Scheduling of parallel processing systems based on algorithmic skeletons

摘要

We motivate the use of algorithmic skeletons for modeling parallel applications in the context of scheduling. Scalability and other features of the skeletal approach are used to improve the quality of scheduling parallel applications in multiprogramming systems. We show how skeletons can be used to model parallel applications that can perform dynamical remapping. This technique modifies schedules during runtime and increases system efficiency.