Architecture, applications, and data analysis of a wireless network location service

The ability to obtain location information on wireless devices is critical to many applications and services. This thesis creates a centralized location service for wireless networks called [email protected] The system has been implemented on the Wireless Andrew network at Camegie Mellon University. It has tracked over 10,000 unique devices using nearly 1,000 access points spread over a 110 acre campus. A taxonomy of architectures for wireless location services is created listing the strengths and weaknesses of the main types. The centralized architecture, as used in [email protected], demonstrates that it is significantly more bandwidth efficient than a distributed service when implemented on a large scale. The [email protected] system uses modularized software components and scales linearly, based on network size and user base, onto a cluster of computers. [email protected] requires no software installation on the wireless devices, but polls the access points directly allowing for an omniscient network view. The architecture of [email protected] and its capacity to store, organize, and retrieve hundreds of millions of data points in an efficient and reliable manner is presented. Data recorded by [email protected] has been analyzed to determine wireless network usage and user mobility. The results of this analysis are surprising in that the average user proved to be relatively stationary, using fewer than three access points per day. However, we also observed sub-groups of users that are highly mobile, and who intensively use the network. This most mobile group increased four-fold in size from 2003 to 2005. A toolkit with 85 functions was developed to support location-based applications. The toolkit reduces location-based application implementation times from days to hours. Three location based applications were implemented as part of research to demonstrate the power and functionality of the location system. These applications utilize access point information on over 10,000 users to answer questions such as: how various diseases would spread on campus, how many potential witnesses were near known crime incidents, and predicting how many people will be at a location in the future. [email protected] has proven to be an end-to-end scalable solution that meets the need of providing location data on access point based wireless networks.