Production simulation using a distributed node-aware system

This study proposes a node-aware distributed simulation system for accelerating production simulation. To optimize simulation performance, this system proposes the performance-surveying module and the node-invoking module based on a MapReduce parallel framework to dynamically invoke valuable nodes during simulation. The performance-surveying module periodically records each node performance and ranks node capability before simulation. The node-invoking module minimizes overall simulation time by surveying current node performance and invoking the best available nodes. Experimental results indicate that using the Hadoop framework to distribute simulation jobs to six nodes reduces mean simulation time to 62.3% of that when a single node is used. Moreover, mean simulation time was reduced 73.6% when the proposed modules were also applied.