Storing and Transferring Data of Parallel Agents to the Parent Process in Agent-Based Simulations.

Agent-based simulations are nowadays frequently used in complex systems to see how heterogeneous agents work together. Parallelization of agents can bring significant speed advantages on today’s processors with high number of threads, but it also brings several problems such as shared data access and inter-process communication. In this paper, we propose an integrated method that combines two separate processes: the process of updating the parent process by transferring the data generated by the agents working in child processes to the parent process, and the process of storing the data by transferring the data between processes through the database. We compare the proposed method with the sequential processing method, the shared memory method, which is a popular inter-process communication (IPC) method and the serial processing method under varying conditions. We observe that under increasing number of agents and data size, the proposed method runs in less time than the other methods.