A Comparative Analysis and Evaluation of MapReduce Cloud Computing Simulators

The application of MapReduce cloud computing simulators for research and development is becoming popular, due to their efficiency and ease of utilization. This ignited the development of several cloud simulators for algorithm testing and performance analysis of dynamic MapReduce environments. The selection of appropriate simulator for a specific research remains a challenge. We have designed a MapReduce classification framework to guide cloud and big data researchers towards suitable tools. We have reviewed eleven MapReduce specific simulators. Our evaluation first revealed thirty general functional requirements for more widely applicable cloud simulators. Then, we focused on specific concerns of MapReduce related simulations and filtered the general requirements down to the most relevant thirteen. Our evaluation highlighted the strengths and weakness of several MapReduce simulators. IoT-based applications, stream processing and replaying of production cluster workloads are key criteria absent from many simulators. Therefore, we identified these as gaps, that simulator developers could focus on when extending their works towards MapReduce oriented simulations. Finally, researchers simulating dynamic behaviors of Hadoop clusters should select simulators efficient in parameter tuning.