Greening campus WLANs: energy-relevant usage and mobility patterns Fatemeh Ganji a,* , Lukasz Budzisz a , Fikru G. Debele b , Nanfang Li b , Michela Meo b , Marco Ricca b , Yi Zhang b , Adam Wolisz a a Telecommunication Networks Group, Technical University of Berlin, Einsteinufer 25, FT 5, 10587 Berlin, Germany b Dipartimento di Elettronica e Telecomunicazioni, Politecnico di Torino, Corso Duca degli Abruzzi, 24, 10129 Torino, Italy Abstract The past years have witnessed a significant increase in the number of WLANs deployed in most of the enterprises, campuses and public areas to provide high-speed Internet connectivity. These WLANs typically consist of APs densely installed to assure enough capacity to meet users demand during the peak period of activity. At the same time, it translates into a serious energy wastage during low-utilization periods, when capacity is not needed and the APs. To reduce this wastage, many proposed solutions consist of adapting the active capacity to the actual needs, introducing switching strategies able to turn on and off the APs. The effectiveness and potential benefit of these strategies strongly depend on the user behavior and traffic patterns. In this paper, we focus our analysis on the real usage characteristics of a dense WLAN (such as users’ be- havior and users’ mobility patterns) in a university campus and evaluate potential energy savings and benefits achievable when introducing AP on/off switching strategies. We discuss different strategies, in which decisions are based either on: (1) historical behavior in the campus, or on (2) current AP utilization. In addition, considering large overlapping coverage available in dense WLANs, we investigate users’ mobility patterns to derive further improvements to AP switching strategies. The results show that, due to the repetitiveness of users’ patterns and large differences in WLAN usage between days and nights, as well as between weekdays and weekends, large savings of up to 40% can be easily achieved. Moreover, by fine-tuning the strategies in different areas of the campus, additional savings are further possible. The deployment of these strategies leads to energy saving and, as a practical consequence, to a remarkable reduction of electricity costs. Keywords: Energy-efficiency, green networking, WLAN, trace-collection analysis. 1. Introduction and motivation Wireless local area networks (WLANs) have nowadays become one of the most preferable solutions to provide reliable and high-speed Internet connectivity [1]. Especially in campus and enterprise scenarios, where providing wireless connectivity is not only preferable but tends to be indispensable, dense WLANs (i.e., featuring thousands of APs per square kilometer) have been deployed to provide coverage and capacity. Different aspects of deployment of dense WLANs, ranging from designing high-speed AP components to defining criteria for management of large-scale WLANs, have been studied for many years. However, only re- cently the excessive energy consumption of dense WLANs has become a concern. The problem of energy wastage in these networks is caused by the inactive APs that are deployed redundantly to provide close-to-the-peak capac- ity even under condition of marginal traffic. On/off switching strategies have emerged in the literature as a very promising solution to overcome this problem. With these strategies, the inactive APs are powered off to the extent that APs remaining in operation can provide the coverage and the capacity needed to meet the current user de- mand, and any additional APs are only powered on, if needed. AP switching strategies rely on measurement and estimation of the user demand to take the on/off decisions. Two main families of approaches exist: c 2014 Elsevier. Personal use of this material is permitted. Permission from Elsevier must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. * Corresponding author Preprint submitted to Computer networks October 7, 2014