Monday, May 31
|11:30-12:30||MAC & PHY layer measurements|
|14:10-14:50||Data collection and analysis of operational networks|
|15:20-16:00||Wireless contact trace analysis|
In this paper, we systematically investigate different factors and their effects on the wireless transmission properties using a full-factorial experimental design of a real-word wireless sensor network. We quantify the impact of primary factors such as the wireless channel, physical position, transmission power, line-of-sight, and their interactions on the wireless communication. While some of our results support conventional assumptions, this study also shows that there are many properties which are in contrast to existing findings. For example, there is no significant correlation in the measured received signal strength between different but equally-distant transmitters, yet the correlation coefficient between two transmitters is above 0.95. Further analysis reveals the strong interaction of transmission frequency and the physical position, while the transmission power and the direction of the transmissions have an isolated, non-interacting effect on the measured received signal strength. Since the analyzed network consists only of a simple, low-cost hardware, the results of this experimental analysis can serve as valuable insights in planning and deploying wireless sensor networks in different application scenarios.
In this paper we present the first field measurements taken using a new approach proposed in  for mea- suring link impairments in 802.11 WLANs. This uses a sender-side MAC/PHY cross-layer technique that can be implemented on standard hardware and is able to explicitly classify lost transmission opportunities into noise-related losses, collision induced losses, hidden- node losses and to distinguish among these different types of impairments on a per-link basis. We show that potential benefits arising from the availability of accurate and reliable data are considerable.
Recent research results have shown how measurements of bandwidth metrics using traditional tools and techniques have to be reconsidered in the presence of WLAN links. The main reason behind this is the CSMA/CA protocol used to regulate the distributed access in the wireless medium. In this paper we identify a set of practical issues that tools based on delay measurements must face to implement end-to-end bandwidth metrics in hybrid paths (i.e., wired/wireless). The paper provides results through extensive simulation to validate the findings of the study.
Building a campus-wide wireless LAN measurement system faces many efficiency, scalability and security challenges. To address these challenges, we developed a distributed Wi-Fi sniffing program called Saluki. Compared to our previous implementation and to other available sniffing programs, Saluki has the following advantages: (1) its small footprint makes it suitable for a resource-constrained Linux platform, such as those in commercial Wi-Fi access points; (2) the frame-capture rate increased more than three-fold over tcpdump with minimal frame loss; (3) all traffic between this sniffer and the back-end server was secured using 128-bit encryption; and (4) the traffic load on the backbone network was reduced to only 30% of that in our previous implementation. In this paper, we introduce the design and the implementation details of this high-performance sniffing program, along with preliminary evaluation results.
Users are generating and uploading multimedia content to the Internet at an unprecedented rate. Residential broadband networks, however, have low upload capacities and large packet latencies. Wi-Fi networks that are used to access the Internet can suffer from high packet losses and contention latencies. All these factors can result in poor video quality for residential users. Using packet traces and active measurements from houses, we study video quality in residential scenarios. We analyze the primary factors that contribute to poor performance and compare the performance over both the wireless hop and broadband hop. Our measurements show that the upload capacities on the broadband links restrict the video bitrate (and hence the resolution) that can be transmitted. Residential wireless networks, however, have much higher capacities than the broadband links and despite being densely deployed, do not see extended periods of high utilization. Our measurements shed light on the video transmission quality that is typically achievable from residences and are used to characterize the reasons behind quality deterioration.
Cellular technology is widely used for Internet access, also because most operators are now offering Mbit/s data rates at affordable prices. Many studies analyzed the performance of these networks using analytical or simulation approaches. However, due to lack of data from operational environments, very little is known about the performance of real cellular networks. In this paper, we assess the performance of the 3G network of one of the major European telecom operators, using several recent traffic traces of TCP connections to port 80 and 8080. After presenting global performance statistics related to all the network users, we focus on a specific set of heavy users. To assess their performance and to uncover the related causes, we introduce an investigation approach easily repeatable in the very common situation where only data traces are available, with no other information such as mapping of users to cells, network capacity, or packet payload. Analyzing both "single long-lived connections" and "multiple long-lived connections", we assess the performance of those users, providing insights on how and why performance can vary significantly over time and among different users.
Understanding current performance and usage of IEEE 802.11 networks is valuable for effectively managing and deploying wireless networks in the future. This paper analyzes a wireless trace from NetGames'08 and compares the results with previous research. The analysis indicates similarities in the dominance of Web traffic but differences in the overall volume and traffic patterns from prior wireless Internet traffic measurements, with emerging p2p applications consuming significant uplink capacities. The workshop schedule greatly influences network usage and the number of attendees using the network in the forum room significantly affects the received signal strength.
By analyzing contact patterns of people over time, it is possible to build efficient delay tolerant networking (DTN) algorithms and derive important parameters for epidemiological studies. Significant research has been performed in the automated acquisition of contact patterns using mobile devices such as Zigbee motes or Bluetooth-enabled cellular phones. However, the limited number of studies described to date do not capture the breadth of human experience or specifically include the acquisition of health related information. In this paper we present Flunet, a mobile contact-tracking network deployed in a university environment during flu season in a Canadian winter. Flunet tracked of contact patterns 36 participants and their proximity to 11 stationary nodes using MicaZ motes over a period of three months. Participants filled out weekly surveys on the state of their health. This study is distinct from others because we incorporate of health information and the impact of sub-zero temperatures on mobility patterns. This paper presents a preliminary analysis of the data set, primarily from a DTN perspective. We present fundamental attributes of the dataset, the efficiency of routing for single pass and flooding-based algorithms and a preliminary look at the impact of reported health on the connectivity of the network. We also report on differences in participant compliance with data acquisition, health survey completion and manual contact tracing. We conclude with recommendations for future work in automated acquisition of contact and health related information.
In this paper we study the influence of hub nodes used to relay data in empirical traces obtained by human-based delay tolerant networks (DTNs). We use a little variation of the standard measure of centrality (or betweenness) over underlying graphs associated to temporal networks in order to establish (in probability) the ability of a networks to forward informations based on a restricted number of active hubs (relay nodes).
All the consideration are made exploiting time-dependent networks associated to four real traces gathered by human mobility experiments. Empirical cumulative distribution of the node betweenness and on the shortest paths length (or geodetic) are derived and characterized. In particular it is shown that the geodetic path length follows a lognormal (skewed) distribution. It is also observed that betweenness measure, interpreted as probability distribution, seems to exhibit an exponential decay, even though less accurate with respect to geodetic length fitting, being in some cases very high for few nodes and very low for almost the whole body of nodes.
Based on this knowledge, we study the delivery success ratio when a set of nodes with low betweenness are inactivated and the forwarding schema is based on single copy strategy. Under these assumptions, we give the probability that a $k$-length paths connecting an arbitrary source-destination pair belongs to the set of activated hub nodes. The results show the trade-off between the number of relays activated in a temporal network and the delivery success ratio when the forwarding is allowed only for these hubs.