Compute mass of large-scale graph problems is an important part of modern data processing. Each social network or even Open Street Map with its huge road map handle problems based on graph structured data. The amount of data will increase every day. Something is needed to handle this amount of data. Adding new Hardware to increase the computational power isn’t the only option to encounter this problem. Optimization approaches will help to make existing algorithms more efficient. In some cases, the optimizations will not lead to a satisfying result. The reasons are often hard to find. Therefore, in this thesis there is a practical evaluation of different shortest path implementations on GraphX. The main goal is to show three different approaches and evaluate their influence on the computation time and the hardware usage. The evaluation is done in several graph environment so different real-world problems are reflected. Furthermore, this thesis provides an analysis of the evaluation results in direct comparison.