Wireless Networks 11, 451–469, 2005 C  2005 Springer Science + Business Media, Inc. Manufactured in The Netherlands. Self-Tuning Wireless Network Power Management MANISH ANAND, EDMUND B. NIGHTINGALE and JASON FLINN Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, MI 48105 Abstract. Current wireless network power management often substantially degrades performance and may even increase overall energy usage when used with latency-sensitive applications. We propose self-tuning power management (STPM) that adapts its behavior to the access patterns and intent of applications, the characteristics of the network interface, and the energy usage of the platform. We have implemented STPM as a Linux kernel module—our results show substantial benefits for distributed file systems, streaming audio, and thin- client applications. Compared to default 802.11b power management, STPM reduces the total energy usage of an iPAQ running the Coda distributed file system by 21% while also reducing interactive file system delay by 80%. Further, STPM adapts to diverse operating conditions: it yields good results on both laptops and handhelds, supports 802.11b network interfaces with substantially different characteristics, and performs well across a range of application network access patterns. Keywords: power management, self-tuning, 802.11 1. Introduction Wireless networks provide mobile computers with continuous Internet connectivity. Yet, power management is needed to en- sure that the network interface does not overly tax the limited battery capacity of a mobile device. For example, our mea- surements show that using a 802.11b network card without power management can shorten the battery lifetime of a HP iPAQ 3870 handheld by almost 50%. The popular IEEE 802.11 standard [10] provides a power- saving mode (PSM) that periodically disables the network in- terface during periods of no activity. However, PSM does not adapt to the power characteristics of the network interface and mobile computer, the intent and access patterns of applica- tions, or the needs and expectations of users. While PSM pro- vides excellent energy conservation in some circumstances, it can also substantially degrade interactive application perfor- mance and even increase the energy needed to perform certain activities. For instance, PSM causes an unacceptable 16-32x slowdown in the time to list directories stored in NFS. We show that different power management strategies are needed in different circumstances. Rather than take a “one size fits all” approach, we propose self-tuning power manage- ment (STPM) that adapts to the characteristics of the network interface, mobile computer, and applications. We have imple- mented STPM as a Linux kernel module that runs on both handhelds and laptops. STPM differs substantially from other adaptive strategies such as the PSPCAM mode of the Cisco Aironet 350 card [4] and the bounded slowdown protocol of Krashinsky and Balakrishnan [13]. STPM explicitly considers the time and energy costs of changing power modes. These transition costs can be quite large for current 802.11b cards—several hundred milliseconds in most cases. STPM also explicitly considers the base power usage of the mobile computer. Finally, STPM provides a simple interface that allows applications to dis- close hints about their intent in using the network interface. For legacy applications that have not yet been modified to dis- close such hints, STPM uses passive monitoring and heuristics to generate hints on their behalf. STPM then tunes its power management strategy to observed network access patterns. Our results show that STPM provides significant energy conservation with minimal performance impact for applica- tions such as distributed file systems, streaming audio, and thin-client remote X displays. For instance, STPM reduces the total energy usage of an iPAQ running the Coda distributed file system by 21% compared to PSM, while also reducing interac- tive file system delay by 80%. Further, STPM shows benefits across a diverse set of network interfaces and mobile devices. We begin with a discussion of the limitations of current wireless power management. Section 3 outlines the principles we followed in the design of STPM. In Sections 4 and 5, we describe our implementation and compare its performance and energy conservation to that of other static and adaptive power management strategies. Finally, we conclude with a discussion of related and future work. 2. Motivation Current 802.11b power management schemes can severely de- grade the performance of latency-sensitive applications. For example, figure 1 shows how power management affects the time to list directories of varying sizes stored in the Network File System (NFS) [19]. These results were generated by exe- cuting ls on a HP iPAQ 3870 handheld with a Cisco Aironet 350 802.11b card. The solid line at the bottom labeled “CAM” shows perfor- mance in continuously-aware mode (i.e. without power man- agement). The dashed line at the top labeled “PSM-static” shows performance with the default 802.11b power saving mode (PSM). Thedifference between these two shows that