|
|
|
|
|
|
|
Today, the "de facto" standard for the wireless Internet is undoubtedly represented by the IEEE 802.11 technology. Born as a niche technology for providing wireless connectivity in small office/enterprise environments, 802.11 has in fact become a common and cheap access solution to the Internet, thanks to the large availability of wireless gateways (home modems, public hot-spots, community networks, and so on.). One of the key factors for the wide acceptance and success of this technology is the simplicity and robustness of the Medium Access Control (MAC) protocol, employing the Distributed Coordination Function (DCF). Based on the well-known carrier sense paradigm, DCF is able to work in presence of interference, which is very critical for networks operating in unlicensed spectrum. Interference source may include stations belonging to the same network (/endogenous interference/), but also other external noises and overlapping networks (/exogenous interference/). Although DCF has been natively designed for coping with interference, the incredible proliferation of independent 802.11 networks is creating serious performance problems, despite of the availability of capillary connectivity. In this talk we discuss some possible strategies for enabling performance optimizations in highly interfered 802.11 networks. First, we propose a simple mechanism for evaluating congestion and channel conditions in error-prone 802.11 links. The mechanism is based on channel observations, independently carried out by each station, and is able to distinguish /collision-induced/ and /channel-induced/ errors, thanks to a bi-dimensional kalman filter coupled with a change detector. The estimation of these independent error sources allows: i) the correct tuning of the contention window as a function of the network load; ii) the correct selection of the most suitable modulation/coding scheme as a function of the link quality. Second, we discuss alternative im! plementations, employed by commercial cards based on Atheros chipsets and MadWiFi drivers, devised to reduce the interference effects by dynamically tuning the receiver sensitivity. We prove how the current solutions for estimating the interference conditions in actual cards are quite poor. Specifically, we document some critical side-effects which arise in strongly attenuated links, where weak signals can be erroneously considered as interference. Our conclusion is that interference estimation has to be very carefully managed, in order to provide effective protocol adaptations.
|
|
|
|
|
|
|
|