آخرین دستاوردهای علمی و پتانسیل برای تحقیق در عملیات
Abstract
1- Introduction
2- An overview of distributed simulation
3- Distributed simulation approaches
4- Distributed simulation technologies
5- Distributed simulation applications
6- The potential impact of distributed simulation on operational research
7- Conclusions
References
Abstract
In Operational Research conventional simulation practices typically focus on the conceptualization, development and use of a single model simulated on a single computer by a single analyst. Since the late 1970s the field of Distributed Simulation has led research into how to speed up simulation and how to compose large-scale simulations consisting of many reusable models running using distributed computers. There have been significant advances in the theories and technologies underpinning Distributed Simulation and there have been major successes in defence, computer systems design and smart urban environments. However, from an Operational Research perspective, Distributed Simulation has had little impact on mainstream research and practice. To argue the potential benefits of Distributed Simulation for Operational Research, this article gives an overview of Distributed Simulation approaches and technologies as well as discussing the state-of-the-art of Distributed Simulation applications. It will investigate the potential advantages of Distributed Simulation for Operational Research and present a possible sustainable future, based on experiences from e-Science, that will help Operational Research meet future challenges such as those emerging from Big Data Analytics, Cyber-physical systems, Industry 4.0, Digital Twins and Smart environments.
Introduction
Distributed Simulation (DS) is a field with roots in Computer Science, especially parallel and distributed computing, and has contributed to major successes in the simulation of large systems in defence, computer systems design and smart urban environments. As noted by Fujimoto (1990; 2016) the field emerged from two communities. In the 1970s, the Parallel Discrete Event Simulation (PDES) community investigated how to speed up simulations using multiple processors in high performance computing systems. Later, in the 1980s, principally led by efforts in the defence sector to enable simulation reuse, the DS community used PDES techniques to interconnect simulations together running on distributed computers connected by a network. Today, researchers active in these areas informally call the field Parallel and Distributed Simulation (PADS). However, as the wider simulation community often refers to this as just DS, we shall therefore use this term in this review. The main goals of DS are to use parallel and distributed computing techniques and multiple computers to speed up the execution of a simulation program and/or to link together simulations to support reusability (Fujimoto, 2000). Some authors have also used DS to refer to approaches that run simulation experiments and/or replications on distributed computers in parallel with the goal of reducing the time taken to analyse a system (Heidelberger, 1986). Following the “modes” of simulation introduced by Robinson (2002), Fig. 1 shows these three main modes of DS: Mode A to speed up a single simulation, Mode B to link together and reuse several simulations, and Mode C to speed up simulation experimentation. In Mode A, a simulation that might be simulated on a single computer is subdivided on some basis into separate simulations that are run on different computers interacting via a communications network – the possibility of speed up arises from the parallel execution of the separate simulations. In Mode B, several simulations running on different computers are linked together to form a single simulation again with interactions between models carried out via a communications network–larger models beyond the capability of a single computer can be created and simulations can be reused by connecting them to other simulations (so potentially reducing the cost of developing new simulations).