Probability based virtual machines placement for green data center

Virtual Machine Placement (VMP) is regarded as an important criterion to improve resource utilization and reduce energy consumption for cloud data centers. The existing VMP schemes simply set the VM resource requirements fixed values and ignore their fluctuation characteristics. Assuming normal distribution resource requirements, we firstly present a model for data centers based on a more accurate energy consumption model for single machine. Then, an effective genetic algorithm is adopted to solve this model. In the algorithm, some important issues, such as the number of population, fitness function and calculating method of energy consumption are discussed. In the end, we validate our method by experiments. © 2016 SERSC.