Comparing conventional and distributed approaches to simulation in a complex supply-chain health system

Abstract

Decision making in modern supply chains can be extremely daunting due to their complex nature. Discrete-event simulation is a technique that can support decision making by providing what-if analysis and evaluation of quantitative data. However, modelling supply chain systems can result in massively large and complicated models that can take a very long time to run even with today's powerful desktop computers. Distributed simulation has been suggested as a possible solution to this problem, by enabling the use of multiple computers to run models. To investigate this claim, this paper presents experiences in implementing a simulation model with a ‘conventional’ approach and with a distributed approach. This study takes place in a healthcare setting, the supply chain of blood from donor to recipient. The study compares conventional and distributed model execution times of a supply chain model simulated in the simulation package Simul8. The results show that the execution time of the conventional approach increases almost linearly with the size of the system and also the simulation run period. However, the distributed approach to this problem follows a more linear distribution of the execution time in terms of system size and run time and appears to offer a practical alternative. On the basis of this, the paper concludes that distributed simulation can be successfully applied in certain situations.

Keywords:

• distributed simulation
• supply chain systems
• healthcare operations
• simulation software
• Simul8

Acknowledgements

The authors thank the following people for their time and their help with data and model validation: Crispin Wickenden and Andrew Oliver from the NBS; the employees of the Southampton PTI Centre, particularly Mike Northcott; Tracey Lofting from Southampton General Hospital; Rob Hick and Judith Chapman from the Blood Stocks Management Scheme. The authors would also like to thank Dr Mark Elder (founder and CEO of Simul8 Corporation) for providing the Simul8 licences and his generous ongoing support. Finally, the authors would like to acknowledge the contribution of Dr Allan Tucker, John Saville and Dr Steven Swift from Brunel University who generously offered assistance during the experimentation process.