An apparatus, system, and method are disclosed for generating a resource utilization description for a parallel data processing system. The method comprises developing a system model using the data flow graph for a target data processing job. The system model is then tuned based on a sample data set. The tuned system model is utilized to calculate the resource utilization description for the target data processing job. The resource utilization description may include disk space usage, scratch space usage, and processor utilization for each partition in the system. The method may further include determining a more efficient data flow graph to achieve the same results of the target data processing job.