Simulating rare events is extremely difficulty and requires massive computational resources and complex data processing workflow, which is determined by the nature of stochastic systems. To help computational scientists discover hard scientific problems in this area, we built a large-scale molecular dynamics simulation framework integrated with forward flux sampling (FFS) technique on Hadoop ecosystem. In this project, we port the customized FFS workflow to underlying MapReduce-based computing pipeline by using dataflow-driven design pattern and Gromacs application. The early works show that our framework is able to provide a scalable, fault-tolerance and efficient rare events simulation environment over varieties of computing infrastructures, while preserving the flexibility of the original scientific application.