Decentralized workload adaptive control plane for edge software-defined networked systems

Abstract

There is a recent increasing interest on novel Software Defined Net-working (SDN) solutions for controlling edge networked systems. During the daily-basis operation of these systems some unexpected high and varied number of novel flows, trying to use the edge systems resources, could arise at the control system level to be efficiently handled. For this, the current work designs, deploys, and evaluates a new decentralized and online solution that orchestrates a variable number of SDN controllers, following the workload dynamics on the control channel protocol. The orchestration protocol is based on group communication over UDP. Our results show the merits of the proposed online decentralized de-sign on correctly adjusting the number of active controllers to the time-based load variations induced by new data flows traversing the edge networked system.