Home navigate_next Journals navigate_next ITCNC navigate_next Vol. 5, No. 3navigate_next An Efficient Evolutionary Approach for Task Scheduling in Supercomputing Environments
ICSES Transactions on Computer Networks and Communications
Vol. 5, No. 3, Sep. 2019


An Efficient Evolutionary Approach for Task Scheduling in Supercomputing Environments

Original Paper
1.3k
Visits
343
Downloads
a Sama College, IAU, Shoushtar Branch, Shoushtar, Iran

 

Highlights and Novelties
1- An new evolutionary approach based on the Genetic Algorithm is proposed to tackle static task-graph scheduling in homogeneous multiprocessor systems.

2- A novel chromosome encoding mechanism is used to model a complete scheduling via a string of gens. Whatever the mutation and cross-over use, the encoding mechanism produce feasible and valid schedules.

3- In addition, the utilization of task priority measurements as the background knowledge of the problem has made the proposed method very robust and efficient.

4- Different experiments on a set of not only random task graphs with different structural parameters but also task graphs of real-world applications reveal the superiority of the proposed approach.

 

Manuscript Abstract
Optimized task scheduling is key to achieve high performance in supercomputing environments e.g. in data centers that are providing variety of services for the community. In this paper, a novel and efficient evolutionary approach based on the Genetic Algorithm (GA) is proposed to tackle static task-graph scheduling in homogeneous multiprocessor systems, as the dominant infrastructure for supercomputing environments e.g. data centers. A novel chromosome encoding mechanism is used to model a complete scheduling via a string of gens. Whatever the mutation and cross-over use, the encoding mechanism produce feasible and valid schedules both from the sequencing the tasks and assigning them to the processors. In addition, the utilization of task priority measurements as the background knowledge of the problem has made the proposed method very robust and efficient. Different experiments on a set of not only random task graphs with different structural parameters but also task graphs of real-world applications reveal the superiority of the proposed approach in comparison with the state-of-the-art and traditional counterparts from the performance perspective.

 

Keywords
 Evolutionary Approaches   Genetic Algorithm   Task-Graph (or DAG) Scheduling   Parallel and Distributed Systems   Supercomputing Environments. 

 

Copyright and Licence
© Copyright was transferred to International Computer Science and Engineering Society (ICSES) by all the Authors. This manuscript is published in Open-Access manner based on the copyright licence of Creative Commons Attribution Non Commercial 4.0 International (CC BY-NC 4.0).

 

Cite this manuscript as
Hamid Reza Boveiri, "An Efficient Evolutionary Approach for Task Scheduling in Supercomputing Environments," ICSES Transactions on Computer Networks and Communications, vol. 5, no. 3, pp. 12-23, Sep. 2019.

 

For External Scientific Databeses
--BibTex-- --EndNote-- --Dublin--
star The old version of this page can be accessed via here, and is supported till 2020.
Purchase and Access

lock_open Open-Access

Bibliography

Manuscript ID: 304
Pages: 12-23
Submitted: 2019-08-21
Revised: 2019-09-12
Accepted: 2019-09-13
Published: 2019-09-30


Cited By (0)
Journal's Title
ITCNC Cover Page

Journal

ICSES Transactions on Computer Networks and Communications
ISSN: 2588-5847

ISSN: 2588-5847
Frequency: Quarterly
Accessability: Online - Open Access
Founded in: Mar. 2015
Publisher: ICSES
DOI Suffix: 10.31424/icses.itcnc