Energy Saving vs. Performance: Trade-offs in Optical Networks
Abstract: The energy consumption of communication networks is continuously growing. Many energy saving approaches have been proposed at the device, system, and network level. The most promising way to address this problem is to utilize photonic technologies as much as possible thanks to their low energy consumption per bit performance. Moreover, several approaches have been proposed to further reduce the energy consumption in optical networks. One popular technique exploits low power modes (e.g., sleep or doze mode) for devices that are not used. However, sleep mode based approaches may affect the way optical connections (i.e., lightpaths) are routed, or alter the characteristics of some devices. This in turn may have a detrimental impact on crucial network/device performance parameters. In other words a green approach may introduce additional delay, change the level of resource utilization in the network, or even impact the lifetime of a device, resulting in increased network operational cost. This thesis provides a study that carefully assesses, in both access and core networks, the trade-off between the benefits of sleep-based energy-efficient schemes and their possible side-effects.In fiber access networks putting a device into sleep mode and waking it up can introduce a significant energy overhead. Already proposed energy-efficient approaches reduce this overhead by aggregating as much as possible the traffic before a transmission. However, aggregating data may cause an additional delay that in some cases might not be acceptable. This thesis investigates the trade-off between energy saving and additional packet delay in the case of a LTE backhaul network based on wavelength division multiplexing passive optical network (WDM-PON). The thesis proposes a novel energy-efficient approach based on the dozing concept able to precisely control when a transmitter needs to wake up in order to maximize the time spent in sleep mode, while assuring that packet transmissions are completed before a given deadline. The proposed scheme is also able to exploit possibly diverse traffic delay requirements to further improve energy saving performance.In optical core networks, one way to decrease the energy consumption is to minimize the number of used active devices by aggregating the lightpaths on the lowest possible number of active fiber links. Routing strategies based on this intuition are beneficial in terms of energy saving, but on the other hand may impact the network performance (e.g., blocking probability) by affecting length of the lightpaths and link occupancy distribution. This trade-off is evaluated in the thesis with the help of a specially designed routing and wavelength assignment (RWA) strategy referred to as weighted power aware lightpath routing (WPA-LR). The WPA-LR strategy permits the fine tuning between the minimization of two objectives: energy consumption and network resource (i.e., wavelength) utilization. Evaluation results confirm that energy efficiency and network performance are conflicting objectives. However, the proposed WPA-LR strategy offers energy minimization with acceptable impact on the network performance.The thesis also investigates the impact that sleep-based energy-efficient strategies have on the lifetime of a number of optical network devices, in both access and core networks. In fact, utilizing a sleep mode functionality may change the operational conditions of the device which can impact the device lifetime. This is a crucial aspect to consider because it may directly affect the network operational cost related to fault management. The thesis provides a methodology to assess under which conditions and for which devices an energy-efficient scheme may lead to overall cost benefit vs. a (possible) increase of reparation cost. It was found that in access networks and with business customers a small lifetime variation in optical line terminals (OLTs) or in optical network units (ONUs) can lead to significant cost increase that cannot be covered by the profits coming from the energy saving. In core networks erbium doped fiber amplifiers (EDFAs) are the most vulnerable devices in terms of impact on their lifetime. For this reason it was found that the usage of green routing algorithms based on putting EDFAs into sleep mode may not always be economically beneficial.In conclusion this thesis provides a different perspective on sleep mode based energy-efficient algorithms where the potential benefit in terms of energy saving is weighted against the impact of a possible degradation of the network performance and devices lifetime. On the other hand these performance degradations can be controlled and limited by the proposed algorithms.
This dissertation MIGHT be available in PDF-format. Check this page to see if it is available for download.