Abstract: Cloud task scheduling is a hotspot of current research, in order to shorten the task completion time on the premise of a given user satisfaction. Mathematical model of Cloud task scheduling with user satisfaction constraint is established. And an improved simulated annealing cloud task scheduling algorithm with multi-dimension QoS constraints is proposed to solve the model in the paper.The task completion time as the goal, time greedy strategy of multi-dimension QoS constraints is introduced to produce initial solution. The algorithm adopt simple simulated annealing process to search the best scheme under the user's QoS constraints.Simulation experiments show that the scheduling policy to ensure customer satisfaction and shorten task completion time. It is a kind of effective scheduling strategy that consider both users and providers of cloud services.