Abstract
The ever-increasing number of network-capable devices places a massive burden on modern networks. Communication infrastructure should provide quality-of-service essentials in terms of high-bandwidth capacity, scalability, resiliency, and security. Programmable networks are viewed as the prevailing method of encountering the challenges introduced by the accelerated expansion. The ability of software-defined networking (SDN) to separate the control plane from the data plane and enable the programmability of the network creates new ways to architect the network. The centralization of control logic introduces complexities in large-scale, distributed networks such as performance bottlenecks and reliability. Distributed SDN controllers have been proposed to overcome the performance concerns. The lack of a communication standard among distributed controllers, referred to as the East/West interface, presents a challenge in the adoption of SDN in large-scale, distributed networks. In this paper, we propose Distributed SDN control plane Framework (DSF) - a framework for the East/West interface for heterogeneous, distributed SDN controllers to synchronize topologies using a standardized data-centric real-time publish/subscribe paradigm known as the Data Distribution Service (DDS). Distributed control plane architectures are proposed using DSF: flat, hierarchical, and T-model. The DSF interface is implemented on multiple SDN control plane platforms to evaluate performance: Floodlight and Open Network Operating System (ONOS) controllers. Test cases with different configurations are designed for performance evaluation of the proposed interface in homogeneous and heterogeneous SDN control planes. In addition, a performance comparison is presented of DSF-based ONOS controllers versus Atomix-based ONOS cluster solutions.
Highlights
Present-day and future networks face an ever-increasing demand of providing fast and reliable interconnectivity among a growing number of network capable devices, estimated to reach 29.3 billion by the year 2023 [1]
ANALYSIS OF NETWORK CONVERGENCE POINT Fig. 11 describes the time taken for networks to converge into a single, holistic topology after a link discovery update packet is published by a control plane entity
The standard error of the mean (SEM) of the 10 repetitions is roughly equal between the different network configurations except for the network with two FL-Distributed SDN control plane Framework (DSF) controllers which takes less than half-a-second to converge
Summary
Present-day and future networks face an ever-increasing demand of providing fast and reliable interconnectivity among a growing number of network capable devices, estimated to reach 29.3 billion by the year 2023 [1]. Enterprise and Data Center Networks (DCNs) further the concern by necessitating Quality-of-Service (QoS) essentials in terms of high-bandwidth capacity, scalability, resiliency, and security. Software-defined networking (SDN) [2] is a programmable network concept designed to overcome the limitations of conventional networks such as configuration and. SDN proposes a programmable network control, decoupled from forwarding. The concept basis on two main characteristics: separation of the control and data planes and programmability of the control plane [3]. Control functions are transferred from data forwarding devices to a logically centralized network entity, the controller. The programmability of SDN is derived from the instructions given to the controller via application programming interfaces (APIs). Network-wide traffic forwarding decisions can be set by the controller simplifying network configuration, policy enforcement, and evolution [4]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
