Cloud computing is a new business model based on Internet and aims to provide information service for users. Resource scheduling is one of the key technologies of Cloud Computing. Analyzing the representative achievement of cloud computing resources scheduling, and then according to the problem that there are large gap between most algorithm task design and actual service needs, the authors present a task scheduling algorithm on the basis of multi-QoS constrains and genetic algorithm. It supports different users choose different scheduling goals according to their own needs. By doing experiment on cloud computing simulation platform called CloudSim, the results shows that this algorithm can satisfy the QoS constrains, at the same time, it can guarantee system load and improve the efficiency of the task scheduling.