With the growing popularity of IEEE 802.11-based wireless networks, it has become increasingly important to understand the characteristics of wireless 802.11 traffic and the wireless medium itself. A number of measurement studies have examined traffic characteristics in wireless networks 10 (10); 1 (1); 8 (8); 5 (5); 7 (7). These studies have measured the wired portion of the network, using wired network sniffers and SNMP polling. Wired network monitoring (WDM) can provide accurate traffic measurements as seen in that portion of the network. They may not, however, disclose characteristics of the wireless medium (the 802.11 MAC/PHY), as wired devices can only see the traffic that is successfully transmitted to the wired side of the AP. While SNMP-based approaches may be able to retrieve such detailed wireless MAC/PHY information through the use of a properly defined MIB (Management Information Base), most existing SNMP MIBs for APs (MIB-I (RFC 1066), MIB-II (RFC 1213), and 802.11 MIB (IEEE Std 802.11-1999)) provide very limited visibility into MAC-level behavior. A further drawback of SNMP-based approaches is that they require an interval between SNMP polls (typically every 1-5 minutes), and it has been shown that long poll intervals may miss wireless clients that associate with APs for less than this poll interval 7 (7).
To overcome the shortcomings of SNMP and WDM, it is necessary to sniff the wireless medium itself. We refer to this technique as wireless monitoring (WM). Like WDM, WM involves a set of devices, commonly referred to as sniffers, which observe network traffic, but in WM, the sniffers are equipped with wireless cards for sniffing the wireless medium. WM has recently been adopted in both wireless networking research, e.g., 9 (9), and commercial WLAN (Wireless Local Area Network) management product development.
There are three advantages to using WM. First, WM captures detailed wireless-side traffic statistics. Second, WM provides per-frame wireless MAC/PHY information, such as 802.11 MAC headers. Third, WM does not require any interaction with the existing network, unlike WDM, where network sniffers need to be attached directly to wired switches.
The data collected by WM can be used for many purposes. Physical layer information, such as error rates, can be used to develop accurate error models for 802.11 WLANs, and for site-planning to determine the signal strengths required to achieve a certain throughput or error rate. Link-layer data, such as the characteristics of data, control and management frames, can be used to develop 802.11 simulation models, and to identify anomalies in the operation of the 802.11 MAC protocol. The overall traces themselves can also be used for emulating 802.11 networks.
WM, however, can be complicated to conduct in practice. Unreliable and varying wireless channel conditions may lead to measurement loss. The goal of this study is to demonstrate that WM can perform reliable and accurate measurement under such non-ideal conditions. We first demonstrate how to improve the capture performance, that is, the amount of the actual wireless traffic captured by a particular measurement technique, by merging multiple sniffer traces. Then, through a controlled experiment, using clients with varying signal conditions, we quantify WM's capture performance in terms of IP and MAC layer statistics.
In this paper, we address all the above problems for accurate measurement technique. However, WM has another big challenge: scalability, i.e. that the cost and management overhead can be significant for the deployment and management of a large number of sniffers. In this work, we limit our work to the fixed number of sniffers for relatively small coverage area (e.g., WLAN in a single floor with less than 10 APs). Based on the promising results of this work, we are currently working towards addressing the scalability problem in more general WLAN environment.
The rest of the paper is organized as follows. In Section 2, we discuss previous measurement studies of 802.11 WLANs. Section 3 describes the setup and implementation of our WM system. In Section 4, we describe a controlled experiment to demonstrate the accuracy of the WM technique in capturing IP and MAC/PHY layer statistics in an actual environment. Finally, we conclude the paper in Section 5 and highlight our ongoing work.