With the popularity of cloud computing, a large number of data processing is completed by cloud services. Current algorithms seldom consider the different computing capability of the virtual machine in the heterogeneous system, so the waiting time is too long. We propose a scheduling algorithm which adapts to the size of real-time virtual machine load. Given the virtualization of cloud computing resources, we propose a method for evaluating the computing capability of virtual machines. Based on the virtual machine capability and the changes of the process, the adaptive task size is adjusted to meet the real-time requirements. Task completion time is unified and the load balancing of the virtual machine is maintained so the total execution time is shortened and the system throughput and the overall service capability are improved, hence the benefit is improved. Experimental results show that the algorithm can adaptively adjust the size of tasks according to the operating conditions, and maintain the load balance of the virtual machine.