Cloud computing enables users to have access to resources on demand. This leads to an increased number of physical machines and data centers in order to fulfill the needs of users which are continuously on the increase. The increase in the number of active physical machines is directly proportional to the increase in the energy consumption. Thus, minimization of energy consumption has become one of the major challenges of cloud computing in recent years. There are many ways to power savings in data centers, but the most effective one is the optimal placement of virtual machines on physical machines. In this thesis, the problem of dynamic placement of virtual machines is solved in order to optimize the energy consumption. A cloud computing model is built along with the energy consumption model considering the states of physical machines and the energy consumption during live virtual machine migrations and the changes in the states of physical machines. The intelligent algorithms having a centralized approach, like genetic algorithm and simulated annealing algorithm have been used to solve the dynamic virtual machine placement problem in earlier research works but many unreachable solutions may result. Thus, a decentralized approach based on game theoretic method is used here in order to reach optimal solutions and also a list of executable live virtual machine migrations is provided to reach the optimal placement. In real world scenario, physical machines may or may not cooperate with each other to arrive at an optimal solution. Therefore, in this thesis both cooperative as well as non-cooperative game theoretic approaches have been used to find optimal solution to the dynamic virtual machine placement problem. It is seen that Nash equilibrium is achieved in polynomial time. The experimental results are compared with the results of best fit approach. Results show that energy consumption is minimized by modifying the placement of virtual machines dynamically.
展开▼
