Search trees in Constraint Programming are a fertile ground for parallelization. However, it is difficult to propose a global parallelization schema, due to the great Constraint Satisfaction Problems (CSPs) variety and the plethora of the serial search methods that are available to solve a CSP. In this work, we exploit a serial search methods framework to make an arbitrary search method parallel by simulating its serial execution. We record the visited search tree parts and then try to restore them on different Mappers-workers in a MapReduce instal- lation. The success of this approach depends on how equal the recorded search tree parts are. We evaluate our proposal by solving N Queens and Number Parti- tioning problem instances on a Hadoop system installed on a 64-core cloud infrastructure.