A Predictive Task Migration Algorithm with Adaptive Migration Threshold for Dynamic Thermal Management of Multi-Core Processors

Authors
Bagher Salami
Mohammadreza Baharani
Hamid Noori
Abstract
Request for more computation power steadily forces designers to provide more powerful processors by using more number of cores on a single chip. The increasing complexity of processors leads to higher integration density, power density and temperature. For avoiding thermal emergencies, various dynamic thermal management techniques have been presented. In this paper, we present a task migration algorithm with novel online self-adjusting migration threshold schema for better dynamic thermal management to minimize both average and peak temperature with very low performance overhead. Our proposed algorithm adjusts migration threshold according to work-load and hardware platforms. The experimental results indicate that our technique can significantly decrease the average and peak temperature in most cases compared to Linux standard scheduler, and two well-known thermal management techniques: PDTM, and TAS.
Keywords
Dynamic Thermal Management
Multi-Core Processors
Migration Threshold
DVFS (Dynamic Voltage Frequency Scaling)
Task Migration
The CSI Journal on Computer Science and Engineering
Volume 9, Issue 2
January 2012