Software Defined Network Partitioning with Graph Partitioning Algorithms Shivaleela Arlimatti 1(&) , Walid Elbrieki 2 , Suhaidi Hassan 2 , and Adib Habbal 2,3 1 Shaikh College of Engineering and Technology, Belgaum, India sarlimatti@gmail.com 2 InterNetWorks Research Laboratory, School of Computing, Universiti Utara Malaysia, Changlun, Malaysia 3 Computer Engineering Department, Faculty of Engineering, Karabuk University, 78050 Karabuk, Turkey Abstract. Software Defined Networks is an emerging paradigm in Internet communication world that increases the flexibility of today’s networks by decoupling control plane and data plane of the network devices. The funda- mental aim is to centralize the control and reduce the complexity of the net- works. The communication medium between control and data plane is through OpenFlow protocol, an open standard network protocol designed to manage the network traffic by software programs. To increase the scalability and flexibility of controllers the OpenFlow controllers are distributed based on location and network types. However, most critical issue is minimizing the communication cost between the controller domains. In this paper, two graph partitioning algorithms Fiduccia-Matthyses algorithm and Kernighan-Lin algorithm are used to minimize the communication cost between distributed OpenFlow controller domains. The implementation of the algorithms is under Matlab simulation environment. The methodology used for the proposed algorithms is to inter- change the elements from one domain to other domain to calculate the gain. The simulated results show that Kernighan-Lin algorithm minimizes more commu- nication cost rather than the Fiduccia-Matthyses algorithm. Keywords: Fiduccia-Matthyses  Kernighan-Lin  Communication cost  OpenFlow 1 Introduction The current economic slowdown is affecting the technology, industries and their market, networking being one of them. The economic-technological trends of net- working and computing domains are converging. Conventionally, network operators configure network policies separately for individual network devices for efficient deployment of network applications. With the limited tools, they manually transform high-level policies to low-level commands for configuration of network conditions. Considering all these factors, future network infrastructure needs simple, cost- effective and dynamic network management systems that rely more on software rather than hardware. The main point in current networks is the vertical integration of basic © Springer Nature Switzerland AG 2020 F. Saeed et al. (Eds.): IRICT 2019, AISC 1073, pp. 583–593, 2020. https://doi.org/10.1007/978-3-030-33582-3_54