Pushing down user information to enhance smart device system design

Recent popularity in smart devices, for example smartphones, has created an increased interest in carrying small devices with limited battery capacity. Unfortunately, smartphones are notorious for consuming energy far too quickly. Although certain advances have been made on the hardware side such as better batteries, the first part of the talk will be focused on improving energy management software in the lower layer system to make better use of existing batteries. We aim to save smartphone radio energy with a network traffic aware approach. More specifically, we exploit application priority with machine learning for energy savings; we also exploit delay tolerant time periods within high priority applications, such as real-time applications, for energy savings. The second part of the talk will be focused on a learning-based approach to analyze disk I/O traffic, rather than network I/O traffic, from users to optimize lower layer system design for both smartphone energy savings and delay reduction.