Revisiting 802.11 power consumption modeling in smartphones

WiFi activity is a major source of power consumption in today's smartphones. Consequently, accurately WiFi power consumption models are extremely useful for researchers and app developers. Among a large number of models proposed recently, a model introduced by Serrano et al. was the first to add a new component - a per-frame energy toll incurred as a frame traverses the protocol stack - to the power consumption of the wireless NIC. The authors called this new component cross-factor and validated the accuracy of the model on a large number of devices, mostly 802.11g wireless routers and APs. This paper examines the validity of the model introduced by Serrano et al. on today's smartphones. We try to answer two questions: (i) Can the model accurately estimate the power consumption due to WiFi activity in today's smartphones given the complexity of modern smartphone architectures? (ii) Does the model remain valid in the case of 802.11n/ac interfaces, and if yes, can it reflect the impact of the new MAC features (e.g., MIMO, channel bonding) on the WiFi power consumption? Additionally, we study the impact of the power saving mode (PSM) which was ignored in the original model and show that ignoring PSM results in significant overestimation of the total power consumption at low frame generation rates. Accordingly, we propose a new model that works across the full range of frame generation rates and verify its accuracy for a wide range of parameters and devices.