Faucets: On-Demand Resource Allocation on Computational Grids
Award year: 2005-2006
The focused area of this research is the scheduling of resources on a computational grid in an on-demand fashion such that work can be completed by a guaranteed deadline time. This approach contrasts traditional scheduling methods that typically involve executing work in a first-in first-out manner. The benefit of this approach is that grid computing resources, such of those that make up the national TeraGrid environment, can be scheduled in a much more efficient manner, improving resource utilization. The work proposed here represents a collaborative effort between the Parallel Programming Laboratory, led by Professor Laxmikant Kale in the Department of Computer Science, and William Yurcik's Security Research Group at NCSA.
The project proposed here integrates Parallel Programming Laboratory technologies, such as the Charm++ and Adaptive MPI runtime systems and the Faucets grid computing framework, with technologies from Yurcik's group, such as the NVisionCC cluster security monitoring software. By leveraging the strengths of both groups, we expect to create a secure and reliable solution for on-demand computing. Users of this system will be able to submit jobs with a variety of criteria related to desired job completion time, number of processors required, processor architecture, etc. and the system will match the job request to the computational resources that can most effectively complete the specified job.