Analysis of User Mobility on the Wireless EPFL Campus Network

A few notes from the presentation of the analysis done by Guillermo Fernandez Castellanos (Professor Matthias Grossglauser) of the User Mobility on the EPFL wireless network. He built an analysis tool (BEGO) of WiFi access points’ logs in order to understand the roaming patterns and propose new roaming models. Data where collected only from the INF building.

2 types of WiFi networks:
- clients build an ad-hoc network
- infrastructure with access points

3 types of mobile users:
- Static user (connects always from the same place)
- Pseudo-mobile user (connects from different places, but disconnects while moving)
- Mobile users (moves while connected)

There where virtually no mobile users detected. The “mobile” population splits more or less in two halves between static and pseudo-mobile users. The heaviest wireless use are at the cafeteria and in the computer room. The average highest simultaneous use of access points is by 4 users. Over 9 monthes, 50% of people connected to 4 APs and 75% connected to only 5 APs. 25% of the people used the wireless network for 1 day over 9 monthes and 66% only 5 days.

The EPFL wireless network is formed by wireless islands.

It is difficult to analyse roaming from AP logs because it seems like they are either badly installed and/or have a bad implementation of the roaming algorithm. In most cases, roaming does not mean mobility.

A factor for mobility could be: (number of access points)/(number of sessions). The higher the more mobile.