Abstract: The thesis analyzed cloud computing virtual machine resource model,and aiming at the mapping between virtual machines and physical machines in the model,and the characteristics of virtual machine multi-resource factor and multi-objective optimization,translated the virtual machine allocation problem into multi-dimensional packing problem,introduced multi-objective evolutionary algorithm to solve problem．Firstly,the algorithm designs the growing chain coding and chromosome evaluation function for virtual machine distribution problem,and according to the Encoding,designs two crossover operators and a smart mutation operator,finally,introduces the Hypervolume Measure as population selection Criterion,and designs cloud virtual machine allocation algorithm based on SMS-EMOA．In order to verify the performance of SMS-EMOA,the simulation experiment was made using PMH,CGA,SMS-EMOA,and the experimental results show that SMS-EMOA is better in virtual machine allocation performance.

Key words: Cloud computing,Virtual machine allocated,Multi-objective evolutionary algorithm