Analysis On Redundancy Of User Data In WLAN Traffic

Wireless Local Area Networks (WLANs) bring us much convenience when we access Internet, and become an indispensable part of our daily life. Recently, as vari-ous types of mobile devices (e.g., smartphones, laptops and tablets) get connected, the dramatic demands for wireless bandwidth challenges efficient maintenance of enter-prise WLANs. In Wide Area Networks(WANs), researchers have found certain degree of redundancy embedded in network traffic. Here, redundancy refers to duplicated data transferred between the two ends of network communications, the sender and the re-ceiver. By reducing the transfers of duplicated data segments, redundancy elimination is an efficient method to alleviate network traffic.Redundancy detection is a key phase of redundancy elimination. To detect all the redundancy, every data segment should be compared with data appeared before. Ob-viously, it is time and storage consuming to check every data segment. So researchers have proposed several sampling algorithms, which aim to detect redundancy with less time and storage. After sampling, a subset of redundancy, related to the duplicated samples, is detected.However, these existing algorithms simply sample the data in a random way which leads to low efficiency of detecting redundancy. In this thesis, we first explore the extent of the redundancy in real traces collected in an enterprise WLAN of a university in China. Moreover, we conduct simulations on existing sampling algorithms with the real traces, and focus on their performance such as rate of redundancy detected, real sampling rate and hit rate of samples. We find that there is a large potential for sampling algorithms to improve their performance. We further investigate the distribution of redundancy which exhibits strong spatial correlations. Our observations thus provide solid foundation and for designing new sampling algorithms which can provide highefciency for redundancy elimination.