The capability of mobile devices has increased to the point where a mobile device is often the primary computing device for many people. As such, mobile devices are being used for bandwidth-intensive and/or delay-sensitive applications, such as HD video streaming and VoIP. To meet the demands of such applications, mobile devices are being equipped with the latest high-speed wireless technology, 802.11n.
However, recent work has shown that 802.11n is significantly power-hungry, with power consumption exceeding 2W in some cases. This is a significant concern for mobile devices, whose usability strongly depends upon the lifetime between charges.
Motivated by this, in this paper we consider the problem of energy management in 802.11n devices. Energy management for 802.11n is qualitatively different from that for its pre- decessor standards. This is because 802.11n has additional power states: beyond a low-power sleep mode, 802.11n MIMO technology offers the possibility of selectively disabling one or more RF-front ends (RF-chains) associated with its antennas, thereby saving energy.
These additional power states motivate us to consider two mechanisms for energy management in 802.11n simultaneously: micro-sleeping which en ables the 802.11n NIC to be put into low-power sleep state for small intervals of time (often a few milliseconds), and antenna configuration management which dynamically adapts the number of powered RF-chains. The design of such mech- anisms is challenging because they are inter-dependent: changing the antenna configuration affects airtime (as defined in of transmissions, which, in turn, affects the sleep opportunities available for nodes in the WLAN.
We have designed Snooze, a novel energy management system for 802.11n WLANs that has several properties. Snooze adapts client sleep durations and antenna configuration to traffic entering and leaving the WLAN, in order to achieve energy-savings; this adaptation creates sleep opportunities by shaping traffic while minimally affecting latency-sensitive applications.
It also takes into account the inter-dependence between micro-sleeping and antenna configuration. This adaptation is AP-directed, application-agnostic, works correctly in the presence of rate adaptation, seamlessly incorporates uplink traffic, and supports multiple concurrent heterogeneous ap- plications per client. Finally, Snooze’s control messages incur minimal overhead by leveraging the frame aggregation capabilities in 802.11n.
We show that it can achieve between 30-85% NIC energy savings over 802.11 CAM (Con-stantly Awake Mode, in which all antennas are active), over a wide variety of applications ranging from VoIP to HD video streaming, large file downloads and chat.
Both micro-sleeping and antenna configuration management contribute significantly to these energy savings, although the proportions they contribute depend on the application. Snooze outperforms 802.11’s PSM (Power Save Mode) even for traffic conditions for which PSM was designed, such as that generated by chat sessions.
Moreover, Snooze’s NIC energy consumption is only 14-39% higher than that for an ideal energy management scheme in some cases. Snooze’s energy savings come at very mini- mal costs: for file downloads it pays a throughput penalty of 2.5% relative to CAM, and for delay sensitive pps it adds up to 5 ms additional latency on average. Finally, Snooze’s control messages consume less than 0.5% airtime and have the property that the control overhead decreases for high-bandwidth applications (where it counts most) because Snooze leverages frame aggregation.
To read more of this external content, download the complete paper from the open online archives at Sigcom.org.