TY - GEN
T1 - Usage history-directed power management for smartphones
AU - Li, Xianfeng
AU - Wen, Wen
AU - Wang, Xigui
N1 - Publisher Copyright:
© Springer International Publishing Switzerland 2015.
PY - 2015
Y1 - 2015
N2 - Smartphones are now equipped with powerful application processors to meet the requirement of performance demanding apps. This often leads to over-provisioned hardware, which drains the battery quickly. To save energy for battery-powered smartphones, it is necessary to let the processor run at a power-saving state without sacrificing user experience. Android has implemented a set of CPU governors by leveraging dynamic voltage and frequency scaling (DVFS) according to the computational requirements of apps. However, they commonly adopt very conservative policies due to limited information, therefore leaving considerable energy reduction opportunities unexplored; or they are not responsive to user interactions, leading to poor user experiences. In this work, we start from an important observation on the repetitive patterns of smartphone usage for each individual user, and propose UH-DVFS, a usage history-directed DVFS framework leveraging on this observation. UH-DVFS can identify repetitive user transactions, and store their execution history information within a table. When such a user transaction is launched again, the table is consulted for an appropriate CPU frequency adaptation to reduce the energy consumption of the user transaction without sacrificing user experience. We have implemented the proposed framework on Android smartphones, and have tested it with real-world interaction-intensive apps. The results show that our framework can save energy from 10% to 36% without affecting the quality of user experiences.
AB - Smartphones are now equipped with powerful application processors to meet the requirement of performance demanding apps. This often leads to over-provisioned hardware, which drains the battery quickly. To save energy for battery-powered smartphones, it is necessary to let the processor run at a power-saving state without sacrificing user experience. Android has implemented a set of CPU governors by leveraging dynamic voltage and frequency scaling (DVFS) according to the computational requirements of apps. However, they commonly adopt very conservative policies due to limited information, therefore leaving considerable energy reduction opportunities unexplored; or they are not responsive to user interactions, leading to poor user experiences. In this work, we start from an important observation on the repetitive patterns of smartphone usage for each individual user, and propose UH-DVFS, a usage history-directed DVFS framework leveraging on this observation. UH-DVFS can identify repetitive user transactions, and store their execution history information within a table. When such a user transaction is launched again, the table is consulted for an appropriate CPU frequency adaptation to reduce the energy consumption of the user transaction without sacrificing user experience. We have implemented the proposed framework on Android smartphones, and have tested it with real-world interaction-intensive apps. The results show that our framework can save energy from 10% to 36% without affecting the quality of user experiences.
KW - Android
KW - DVFS
KW - Energy reduction
KW - Smartphone
KW - User transaction
UR - http://www.scopus.com/inward/record.url?scp=84959350994&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-27119-4_20
DO - 10.1007/978-3-319-27119-4_20
M3 - Conference contribution
AN - SCOPUS:84959350994
SN - 9783319271187
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 288
EP - 302
BT - Algorithms and Architectures for Parallel Processing - 15th International Conference, ICA3PP 2015, Proceedings
A2 - Perez, Gregorio Martinez
A2 - Zomaya, Albert
A2 - Wang, Guojun
A2 - Li, Kenli
PB - Springer Verlag
T2 - 15th International Conference on Algorithms and Architectures for Parallel Processing, ICA3PP 2015
Y2 - 18 November 2015 through 20 November 2015
ER -
