軟體定義網路 (Software-defined network, SDN) 是一種新型的網路架構，它得以使網路集中且智能地被控制。 藉由分離控制網路封包的元件與在交換機及路由器 (switches and routers) 中轉傳封包的硬體，控制層得以被程式化，這有助於開發商及管理員更有彈性地管理網路。 網路管理者可以透過北向的API操作控制器 (controller)，而控制器可以透過南向的API與交換機及路由器進行溝通。

現今，由於鏈接容量的限制，在傳統網路中，單一路徑傳送已經無法滿足我們的需求。為了解決這一問題，我們提出在軟體定義網路下的多路徑傳輸。為了保證所有起始點與終點配對的傳送速率，我們將多物網路流問題 (Multicommodity Flow Problem, MCFP) 演算法應用於軟體定義網路，並提出基於SDN的MCFP機制來改善吞吐量 (throughput) 的效能。 在設計中，我們使用一個集中式的SDN控制器來決定多種商品的最佳路徑，並且控制器會將這些路徑規則轉送到這些路徑上。 實驗結果顯示，我們的方法能夠在多條路徑上分配流量，且可以滿足所有配對的速率需求。

Software defined networking (SDN) is a new network architecture enables the network to be centrally and intelligently controlled. By decoupling control network packet unit from packet forwarding unit in switches/routers, the control plane could be programmable which helps developers and administrators manage the network more flexibly. Network managers could program controller via northbound API, and controllers will communicate to switches and routers via southbound API.

Nowadays because of the limit of link capacity, single-path transmission used in traditional networks is insufficient. To solve this problem, we address multi-path transmission in software defined networks. To provide rate guarantee for all source-destination pairs, we apply the multicommodity flow problem (MCFP) algorithm to SDN and propose SDN-based MCFP scheme to improve throughput performance. In this design, a centralized SDN controller could decide optimal paths for multiple commodities and the controller would sending the flow rules to them. Simulation results show that our method which is able to distribute flows over multiple paths can meet the rate-demand for all source-destination pairs.

[1] Iperf3. https://iperf.fr/iperf-download.php.
[2] Mininet. http://mininet.org/.
[3] Netconf. https://tools.ietf.org/wg/netconf/.
[4] NOX/POX. https://github.com/noxrepo.
[5] ONOS. https://opennetworking.org/onos/.
[6] OpenDaylight. https://www.opendaylight.org/.
[7] OpenFlow Switch Specification Ver. 1.3.0. https://opennetworking.org/wpcontent/uploads/2014/10/openflow-spec-v1.3.0.pdf.
[8] Ryu. https://github.com/faucetsdn/ryu.
[9] Wireshark. https://www.wireshark.org/.
[10] Mark Fedor, Martin Lee Schoffstall, James R. Davin, and Dr. Jeff D. Case. Simple Network Management Protocol (SNMP). RFC 1157, May 1990.
[11] Lisa K Fleischer. Approximating fractional multicommodity flow independent
of the number of commodities. SIAM Journal on Discrete Mathematics,
13(4):505–520, 2000.
[12] Mordechai I. Henig. The shortest path problem with two objective functions. European Journal of Operational Research, 25(2):281–291, 1986.
[13] Chih-Lin Hu. Improving multipath tcp routing performance in softwaredefined networks. https://hdl.handle.net/11296/44tja6, 2020.
[14] Xiangshan Sun, Zhiping Jia, Mengying Zhao, and Zhiyong Zhang. Multipath load balancing in sdn/ospf hybrid network. In IFIP International Conference on Network and Parallel Computing, pages 93–100. Springer, 2016.
[15] Phyo May Thet, Parichat Panwaree, JongWon Kim, and Chaodit Aswakul. Design and functionality test of chunked video streaming over emulated multipath openflow network. In 2015 12th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON), pages 1–6. IEEE, 2015.
[16] George Tsaggouris and Christos Zaroliagis. Qos-aware multicommodity flows and transportation planning. In 6th Workshop on Algorithmic Methods and Models for Optimization of Railways (ATMOS’06). Schloss Dagstuhl-Leibniz-Zentrum f¨ur Informatik, 2006.
[17] Cheng-Ying Wu. Balanced service chaining with traffic steering
in software defined networks with network function virtualization.
https://hdl.handle.net/11296/9pf84q, 2015.