Optimizing the performance of Hadoop can greatly improve the efficiency of Map Reduce,reduce the execution time of tasks and increase the throughput of distributed computing.The implementation performance of Hadoop is impacted by a list of configuration properties which are interdependent.Based on genetic algorithm,the configuration schemes are updated by constantly selection,crossover,mutation,then approximate optimal configuration scheme is produced after several iterations.Experimental results show that this configuration scheme can effectively improve the overall performance of system and has stability and applicability.