RDSNET: A proposal for control architecture for software defined MANETs
Sergio Francisco Mora, Jhon Edwin Vera · International Journal of Engineering and Technology · 2018
Software-defined networks (SDN) have quickly gained importance in telecommunications.This paper has examined the performance of SDN's techniques used in wireless networks, specifically in ad hoc mobile networks (MANET).The architecture of SDN control specially designed for MANETs has been proposed under the name RDSNET.RDSNET provides a solution to the design and deployment of the controllers in these networks.Several simulation scenarios were built in OMNET ++ to decide whether the design of a distributed or centralized controller was the most suitable for implementation.The results showed that the RDSNET distributed configuration yielded better performance.It was also shown that, despite the additional overload introduced by synchronizations, the performance was similar to a conventional MANET without SDN for the message success rate.However, the latency with RDSNET does not improve over SDN networks.Thus, though flexibility is the advantage of SDN for MANETS, latency is the main disadvantage.Keyword -Software defined networking, Control, MANET I. INTRODUCTION Software-defined networks are a new paradigm of network architecture.They have aroused great interest in the telecommunications community lately.The SDN permit better control and administration, rapid reconfiguration, and dynamic traffic.They also have the potential to evolve and facilitate innovation in networks [1].These characteristics make them interesting to companies and service providers enabling them to build highly scalable, flexible and adaptive networks that cater for the different needs of the industrial sector.Research carried out with SDN so far has focused mainly on wired networks and wireless networks [2-3].However, little attention has been paid to the use SDN in ad-hoc mobile network (MANET).There are potential benefits in the use of SDN in the MANETs, where the characteristics of the SDN could be immensely useful.Such benefits must be weighed against the cost of wireless channel solutions needed to replace the conventional wired SDN implementation.Most works related to wireless SDN in cellular networks,however, their use has been extended to several types of networks, for example, the authors of [4] identified use cases for SDN in cellular networks and proposed interesting extensions to SDN wired that overcome some inherent limitations to this type of network.For wireless sensor networks (WSN), SDN architectures such as those introduced in [5] have been presented, although in this work they do not present simulations or test results that allow the design to be validated.Recently, the Open Networking Foundation (ONF) has initiated a working group to find use cases for SDN in wireless networks [6].A key feature of the SDN is the decoupling of the control and data planes.The controller is central to this type of network since it has maintaining and configuring the network, ensuring that the general aims of the network and the administration of it are being met.The intelligence of the SDN resides mainly in the controllers whereas the routers focus on the transmission of data.Currently, most wired SDN use the OpenFlow protocol at first proposed in [7].At first thiswas implemented with a single controller scheme, until [8] it was determined that multiple controllers improved the performance until a limit of operation of the net was reached.The most commonly used controller in SDN is NOX and its Python-based POX successor [9], both run with OpenFlow.The scalability and location of the SDN drivers are a still open topic in wireless networks [10,11].This paper seeks to analyze the challenges of using SDN in MANET networks and proposes RDSNET, a driver architecture specially designed for MANETs' operationalconstraints and requirements, such as frequent route changes and mobility.The performance of RDSNET was evaluated using simulations in OMNET ++ in two scenarios (i) a centralized controller design, (ii) a distributed system of controllers.