Abstract: Cloud computing is a hot spot in computer research field, in which the performance of task scheduling algorithm directly affects total performance of cloud computing systems. For satisfying users' different service requirements, a dual fitness genetic annealing algorithm (DFG2A) is presented. Task scheduling scheme based on this algorithm can balance different attribute requirements for each task, and improve customer satisfaction index of platform. The simulation results show that DFG2A can achieve a balance between task running time and user requirements.