Research Area:  Software Defined Networks

Abstract:

Software-Defined Networking (SDN) controllers are nowadays expected to manage large infrastructures and services in a practical, efficient, and optimized way. To achieve such control, SDN networks should increase the automation of orchestration functions to fullfil both service and network management lifecycles. However, network management applications have stringent resource demands that orchestrators need to honor. Most of the available orchestrators focus on high-level approaches that rely exclusively on the Northbound API (NB-API). A few strategies have also been designed to optimize the controllers internal mechanisms through features enabled by the Southbound API (SB-API). However, such solutions do not focus on optimal information delivery to the orchestrator, a crucial property to provide prompt feedback in response to network events, and to potentially drive self-healing and auto-scaling properties. To address such need, we present NOTORIETY, an orchestration system that provides an abstraction for real-time SDN network controller event message handling. NOTORIETY is able to maximize the orchestration capabilities of SDN controllers including innovative features for processing and delivering network event information. NOTORIETY-s design consists of mechanisms that empower SDN-controlled entities by applying filtering rules to efficiently control and optimize data flows provided by the SDN controller. We implemented a testbed to assess our NOTORIETY proposal following the guidelines provided in the RFC 8456, and we benchmark its performance over three SDN controllers (OpenDaylight, ONOS, and Floodlight). The evaluation results reveal the effectiveness of NOTORIETY in reducing the execution time of SDN requests, load processing, and overall data marshaling volume in several scenarios.

Keywords:  
Software-defined networking
SDN controller
orchestration
network programmability

Author(s) Name:  Kevin Barros Costa; Felipe S. Dantas Silva; Lucas M. Schneider; Emídio Paiva Neto; Augusto Venancio Neto; Flavio Esposito

Publisher name:  IEEE

DOI:  10.1109/ACCESS.2020.3033486

Volume Information:  ( Volume: 8) Page(s): 195487 - 195502

Paper Link:   https://ieeexplore.ieee.org/document/9237921