The research direction of the University of Sciences and Humanities has integrated a Big Data architecture to make predictive models with large volumes of data. Therefore it was implemented with the purpose that in future research, this architecture can be used efficiently. In this study, the theoretical concepts of Hadoop version 2.0 will be discussed, as well as the next scalability in a Beowulf cluster implemented in one of the University's laboratories and the configuration of Hadoop Spark and how they were able to work in conjunction. Finally, in the results section, tests will be carried out to validate that this architecture works perfectly.