Abstract: With the continuous development of cloud computing, task scheduling problem become a crucial aspect. How to deal with tasks quickly, not only meet the needs of users, but also to achieve load balancing and make the completion time, cost to achieve relatively optimal.By comparing and analyzing the existing task scheduling algorithms.we proposed the genetic and difference algorithm policy to solve the problem. we employs population updating scheme to improve the robustness of the algorithm.Using secondary mutation strategy to improve population diversity, accelerate the convergence speed. Adding difference operator to improve the algorithm's local search ability.The performance is analyzed using Cloudsim simulator and compared with existing GA, Min-min algorithm. Simulation results demonstrate that the proposed algorithm has better performance in load balancing, finish time and costs.