4.5 Article

Slack extender mechanism for greening dependent-tasks scheduling on DVFS-enabled computing platforms

Journal

JOURNAL OF SUPERCOMPUTING
Volume 78, Issue 5, Pages 6275-6295

Publisher

SPRINGER
DOI: 10.1007/s11227-021-04112-9

Keywords

Dynamic voltage; frequency scaling; High-performance computing; Dependent-task scheduling; Greening computing; Energy aware task scheduling

Ask authors/readers for more resources

This paper proposes a mechanism called BlackLight to extend the slack of tasks in schedules produced by dependent-task scheduling algorithms, aiming to reduce energy consumption.
The task's slack is the key issue to reduce the energy consumed by DVFS-enabled computing platforms. Despite the large number of scheduling algorithms that are presented in the literature, only a unique scaling axiomatic approach (SAA) is utilized in the scaling phase of the algorithms. SAA simply extends the execution of the task within its slack if a suitable scaling frequency is available. Unfortunately, when dependent-tasks applications are scheduled on such platforms, scheduling algorithms minimize the tasks' slacks to reduce the overall completion time of the application tasks. This paper presents a mechanism that can be applied to any schedule produced by a dependent-task scheduling algorithm for both homogeneous and heterogeneous DVFS-enabled computing platforms. The proposed mechanism is called BlackLight. BlackLight attempts to extend the tasks' slacks via rescheduling the application tasks without violating the overall completion time of the application tasks. The proposed mechanism is applied to a large number of dependent-tasks schedules of both random generated application graphs and two real-world application graphs. The experimental results based on a computer simulation show that the proposed mechanism significantly extends the tasks' slacks compared with SAA , which leads to more reduction in the consumed energy.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.5
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

No Data Available
No Data Available