An efficient framework for ensemble of natural disaster simulations as a service

• Main Authors: Ujjwal KC, Saurabh Garg, James Hilton
• Format: Article
• Language: English
• Published: Elsevier 2020-09-01
• Series: Geoscience Frontiers
• Subjects: Wildfire prediction; Ensemble simulation; Cloud computing; Natural disaster models
• Online Access: http://www.sciencedirect.com/science/article/pii/S1674987120300438

Calculations of risk from natural disasters may require ensembles of hundreds of thousands of simulations to accurately quantify the complex relationships between the outcome of a disaster and its contributing factors. Such large ensembles cannot typically be run on a single computer due to the limited computational resources available. Cloud Computing offers an attractive alternative, with an almost unlimited capacity for computation, storage, and network bandwidth. However, there are no clear mechanisms that define how to implement these complex natural disaster ensembles on the Cloud with minimal time and resources. As such, this paper proposes a system framework with two phases of cost optimization to run the ensembles as a service over Cloud. The cost is minimized through efficient distribution of the simulations among the cost-efficient instances and intelligent choice of the instances based on pricing models. We validate the proposed framework using real Cloud environment with real wildfire ensemble scenarios under different user requirements. The experimental results give an edge to the proposed system over the bag-of-task type execution on the Clouds with less cost and better flexibility.