Allocating Tactical High-Performance Computer (HPC) Resources to Offloaded Computation in Battlefield Scenarios.

摘要

We envision a battlefield of the future where computation plays a pivotal role in providing the Army an advantage over its enemy. The ability for computationally intensive tasks to be carried out in real-time at the front lines will give Soldiers access to a wealth of information that they currently do not possess. This information could range from suggesting optimal locations from which to observe points of interest to locating snipers based on the shock waves of bullets. In order for these, and yet unknown applications, to be implemented a considerable computational capability has to be mobilized with an army into the battlefield. Mobile devices will soon become an integral part of a Soldier s toolkit as the military is attempting to introduce smartphones into combat. In order for the potential of these tools to be realized, their fundamental weakness of limited battery life has to be addressed. One approach to increasing energy efficiency of mobile devices is to offload computationally intensive applications to more capable machines. While a number of approaches to offloading have been proposed, none so far have addressed the issue of intelligently scheduling the offloaded applications to a cluster of heterogeneous machines. This paper describes the state-of-the-art in techniques necessary to implement such a scheduler and presents a plan for a project to implement such a scheduler.