Power reduction strategies with differentiated quality of protection in IP-over-WDM networks

This work considers two important aspects of modern communication networks, network survivability, and energy efficiency. Survivability is a design requirement to provide failure recovery against network outages such as fiber cuts. To ensure survivability, traditionally, spare network (i.e., backup) resources are reserved in case of failures of primary (i.e., working) network resources. On the other hand, energy efficiency is required to cope with the continuous growth of the Internet traffic. Backup resources increase the energy consumption, especially if constantly powered-on. Hence, a trade-off between energy efficiency and network resiliency arises. To increase transport capacity and reduce energy consumption, wavelength division multiplexed (WDM) networks are adopted as the most practical solution for data transport in core networks, which are the focus of our work. In WDM networks, different levels of protection (i.e., dedicated, shared) can be implemented for different traffic demands, and different protection levels are characterized by different power consumptions. We consider a differentiated quality of protection (Diff-QoP) scheme where different levels of protection are provided. In this context, we investigate on two different power-reduction strategies to be used in protected WDM networks: (i) setting backup devices into low-power modes (sleep mode) and (ii) adapting the devices (i.e., transmitting/receiving equipment) usage to hourly traffic variations. We present exact modeling through Integer Linear Program (ILP) of the three scenarios and a provisioning algorithm to solve the problem of power minimized design of resilient optical core networks, under static and dynamic traffic conditions. We evaluate the performance of the proposed algorithm under static traffic conditions by comparing the obtained results with the optimal solution, in absence of traffic grooming. We show that the proposed heuristic reduces the computational time by three orders of magnitude with an optimality gap ranging from 8.88 up to 23.88%. Furthermore, we include traffic grooming and solve the problem under dynamic traffic conditions. Our findings show that enabling sleep mode (SM) for backup devices can help reduce total power consumption up to 20 and up to 38% when considering with Diff-QoP. Finally, adapting the number of transmitting/receiving devices to actual traffic needs guarantees power savings up to 80%, especially during off-peak hours.