Mr Lance Holden1, Dr Antonio Giardina2, Mr Denis Shine1
1DSTG, Edinburgh, Australia,
2Deakin University, Melbourne, Australia
When the Defence Science Technology Group (DSTG) analyse close combat via simulations, they are required to generate a large amount of interaction data from many small processing tasks. This data is used to execute simulations with enough replications to provide a large enough sample set from where to gain statistically significant simulation insights. Initially DSTG had a slow, manually controlled execution pipeline for all required processing and simulation tools across ad-hoc and poorly controlled computing resources. The tools used often lacked any built in support for distribution and DSTG managed this process manually.
A collaboration with Deakin University produced a new software system to manage the distribution of these tasks across a fixed network of known processing capability. This new system, called Virgil, used a mix of available open source components and custom scheduling modules. Experience in the use of Virgil by DSTG exposed further areas where improvements could be made, such as the need of better methods for creating dependencies between task executions and the need to easily configure the required processing/simulation tools across all remote clients.
A new collaboration between DSTG and Deakin University has started to explore creating a task specification language and a reconfigurable network environment. These new features will provide a scalable and adjustable distributed processing platform that will improve the speed and reliability of data generation and simulation execution.
Mr Denis Shine is a researcher working for the Defence Science and Technology Group. He specialises in the application of Land Combat Simulation to support Army decision making.