EUVIMEDEUVIMED entdecken
Crossref

Energy-efficient DAG scheduling based on duplication and slack-aware DVFS in heterogeneous fog–edge–cloud environments

Zhihao Peng, Behnam Barzegar, Parmida Tavakoli, Milad Akbarnia Mansor, Poria Pirozmand

The Journal of Supercomputing · 07.08.2026 · 82 (12)

DOI: 10.1007/s11227-026-08756-3 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Journal of Supercomputing, online erschienen am 07.08.2026. Verfasst von Zhihao Peng und 4 weiteren Autor:innen. Die Arbeit nennt 55 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This paper presents a multi-objective scheduling framework, ECD-DVFS (Energy and Cost-Efficient Duplication-Based Dynamic Voltage and Frequency Scaling Scheduler), for scientific workflows in hybrid cloud environments. The main challenge is to balance several conflicting objectives, including total completion time, energy consumption, reliability, cost, and Quality of Service (QoS). Critical path-based ranking, dependency-aware clustering, selective task duplication, and dynamic voltage and frequency scaling (DVFS) are the four coordinated parts of the method. The upward ranking of jobs on the critical route, taking into account both computing costs and communication delays, is the first method used to identify the delay-sensitive tasks. Second, in order to minimize inter-node data transfers and communication overhead, we cluster highly dependent parent–child activities using dependency awareness. In order to improve reliability, critical tasks are then selectively replicated to separate nodes while controlling the additional energy overhead resulting from replication. Meanwhile, non-critical tasks use a slack-aware DVFS algorithm to reduce power consumption without increasing the system’s makespan. Simulation results on common scientific workflows (Montage, CyberShake, Epigenomics, and LIGO) showed that the ECD-DVFS framework reduced the average makespan by about 22.2% and the average energy consumption by about 28.8% compared with the HEFT algorithm. Moreover, the average task failure rate was reduced by about 58.5% compared with HEFT, SLA violations reduced from 14.5 to 5.8%, and the average total workflow execution cost was reduced by approximately 22%. Overall, the results show that ECD-DVFS provides a balanced trade-off among performance, energy efficiency, reliability, cost, and QoS through adaptive and energy-aware workflow scheduling in hybrid fog–edge–cloud environments.

Zitieren

APA

Peng, Z., Barzegar, B., Tavakoli, P., Mansor, M. A., & Pirozmand, P. (2026). Energy-efficient DAG scheduling based on duplication and slack-aware DVFS in heterogeneous fog–edge–cloud environments. The Journal of Supercomputing, 82(12). https://doi.org/10.1007/s11227-026-08756-3

Vancouver

Peng Z, Barzegar B, Tavakoli P, Mansor MA, Pirozmand P. Energy-efficient DAG scheduling based on duplication and slack-aware DVFS in heterogeneous fog–edge–cloud environments. The Journal of Supercomputing. 2026;82(12). doi:10.1007/s11227-026-08756-3

BibTeX
@article{Peng2026,
  author = {Peng, Zhihao and Barzegar, Behnam and Tavakoli, Parmida and Mansor, Milad Akbarnia and Pirozmand, Poria},
  title = {Energy-efficient DAG scheduling based on duplication and slack-aware DVFS in heterogeneous fog–edge–cloud environments},
  journal = {The Journal of Supercomputing},
  year = {2026},
  volume = {82},
  number = {12},
  doi = {10.1007/s11227-026-08756-3}
}
Kostenlos starten →