چكيده به لاتين
Over recent years, green computing has been proposed as an emerging strategy to reduce the Carbon footprint produced by the networking sector. The relevance of this trend turns back to its impact on environmental pollution and economic cost. A significant amount of energy conservation can be obtained by switching redundant or unused network components (node and link) to inactive mode during low traffic situation, referred to as sleep-scheduling. However, the quality of service measures should be satisfied during sleep-scheduling.
In this thesis, in the first, a distributed and holistic architecture is proposed for each node that composed of four components including traffic selection، information management، energy management and node management. According to this architecture and based on type of traffic awareness (local or global) quality of service constraints, two local traffic awared methods and on global traffic awared method are proposed in order to select sleep components. In addition, based on type of traffic awareness, two dynamic proposals are introduced in order to decision in sleep-scheduling that decide to switch off/on components according to current traffic situation. Also, two switching off and switching on methods are proposed in order to switch state of components in each node. In addition to sleep-scheduling proposals, in order to achieve more energy saving, an adaptive link rate approach is proposed to reduce capacity of some remaining active links based on their loads.
Finally, we simulate all proposals using real network scenarios and evaluate them based on energy conservation and network performance metrics. Results show that our proposals can switch off network nodes and the maximum number
of switchable links respectively up to 2% and from 7% to 10% more than existing alternatives in the literature, while the network performance is not altered considerably. Furthermore, evaluation results show that our proposed power-proportional approach can acheive 15% to 35% of energy conservation in the network. Moreover, our proposal can dynamically switch on/off network elements according to current traffic loads, which minimizes the degradation of quality of service offered by the network.
