Open-Source Software Defined Networking Controllers: State-of-the-Art, Challenges and Solutions for Future Network Providers

Abstract

Software Defined Networking (SDN) is programmable by separation of forwarding control through the centralization of the controller. The controller plays the role of the ‘brain’ that dictates the intelligent part of SDN technology. Various versions of SDN controllers exist as a response to the diverse demands and functions expected of them. There are several SDN controllers available in the open market besides a large number of commercial controllers; some are developed to meet carrier-grade service levels and one of the recent trends in open-source SDN controllers is the Open Network Operating System (ONOS). This paper presents a comparative study between open source SDN controllers, which are known as Network Controller Platform (NOX), Python-based Network Controller (POX), component-based SDN framework (Ryu), Java-based OpenFlow controller (Floodlight), OpenDayLight (ODL) and ONOS. The discussion is further extended into ONOS architecture, as well as, the evolution of ONOS controllers. This article will review use cases based on ONOS controllers in several application deployments. Moreover, the opportunities and challenges of open source SDN controllers will be discussed, exploring carrier-grade ONOS for future real-world deployments, ONOS unique features and identifying the suitable choice of SDN controller for service providers. In addition, we attempt to provide answers to several critical questions relating to the implications of the open-source nature of SDN controllers regarding vendor lock-in, interoperability, and standards compliance, Similarly, real-world use cases of organizations using open-source SDN are highlighted and how the open-source community contributes to the development of SDN controllers. Furthermore, challenges faced by open-source projects, and considerations when choosing an open-source SDN controller are underscored. Then the role of Artificial Intelligence (AI) and Machine Learning (ML) in the evolution of open-source SDN controllers in light of recent research is indicated. In addition, the challenges and limitations associated with deploying open-source SDN controllers in production networks, how can they be mitigated, and finally how open-source SDN controllers handle network security and ensure that network configurations and policies are robust and resilient are presented. Potential opportunities and challenges for future Open SDN deployment are outlined to conclude the article.