| dc.description.abstract | HWMP (Hybrid Wireless Mesh Protocol) has been set as the default routing protocol for the
ongoing IEEE 802.11s WMN (Wireless Mesh Network) standard. Unlike most multi-hop routing protocols, which operate at the network layer, HWMP operates at the MAC layer and uses IEEE 802.11s Airtime as a routing metric.
In this dissertation, and in an attempt to delve into the subtleties of this new promising technology, we started first by deploying a real-world pre-IEEE 802.11s indoor WMN testbed that spans two separate buildings, and implements the main traits of this new IEEE standard, e.g., architecture, HWMP, and Airtime. We ascertained the practicality of the testbed by testing real-world
trafic such as having Wi-fi enabled stations browsing the Internet, using the deployed WMN as a backhaul. We identified major practical issues and addressed them, e.g., clients association, Internetworking, and supporting multiple gateways. To encourage the use of real-world WMN testbeds,
and to cope with the scarcity of such testbeds, we made the implementation open-source and available online.
IEEE 802.11s Airtime metric depends on the observed loss and transmission rates. However, IEEE 802.11s did not delineate a specific way to measure the loss rate; It is left as an open research issue. To estimate loss rate, most routing metrics, e.g., ETX and ETT, use the broadcast
approach whereby broadcast probe frame losses are measured. Such an approach suffers fundamental
shortcomings which were addressed by introducing the passive approach. This latter uses data trafic frames instead of broadcast frames. However, the passive approach still suffers from the major shortcoming of probing idle links. In this dissertation, we propose a novel loss rate estimation scheme that overcomes the shortcomings of both broadcast and passive approaches. The proposed scheme estimates losses by tracking their causing events, mainly contention and interference.
However, interference measurement remains a challenging task merely because of the impracticality of the current interference models. We propose a novel practical interference measurement
framework that adapts the physical interference model and uses a probabilistic model for frame arrivals. A new interference-aware routing metric, ICE (Interference and Contention Estimator), is shaped and compared to ETX and IEEE 802.11s Airtime routing metrics using a real-world pre-IEEE 802.11s WMN testbed. ICE outperforms both ETX and IEEE 802.11s Airtime with an average of 16%. This outperformance is a significant improvement when taking into account that we were using a single-channel radio approach.
While experimenting with HWMP and Airtime, we noticed a ping-pong effect in the behavior of the network. The very few references to this effect in the literature condemn it as a perilous behavior
but only superficially address the problem. In this dissertation, we also present a thorough study of the IEEE 802.11s Airtime ping-pong effect, and we highlight its correlation to the underlying
rate control algorithms. Using different rate control algorithms (e.g., ARF, AARF, ONOE, AMRR and Constant rate), we prove that transmission rate adaptation is the principal cause behind the effect. We show that the effect is an inherent behavior and not necessarily a perilous one and that an accurate characterization of the effect will improve network performance. We present and discuss a novel research direction consisting of shaping ping-pong-aware mechanisms that, by detecting when a link undergoes such an effect, adapt the network resources for a better performance.
In this context, we present a ping-pong-aware mechanism that is O(1), decentralized, and can be easily integrated into the IEEE 802.11s routing protocol. Using extensive ns-3 simulations, we show that the new mechanism has an average outperformance of 5%. This latter is a slight improvement; However, it remains as a proof of concept for future research in the direction of shaping better ping-pong-aware mechanisms. | en |
