Data warehouse system includes many extract-transform-load(ETL) tasks which have some precedence constraint relations. In a multi-processor environment, how to efficiently schedule these ETL tasks is one of the important aspects for constructing data warehouse. On the basis of classification and summary of data warehouse ETL scheduling rules, a data warehouse ETL scheduling model is established to minimize the total ETL execution time. At the same time, based on the characteristics of the problem, adopting the same layer division strategy, a genetic algorithm based on the same layer division is proposed. Finally, the application case of the model is represented, and the case results show the feasibility and effectiveness of this model and its algorithm.