GRASS: Generic Reactive Application-Specific Scheduling
High resource utilization is important to operate compute infrastructures and data centers efficiently. High utilization is achieved by multiplexing several applications over the same physical infrastructure. Yet, with this approach, the different requirements of each application have to be taken into account when scheduling resources.
We propose GRASS, a reactive domain-specific language that allows specifying application-tailored resource scheduling policies. We demonstrate how the declarative approach of GRASS enables extension and composition of scheduling policies. Our evaluation shows the performance benefits of considering application-specific information in a composition of scheduling policies that adapt at runtime.
Mon 21 OctDisplayed time zone: Beirut change
14:00 - 15:30
|GRASS: Generic Reactive Application-Specific Scheduling|
Vadim Zaytsev Raincode Labs, Belgium
|Multitier Reactive Programming in High Performance Computing|
Daniel Sokolowski Technische Universität Darmstadt, Philipp Martens Technische Universität Darmstadt, Guido Salvaneschi Technische Universität DarmstadtFile Attached