Abstract | ||
---|---|---|
In modern processors, energy savings are achieved using dynamic voltage and frequency scaling (DVFS). For task scheduling, where a task graph representing a program is allocated and ordered on multiple processors, DVFS has been employed to reduce the energy consumption of the generated schedules, hence running the processors at heterogeneous speeds A prominent class of energy-efficient scheduling algorithms is slack reclamation algorithms, which try to use idle times (slack) to slow down processor speed to save energy. Several algorithms have been proposed and under the assumed system model they can achieve considerable energy savings. However, the question arises, how realistic and accurate these algorithms and models are when implemented and executed on real hardware. Can one achieve the promised energy savings? This paper proposes a methodology to investigate these questions and performs a first experimental evaluation of selected slack reclamation algorithms. Using schedules created by three scheduling algorithms for a set of task graphs, we generate code and execute it on a small parallel system. We measure the power consumption and compare the results between the algorithms and relate them to the expected values. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1109/ICPPW.2016.48 | 2016 45th International Conference on Parallel Processing Workshops (ICPPW) |
Keywords | Field | DocType |
task scheduling,energy savings,slack reclamation,DVFS | Fixed-priority pre-emptive scheduling,Fair-share scheduling,Computer science,Parallel computing,Two-level scheduling,Least slack time scheduling,Rate-monotonic scheduling,Frequency scaling,Computer hardware,Dynamic priority scheduling,Energy consumption,Distributed computing | Conference |
ISSN | ISBN | Citations |
1530-2016 | 978-1-5090-2826-9 | 0 |
PageRank | References | Authors |
0.34 | 15 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Andrew Zaliwski | 1 | 0 | 0.34 |
Stefan Lankes | 2 | 152 | 26.39 |
Oliver Sinnen | 3 | 382 | 38.71 |