JCSE, vol. 12, no. 4, pp.157-169, 2018

DOI: http://dx.doi.org/10.5626/JCSE.2018.12.4.157

User-Centric Thermal Management for Smartphones

Wook Song and Jihong Kim
Samsung Research, Samsung Electronics, Seoul, Korea Department of Computer Science and Engineering, Seoul National University, Seoul, Korea

Abstract: For high-performance smartphones, keeping the on-chip temperature under a given critical temperature is a major concern. In order to prevent the temperature from rising above the critical point, modern smartphones widely adopt the dynamic thermal management (DTM) scheme, which limits the maximum CPU frequency when the CPU reaches high temperatures (thus making the CPU temperature drop). In this paper, we propose a novel DTM scheme based on user-perceived response time analysis called SmartDTM. Unlike existing DTM schemes that can significantly degrade the quality of user experience, SmartDTM explicitly accounts for the quality of the user experience in making DTM decisions. We divide an execution of a given user-interactive session into two intervals, one where the system response time directly affects the user experience and the other where the system response time does not affect the user experience. In the user-perceived response time interval, our proposed scheme conservatively makes DTM decisions so that the quality of the user experience is not affected by the reduced maximum CPU frequency. On the other hand, in the user-oblivious response time interval, SmartDTM aggressively lowers the CPU frequency so that the CPU temperature can be quickly decreased to a safe level without negatively affecting user experience. Our experimental results on an ODROID-XU+E board show that SmartDTM can improve the performances of user-perceived intervals by 12.2% and 21.4% over the Android's default DTM policy when the initial temperature was set to 65??and 70?? respectively, under the critical temperature of 85??

Keyword: Smartphone; Operating system; User-centric optimization; Thermal management; Dynamic voltage and frequency scaling

Full Paper:   238 Downloads, 1618 View