Embodiments of the invention may be used to generate a benchmark profile for a computing job configured to execute on distributed systems. The benchmark profile may be used to predict the performance of components of a computing job for a variety of different distributed computing system architectures. A profiling tool evaluates the computing job to identify the particular performance characteristics of the application and match this with benchmarks that are most representative of the identified performance characteristics and store them in the benchmark profile. The identified benchmarks may then be run on different configurations of a distributed computing system in order to predict the performance of the application for a variety of scenarios.