Throughput Maximization for Intel Desktop Platform under the Maximum Temperature Constraint

As processor power consumption continues to grow, thermal issue is becoming critical in the design of computing systems. While there have been extensive theoretical researches conducted on thermal aware computing, most of these researches are based on idealized theoretical models and, sometimes, unrealistic assumptions. In this paper, we develop a practical thermal aware scheduling algorithm to optimize its throughput under a given temperature constraint. We identify limitations of related theoretical work and implement a practical algorithm on an Intel i5 desktop computing running Linux 2.6.34. We also use SPEC2000 benchmark to validate the effectiveness of the proposed algorithm.