Digital-twin-assisted Software-defined PON: A Cognition-driven Framework for Energy Conservation

Abstract

The energy consumption footprint of Passive Optical Networks (PONs) is high due to its widespread deployment. Therefore, a large and growing body of literature has investigated how energy consumption in PONs can be minimized. Existing research indicates that sleep mode is an effective means for reducing energy consumption in PONs. There is also a growing number of contributions that demonstrate the importance of integrating PON with Software-defined Networking (SDN). Such integration facilitates a programmable PON, which can dynamically adjust its operation on-the-fly based on the decision made by a SDN controller. The most important limitations in the previous energy-efficient PON research lies in the fact that they failed to address how the energy saving and traffic performance can be fed in real-time into the network operation decision process. Digital Twin (DT) is an emerging field of research in the communication networking area. DT can bridge between real and virtual world making a significant contribution in defining optimal operation of the network. Therefore, in this paper, we propose a Digital-twin-assisted Software-defined PON framework in order to feed the network performance data in real-time to the network operation decision support system, which aims at maximizing energy conservation while meeting quality of service, and control PON infrastructure on-the-fly. Our DT based solution also integrates a supervised learning Autoregressive Integrated Moving Average (ARIMA) method for making traffic arrival forecasts based on time series traffic traces. The initial findings of our research based on a C++ discrete event simulator show that the proposed solution successfully reduces energy consumption while meeting traffic delay requirements in a PON.