MapReduce Based Experimental Frame for Parallel and Distributed Simulation Using Hadoop Platform

Simulation-based experiment of complex systems is a time consuming-job. Parallel and distributed simulation is one of the methods to reduce the simulation time. To simulate and analyze the system with this method, it is required to design a suitable experimental frame. Therefore, this paper proposes a MapReduce based experimental frame for the parallel and distributed simulation. Because Hadoop MapReduce is the most widely used parallel and distributed computing platform, we use it to design the experimental frame. In our work, the ‘map’ of MapReduce automatically generates and simulates the system, and the ‘reduce’ of MapReduce collects and analyzes the result. We can reuse the existing large scale Hadoop clusters without any modification of the platform, so it is easy to set-up and use the experimental frame. This paper presents an air defense simulation to show the usage and speed up with a 16-node Hadoop cluster.