|Thesis abstract: |
Traffic in telecommunication networks is estimated to substantially grow in the near future. As the Information and Communication Technology (ICT) sector already consumes a relevant portion of the global world electricity, one of the most challenging issues will be the reduction of its power requirement.
The core/transport section of telecommunication networks is the main target of this dissertation. We will focus on multilayer architectures composed by an optical Wavelength Division Multiplexing (WDM) transport layer providing connectivity support to the classical electronic Internet Protocol (IP) layer. Such network structure is commonly identified as an IP-over-WDM network.
Several strategies can be adopted to reduce the energy consumption of transport networks. The adoption of optical switching technologies can enable substantial power savings, mainly due to the significant reduction in the number of used optoelectronic interfaces and the amount of electronic traffic processing, and thanks to the low-power consumed by photonic devices.
In this thesis several aspects of energy-efficient networking are investigated. Specifically, we first evaluate the network energy consumption performing a comprehensive comparison between different flavours of IP-over-WDM architectures. Such architectures mainly differ from the way they perform signal switching, i.e., either in the electronic (e.g., through IP routers or Digital Cross Connects) or in the optical (via opaque or transparent optical switching matrices) domain.
Fault-tolerance is another aspect studied in this thesis, from the power consumption point of view. We highlight the impact of providing resiliency on the overall network power consumption, considering that protection resources can be set into low-power sleep mode and comparing different protection scenarios, namely, dedicated vs shared and link vs path protection strategies.
Then, a recently proposed switching technique, the Time-Driven Switching (TDS) is also studied from the power consumption perspective. The TDS paradigm is a promising energy-efficient solution since it allows performing traffic grooming of ¿fractions¿ of wavelengths directly in the optical domain, thanks to the time-coordination of network elements, with no need for opticalelectronic signal conversions, so that the consumption due to many optoelectronic interfaces can be saved.
Finally, the potential energy savings of a management protocol for switching on and off optoelectronic interfaces in the network is analytically modelled using a probabilistic approach based on Markov chains, under the assumption that some devices are reserved for high-priority traffic and are set into low-power sleep-mode, whereas the other unused devices are fully poweredoff.
We demonstrate that relevant power savings, up to 60%, can be obtained in comparison to traditional IP-over-WDM network architectures where signal switching and regeneration operations are always accomplished in the electronic IP domain. Moreover, up to 15-20% of power savings are also obtained for protected IP-over-WDM networks by setting protection resources into sleep mode. Finally, we show that exploiting an intelligent strategy for managing the activation of optoelectronic devices enables high power savings, around 75-80% compared to the cases where all devices are fully powered-on, with low impact on the overall network performance (i.e., Quality of Service).