Abstract: In recent years, the number and variety of wireless network installations have dramatically increased, from small-scale installations spanning buildings and campuses to much larger-scale installations spanning cities and metropolitan areas. As these wireless networks proliferate, the population of users taking advantage of these networks for communication and access to online services and information also increases. To aid this growing population of users, many research and development efforts focus on upgrading and enhancing the services provided to the users.
As part of these efforts, it is crucial to analyze real wireless networks to understand better how users take advantage of them. These analyses are important for at least two reasons. First, they are helpful for creating more realistic models of users when simulating new services, which is a common technique used in the mobile networking community to predict performance. Second, they are also helpful in focusing research on topics that will impact users the most. This thesis analyzes two very different wireless networks: the Metricom Ricochet packet radio network, a high latency, low bandwidth, metropolitan-area wireless network, and the WaveLAN network installed in the Gates Computer Science Building, a low latency, high bandwidth, local-area network. These analyses answer questions about network utilization, traffic characteristics, and user mobility rates and patterns. Among other results, we find that traffic peaks are usually caused by a single user rather than multiple users, and that significantly asymmetric network throughput would be undesirable for the WaveLAN network users. We also determine that users can indeed be categorized based on their mobility in the Metricom analysis and based on their usage characteristics (applications, time of usage, and number of remote hosts visited) in the WaveLAN trace. We also present clustering and visualization as two tools not commonly used in the networking community that are very useful in doing such analyses.
While the results from these analyses are necessarily valid only for the particular network and user communities studied, we believe many of the conclusions will hold for similar network environments, and the analyses also provide a blueprint for others desiring to perform similar investigations.
Publication Year: 2000
Publication Date: 2000-01-01
Language: en
Type: book
Access and Citation
Cited By Count: 4
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot