Online Redundancy Mining in Enterprise WLAN Traffic

Abstract

As various types of mobile devices (e.g., smart phones, laptops and tablets) get connected via Wireless Local Area Networks (WLANs), the dramatic demands for wireless bandwidth have posed new challenges for efficient operation and maintenance of enterprise WLANs. Previous studies have found certain degree of redundancy embedded in user data, which stimulates new redundancy elimination schemes implemented on gateways to restrain redundant data being transmitted within one enterprise WLAN. Due to both the computation and storage limitations of gateways, it is very hard to process all user data online and sampling methods are adopted to shrink the size of data streams. Existing methods simply sample the original user data in a random way, leading to low efficiency of finding redundant data. In this paper, we conduct an empirical study on effective sampling strategies using real user trace collected from a university WLAN. We first investigate the extent to which WLAN traffic can be redundant. We then further analyze the distribution characteristics of redundant blocks and find that the position of redundant chunks exhibit strong spatial correlations with previous ones. Our observations thus provide solid foundation for designing new sampling schemes which can capture more redundant data embedded in WLAN data and improve the performance of redundancy elimination schemes.