].
[22] K. Wang, S. Kao and M. Kao, "An efficient load adjustment for balancing multiple controllers in reliable SDN systems," IEEE ICASI, pp. 593-596, Chiba, 2018.