Research On Energy-efficient Technology In Green Data Center

Nowadays, human society is making unceasing progress. The problem of energy crisis and climate is getting worse and worse, and all walks of life are seeking ways of energy saving and emission reducing. With the development of ICT(Information & Communication Technology), humans are stepping into the Internet age, the Internet industry makes great contribution to energy consumption. Cloud computing has become the most popular technology of Internet. As the core infrastructure of cloud computing, data center consumed much greater energy than any other parts of data center. Thus, researching on the energy-saving in data center is very meaningful. The researchers are mainly focus on data center network and sever.Firstly, the thesis studies on the energy-saving problem for switch, because data center network is made up of thousands of switches, which consume most parts of electric power in data center network. Many studies show that sleeping mode is quite popular for energy-saving of switch, the main idea of sleeping mode is setting the switch into sleeping mode, which is idle or in a very low-flow statement, then transferring the low-flow demand into another switch. In sleeping mode, switch consumes very tiny electric power, so as to awake it into working mode in case of burst traffic. Based on sleeping mode, the thesis studies on the way to save energy for switch. The thesis has studied on the topology of data center network and analyzed the characteristics of all kinds of typical topologies, such as Fat Tree, DCell and BCube, then the thesis chose the classic Fat Tree as data center network topology. Based on Fat Tree, a GA(Genetic Algorithm) was proposed to save energy in data center network, the simulations reveal that GA is obviously better than the compered energy-saving algorithm Elastic Tree. Meanwhile, with the success of energy-saving, the topology becomes weak to link and node failures. In order to keep topology reliable, the thesis has adjusted the energy-saving topology by adding switches and links to provide equal cost multi-path for each demand.Finally, the thesis also study the energy-saving problem of VDC(Virtual Data Center). By using virtualization technology in data center, the thesis can make contribute to improve the ability of energy-saving. The mapping from virtual data center to physical data center includes virtual machine mapping and virtual link mapping. Based on the case of online virtual data center request mapping, the thesis has proposed two algorithms, ESA_SFM(Energy Saving Algorithm of Server First Mapping) and ESA_JM(Energy Saving Algorithm of Joint Mapping). With these two methods, the thesis has reached the goal of saving power in the whole data center.