Aquilo: Temperature-aware scheduler for millimeter-wave devices and networks

Abstract

Millimeter-wave is the core technology to enable multi-Gbps throughput and ultra-low latency connectivity. But the devices need to operate at very high frequency and ultra-wide bandwidth: They consume more energy, dissipate more power, and subsequently heat up faster. Device overheating is a common concern of many users, and millimeter-wave would exacerbate the problem. In this work, we first thermally characterize millimeter-wave devices. Our measurements reveal that after only 10 s of data transfer at 1.9 Gbps bit-rate, the millimeter-wave antenna temperature reaches 68°C; it reduces the link throughput by 21%, increases the standard deviation of throughput by 6×, and takes 130 s to dissipate the heat completely. Besides degrading the user experience, exposure to high device temperature also creates discomfort. Based on the measurement insights, we propose Aquilo, a temperature-aware, multi-antenna network scheduler. It maintains relatively high throughput performance but cools down the devices substantially. Our testbed experiments under both static and mobile conditions demonstrate that Aquilo achieves a median peak temperature only 0.5°C to 2°C above the optimal while sacrificing less than 10% of throughput.