Analysis of high performance communication and computation solutions for parallel and distributed simulation

Abstract

This paper illustrates the definition and analysis of a collection of solutions adopted to increase the performance of communication and computation activities required by the implementation and execution of parallel and distributed simulation processes. Parallel and distributed simulation has been defined, and a real testbed simulation scenario has been illustrated, based on the ARTIS simulation framework. Three classes of solutions have been proposed to improve the performance of simulations executed over commodity off-the-shelf computation and communication architectures: multi-threaded software and Hyper-Threading support by the processor architectures, data marshalling solutions for shared-memory and network-based communications, and data structure optimization for simulation events' management. All the proposed solutions have been evaluated on a testbed evaluation scenario, under variable configurations. Results obtained demonstrate that a performance improvement can be obtained by adopting and tuning the proposed solutions. © Springer-Verlag Berlin Heidelberg 2005.