Graph databases have become an indispensable tool for the analysis of linked data and interrelated data. As with any graph data representation, the need for using graph database systems emerge when they increase in size and complexity. Affiliated to those needs, graph database benchmarks emerge to assess the performance of such systems in application scenarios, representative of use cases. I proposed graph database benchmarks based on the idea of social network application. The benchmark implements and proposes a data generator that synthetically generate graphs. Also a set of high level application queries on this application that model parts of the behavior of social network users. I also studied the graph database benchmarks results that affects the performance of graph database systems. I proposed graph database benchmarks in terms of data model, query workload, large datasets and usage scenarios. I discussed the characteristics of Neo4j and DEX graph databases to be included in the benchmarks. Also I studied, the characteristics of these graph databases queries that are important in the application of graph analysis.