Simulation based optimisation on cloud

Abstract

Simulation-based optimisation (SBO) aims to obtain a set of optimal simulation parameters by combining the use of 2 steps into an iterative process – simulation modelling and optimisation. Multiple candidate simulations have to be ran in each iteration, and the outcomes are evaluated by the optimisation model, which then create new candidate simulations. The step repeats until the error from the simulations are within a pre-set threshold. This iterative process consumes huge amounts of computing power and time. In addition, most use cases for SBO only requires it to be executed periodically. This presents unique challenges for SBO. Cloud Computing have emerged in the recent decade and is gaining strength in both research and corporate organisations. Key benefits include elasticity of compute/storage resource and lower initial costs. These characteristics fit into the requirements of SBO. This project will analyse the use of different computational software for Simulation based Optimisation in the cloud environment. Multiple dimensions were identified for analysis, and experiments were ran to obtain quantitative results. Lastly, policies were determined based on the quantitative results to provide guidelines on the conditions to use these software systems in the Cloud.