DEXSim: an experimental environment for distributed execution of replicated simulators using a concept of single simulation multiple scenarios

Abstract

This paper presents an efficient and scalable experimental environment for distributed execution of replicated simulators. By taking a performance-centered approach, the proposed technique makes the best use of distributed hardware resources for faster data collection. Accordingly, the primary contribution of this work is to describe how the environment improves scalability and utilizes distributed hardware resources efficiently. To do this, we suggest a new concept of single simulation multiple scenarios and propose a distributed execution simulation framework regarding the following three aspects: (1) layered architecture model design;(2) protocol definitions interacting with them; and (3) framework implementation. The proposed model architecture and protocol definitions guarantee a straightforward structural scalability and an efficient load-balanced utilization between hardware resources. Moreover, the framework operates simulation execution automatically without users' extra work. In order to prove the efficiency of the proposed framework, we performed three extensive experiments with different models, that is, different systems. The experimental results show that simulation performance increases proportionally with the number of hardware resources, minimizing the overhead of the proposed framework's utilization.

Distributed simulation: state-of-the-art and potential for operational research

“…Experiences from this formed the basis for SAKERGRID (Kite, et al, 2011), a desktop grid and computing cluster system in use at Saker Solutions and Sellafield PLC, a cluster-based high performance simulation system in use in the Ford Motor Company and a desktop grid that was used for simulations of biochemical pathways in cancer (Liu et al, 2014). Choi, Seo, and Kim (2014) also developed a similar system for use with dedicated computing clusters. In terms of cloud-based systems, the JADES platform was adapted to run agent-based simulations in parallel on cloud resources (Rak, Cuomo, & Villano, 2012) and the CloudSME Simulation Platform has been used to run simulation experiments over multiple clouds .…”

High Speed Simulation Analytics

“…More recent ones essentially use the same techniques but instead of fixed computing resources these use virtualised ones made available on a cloud. Examples of both of these include: the WINGRID desktop grid system that was used to speed up credit risk simulations in a well-known European bank (Mustafee & Taylor, 2009), SakerGrid, a desktop grid and computing cluster system in use today at Saker Solutions and Sellafield PLC (Kite, et al, 2011), a cluster-based high performance simulation system in use in the Ford Motor Company, a desktop grid that was used for simulations of biochemical pathways in cancer (Liu et al, 2014), and a cluster computing based grid used for a similar application (Choi, Seo, and Kim (2014)). Examples of cloudbased systems include an adaptation of the JADES platform to run agent-based simulations in parallel on cloud resources (Rak, Cuomo, & Villano, 2012) and the CloudSME Simulation Platform is used to run simulation experiments over multiple clouds (S.J.E.…”

A simulation as a service cloud middleware

“…It is founded on a Java-based architecture and is designed to run multiple concurrent simulations while automatically acquiring resources from an ad hoc federation of cloud providers. DEXSim [15] is a distributed execution framework for replicated simulations that provides two-level parallelism, i.e., at CPU core-level and at system-level. This organization delivers better performance to their system.…”