Relatore : Renata CRUZ TEIXEIRA
Co-relazione : MAY Martin
Broadband Internet access is now widespread and many users connect to the Internet from home. Often, Internet users at home are not computer experts. When a performance problem occurs, users have no simple means to diagnose the problem and may call their Internet service provider to fix the problem, even if the problem comes from the user network. This situation frustrates Internet users and incurs a large cost on the Internet service providers which must provision call centers.
In this thesis, we consider techniques for end-hosts to pinpoint whether performance problems occur in the home network or not. We show that some home network configurations affect the end-to-end performance and that existing techniques cannot always pinpoint whether the home network is the performance bottleneck. To get a better understanding of existing home networks at large, we design HomeNet Profiler, a software measurement tool to measure the list of devices active in the home network, the implementation of UPnP in residential home gateways, and the WiFi environment inside home networks. With our dataset consisting of nearly 3,000 homes, we show that home networks are often small but can have up to 20 devices. We demonstrate that UPnP queries can pinpoint cross-traffic from the home network and differentiate local from wide-area losses. We also show that the home WiFi environment is generally dense and has an inherent risk for interference. To leverage this high WiFi density, we design neighbor-assisted diagnosis techniques. These techniques are able to efficiently detect and distinguish uplink and downlink delays and loss rates with small error.
Difesa : 04/22/2013 - 14h00 - Site Jussieu 65-66/211
Konstantina Papagiannaki : Chercheur, Telefonica [Rapporteur]
Nick Feamster : Professeur, Georgia Institute of Technology [Rapporteur]
Ernst Biersack : Professeur, Eurécom
Serge Fdida : Professeur, Université Pierre et Marie Curie
Renata Teixeira : Professeur, Université Pierre et Marie Curie & CNRS
Martin May : Chercheur, Technicolor