Dynamic Resource Allocation in an HPC Environment
Abstract
Analysts within the military strategic planning process need to be able to anticipate and respond in real-time to a dynamically changing battlespace with counter actions. The capability is required to accept current information into a simulation and rapidly peer into the future at any given moment to derive hypotheses about future alternatives. It is virtually impossible to identify or predict the specific details of what might transpire. Our research interest is to develop techniques to assess potential courses of action (COAs) against the adversarial environment. Utilizing HPC technology, multiple force structure simulations can be dynamically executed in parallel to concurrently evaluate the hypothesis of assessing a given COA against a range of adversarial eCOAs. The uncertainty of the adversary decision process requires simulation capabilities that spawn multiple simulations from critical decision points to evaluate alternative eCOAs. The focus of this paper is on the development of a simulation framework that provides the foundation for faster than real-time parallel COA simulations. Within this framework is a requirement to be able to dynamically allocate and reallocate resources in an HPC environment. The authors will describe techniques to clone and create variant simulations in real-time and the dynamic system requirements and use of HPC nodes.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2004
- Accession Number
- ADP023854
Entities
People
- Duane A. Gilmour
- Gary Blank
- James P. Hanna
Organizations
- Air Force Research Laboratory