Energy-saving Technique Research And Green Configura Tion Strategy Design Of Server Clusters In Cloud Environ Ment

With the growing demand for cloud computing, the large scale data centers are established constantly which will need a lot of energy to maintain its operation. The maintenance fees will exceed the costs on purchasing the system hardware with the exhausted of traditional energy resources and the rise of prices. How to solve the problem of operating data center is becoming urgent. In recent years, the research hot spot is based on the conception of "green" configuration. Namely, Data center energy consumption and system load is in direct proportion. However, the recent cloud computing data center, for example, HOTMAIL and Microsoft institute can not reach this goal. Even when the system is idle, energy consumption only reached half values of its peak. As for cloud computing data centers are composed of a large number of servers, an effective green configuration which is accepted widely by domestic and foreign scholars, is the load matching by using the software to dynamic adjust the number of active server system; reasonable distribution of load makes some server in the system task scheduling, in the low load period need not assume any task, entering a low power mode to save energy.The object that this dissertation studied was data center composed of isomorphic servers. The operational process was divided into several time periods. The load information of each period could be got based on historical statistics or online prediction. The "Green Configuration" strategy was divided into offline and online algorithms. Offline algorithm was applicable to the case of the known load per period and the most optimal results of it on the one hand could provide accurate assessment of the number of active servers, on the other hand set a good foundation for the online prediction algorithm.Firstly, the energy consumption during operation and the switching energy between adjacent periods were considered. The minimized power consumption mathematical model of data center was established. The nonlinear constrains arose from the model could be simplified by formula reconstruction. As a basis of the optimal offline algorithm that subsequently discussed, the optimal solution for two special cases (energy consumption without operation and energy consumption without switch) was discussed, then optimal characteristics under normal circumstances were arose and the relevant lemma and the proof were given.Secondly, the optimal offline algorithm of server’s green configuration was given. This algorithm was based on integer dynamic programming idea and generally with an exponential complexity. It had polynomial complexity by eliminating the dynamic programming recursion process. The numerical results of different load change trends indicated that optimal offline algorithm proposed could ensure the minimum energy consumption while calculation process was stationary.Finally, according to the worst-case scenario that workload may meet, online algorithm was designed which followed the ideas of receding horizon control. The conclusion that online algorithm is approximate optimal could be obtained by comparing the results of simulation and optimal offline algorithm.It is an effective energy-saving measure to switch the servers to the idle mode to reduce power consumption while keep an appropriate number of active servers to provide timely services. This paper provides a certain extent theoretical achievement for green configured technology of cloud computing data center.