Volume 17, Issue 3 (JIAEEE Vol.17 No.3 2020)
Journal of Iranian Association of Electrical and Electronics Engineers 2020, 17(3): 129-140 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Niknia F, Hakami V. Optimization of Online Task Assignment in Multi-Core Processing Systems for Dynamic Thermal Management. Journal of Iranian Association of Electrical and Electronics Engineers 2020; 17 (3) :129-140
URL: http://jiaeee.com/article-1-844-en.html
URL: http://jiaeee.com/article-1-844-en.html
Iran University of Science and Technology
Abstract: (2288 Views)
Increase in chip temperature causes more power consumption in multi core processors and decreases the CPU lifetime. The optimization of online task assignment to processing cores is an efficient approach to control chip-wide temperature distribution. However, task assignment faces some uncertainties in the system (including: stochastic task arrival, random task pairings, and time-varying thermal profile variations). In this paper, an online task to core assignment approach is presented which uses Semi-Markov Decision Process to prevent performance reduction and considers randomness and uncertainties in system. As the transitional properties are not accessible and due to high dimension of system state components, the proposed approach uses function approximation to approximate action values in any system state. The simulation results show 6 centigrade decrease in system average peak temperature and 66 milliseconds decreases in task service time.
Keywords: Multi-core Processors, Online Task Assignment, Thermal Management and Reinforcement Learning
Type of Article: Research |
Subject:
Electronic
Received: 2019/03/31 | Accepted: 2020/06/13 | Published: 2020/09/28
Received: 2019/03/31 | Accepted: 2020/06/13 | Published: 2020/09/28
Send email to the article author
| Rights and permissions | |
 |
This Journal is an open access Journal Licensed under the Creative Commons Attribution-NonCommercial 4.0 International License. (CC BY NC 4.0) |
