TPRC has ended
Back To Schedule
Saturday, September 28 • 2:35pm - 3:10pm
The Emperor Has No Problem: Is There Really Wi-Fi Congestion at 2.4 GHz?

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

Download Paper

?Wi-Fi congestion is a very real and growing problem.? So said FCC Chairman Genachowski in a recent proceeding, and this sentiment is widely heard. But is it true? And what does the term congestion mean? We are aware of only three engineering studies that bear on these questions: Sicker et al., TPRC 2006; Mass Consultants for Ofcom, 2009; and Van Bloem & Schiphorst for the Dutch spectrum regulator, 2011. Unfortunately, these seminal efforts do not provide definitive answers. Beyond this, evidence for ?congestion? problems in the 2.4 GHz ISM band is anecdotal and inconclusive at best.

While some users no doubt sometimes perceive ?congestion? in some places, that is not sufficient to prove that there is a policy problem. We therefore set out to develop a list of user experience-oriented service impairment criteria that, if met, would prove that congestion exists to a degree that justifies regulatory intervention. While we focus on Wi-Fi services in the 2.4 GHz band, our proposed method of objective congestion criteria for packet based ("all-IP") communication is generalizable to other cases, e.g. the discussion about a "spectrum crunch" in cellular bands. This is a step towards a larger research program that integrates concerns about occupancy, congestion, utilization, quality of service and user satisfaction into replicable metrics to characterize band use.

We base our work on a review of prior studies, new experimental data, and an analysis of the engineering factors underlying what is perceived as ?congestion?. In order to provide solid experimental evidence, we are planning lab and user experiments on service levels in 2.4 GHz Wi-Fi networks. We aim to collect data including performance logs for a campus network and massive network congestion with large numbers of users; user satisfaction surveys at times of different network load; and bench tests of congestion/degradation for different traffic types and use scenarios.

We understand claims about band congestion and spectrum shortage to be claims about achievable service levels, i.e. the degree to which users in a band can fulfill their service needs. As is well known in the literature, poor performance against commonly used low-level metrics such as spectrum occupancy, utilization or congestion do not necessarily imply service degradation; further, degradation does not necessarily mean that service needs cannot be met; and none of these metrics take economic utility (user willingness to pay for a given service level) into account.

We therefore propose that ?congestion? rises to a regulatory problem if scenario-specific performance metrics are reduced significantly, leading to a significant increase in the percentage of users who can?t complete a valuable task, on a persistent, ubiquitous basis in two or more key scenarios in a band; and that this service degradation occurs in spite of the use of state-of-the-art best practices, such as careful frequency planning, the use of complementary bands, and careful access point placement. The numerical values of parameters for significant performance reduction, percentage of users affected, recurrence and ubiquity will be informed by the experimental work described above.

Based on the collected experimental data, analysis and the resulting criteria we will then evaluate the anecdotal claims of congestion. Our preliminary conclusion, to be tested as work progresses, is that neither engineering studies nor stories in the press support the claim of widespread service level failure, contrary to received wisdom. So far at least, Wi-Fi congestion is not a real problem.


Andreas Achtzehn

Institute for Networked Systems, RWTH Aachen University

Saturday September 28, 2013 2:35pm - 3:10pm PDT
Founders Hall 111

Attendees (0)