Energy Saving Research For Green Communications Networks

WiththerapiddevelopmentofInformationCommunicationTechnology(ICT),networkap-plications and scales expand increasingly, bringing large amount of energy consumption. Thisnot only increases the cost of network providers, but also leads to massive Green House Gas(GHG) emission, becoming an important issue of ICT. Therefore, while enjoying the conve-nience provided by the powerful communication networks, we have to face the energy problemand have deep research in this problem. As current networks are hybrid networks composed ofwired and wireless networks, this dissertation focuses on the energy saving research of hybridnetworks in ICT. The contributions of this dissertation include:1) Current network devices, designed to endure peak loads and degraded conditions, isof low energy efficiency when they are with low traffic and in idle status. In this sense,putting the underutilized devices into sleeping is an efficient way to decrease the energyconsumption in a network. Therefore, this dissertation proposes a prediction-based trafficmanagement scheme to realize energy saving, which tries to aggregate and reroute thetrafficinsomelinkstomakethemidleandthensleepwhentheirtrafficloadsarepredictedlight in the future.2) Energy efficiency is a main issue for cooperative communication in wireless network-s. This dissertation proposes a distributed energy-efficient topology control to improveenergy efficiency, which jointly considers network capacity and energy consumption interms of bits per Joule. This scheme can choose proper transmission patterns through anadaptive coalition formation algorithm to improve the network energy efficiency and pro-long network lifetime. The scheme is proved with convergence property, low complexityand stable coalition structures.3) Hybrid networks composed of all-optical and wireless networks suffer lots of retrans-missions caused by packets drop events, which not only results in performance degrada-tion, but also wastes much energy to retransmit the lost packets. To reduce the energyconsumption for retransmissions, this dissertation proposes a domain-based TCP basedon SEMI-TCP, in which a TCP connection is split into multiple domains along the pathand congestion control is managed domain-by-domain. Simulation study demonstrates that this approach can significantly improve the performance of hybrid networks in termsof energy efficiency.4) In green networking research, energy consumption evaluation methods are required tocalculate how much energy an approach consume. However, due to the complexity ofnetworks, it is difficult for mathematical modeling to achieve this goal. This dissertationproposes a simulation-based energy consumption estimation method, which consider allthe energy consumption aspects from the life-cycle of packets and work status of nodes.This estimation method is completely independent on the details of network protocolswhile it is simple, convenient and low-cost.