Title :
Approximation algorithm for the temperature-aware scheduling problem
Author :
Zhang, Sushu ; Chatha, Karam S.
Author_Institution :
Arizona State Univ. Tempe, Tempe
Abstract :
The paper addresses the problem of performance optimization for a set of periodic tasks with discrete voltage/frequency states under thermal constraints. We prove that the problem is NP-hard, and present a pseudo-polynomial optimal algorithm and a fully polynomial time approximation technique (FPTAS) for the problem. The FPTAS technique is able to generate solutions in polynomial time that are guaranteed to be within a designer specified quality bound (QB) (say within 1% of the optimal). We evaluate our techniques by experimentation with multimedia and synthetic benchmarks mapped on the 70 nm CMOS technology processor. The experimental results demonstrate our techniques are able to match optimal solutions when QB is set at 5%, can generate solutions that arc quite close to optimal (< 5%) even when QB is set at higher values (50%), and executes in few seconds (with QB > 25%) for large task sets with 120 nodes (while the optimal solution takes several hundred seconds). We also analyze the effect of different thermal parameters, such as the initial temperature, the final temperature and the thermal resistance.
Keywords :
approximation theory; computational complexity; processor scheduling; CMOS technology processor; NP-hard problem; approximation algorithm; heat flux; periodic tasks; pseudo-polynomial optimal algorithm; temperature-aware scheduling problem; Approximation algorithms; CMOS process; CMOS technology; Frequency; Optimization; Polynomials; Scheduling algorithm; Temperature; Thermal resistance; Voltage;
Conference_Titel :
Computer-Aided Design, 2007. ICCAD 2007. IEEE/ACM International Conference on
Conference_Location :
San Jose, CA
Print_ISBN :
978-1-4244-1381-2
Electronic_ISBN :
1092-3152
DOI :
10.1109/ICCAD.2007.4397278