Preview

PDF 1,003Kb

The accurate modeling of bushfires is not only complex and contextual but also a computationally intensive task. Ensemble predictions, involving several thousands to millions of simulations, can be required to capture and quantify the uncertain nature of bushfires. Moreover, users’ requirement and configuration may change in different situations requiring either more computational resources or modeling to be completed with a stricter time constraint. For example, during emergency situations, the user may need to make time-critical decisions that require the execution of bushfire-spread models within a deadline. Currently, most operational tools are not flexible and scalable enough to consider different users’ time requirements. In this paper, we propose the SparkCloud service, which integrates features of user-defined customizable configuration for bushfire simulations and scalability/elasticity features of the cloud to handle computation requirements. The proposed cloud service utilizes Data61’s Spark, which is a significantly flexible and scalable software system for bushfire-spread prediction and has been used in practical scenarios. The effectiveness of the SparkCloud service is demonstrated using real cases of bushfires and on real cloud computing infrastructure.

Item Type:	Refereed Article
Keywords:	bush fire, cloud computing, ensemble predictions, deadline-based resource allocation
Research Division:	Information and Computing Sciences
Research Group:	Distributed Computing
Research Field:	Distributed and Grid Systems
Objective Division:	Information and Communication Services
Objective Group:	Computer Software and Services
Objective Field:	Application Tools and System Utilities
UTAS Author:	Garg, S (Dr Saurabh Garg)
UTAS Author:	Forbes-Smith, N (Mr Nicholas Forbes-Smith)
ID Code:	125155
Year Published:	2018
Web of Science® Times Cited:	1
Deposited By:	Information and Communication Technology
Deposited On:	2018-04-03
Last Modified:	2019-02-25
Downloads:	40 View Download Statistics