The wireless LAN market (WiFi or IEEE 802.11) is one of the fastest-growing segments of the communications industry. On March 21, 2003, for example, the business section of the Ottawa Citizen announced that Cisco bought Linksys in a stock deal valued at US$500 million. According to the commentary, this deal moves Cisco into new territory: the consumer and small business market for networking equipment, one of the few rapidly growing segments of the troubled sector. As these wireless devices proliferate further, completely new networking architectures will evolve, in which devices communicate with each other wirelessly and potentially over multiple hops, using protocols such as IEEE 802.11b or IEEE 802.11g, Bluetooth, and others. Additional factors contributing to this development are the Network Processing Elements (NPEs) chipmakers such as Intel have developed: these network processors efficiently support a range of communication technologies, and can be integrated relatively cheaply with other devices, seeding the environment with even more wireless network nodes.

The resulting network architecture is a multi-hop wireless networks: due to the limited radio range and the existence of propagation impairments, not all the nodes that want to talk to each other will be in direct range, necessitating routing through intermediate nodes. Depending on the particular application scenarios, all, some, or none of the nodes may be mobile. Since the term MANET (mobile ad-hoc network) focuses on pre-dominantly mobile networks, we will use the term mesh network in the following to refer to all multi-hop wireless networks. With devices getting smart and connected, ubiquitous networking and ambient intelligence are becoming an integral part of people's life. Mobile and embedded devices will be a part of a global communication infrastructure by integration of, for example, vehicle, home, personal and body area networks with the Internet. For mesh networks to live up to their potential, a number of challenging issues need to be addressed:

Establish ubiquity in networking: connect smart devices with each other (e.g. vehicle or home ad hoc networks) and integrate these ad hoc networks to the Internet.
Reduce human interaction required for configuration, adaptability, and dependability to improve usability and cost effectiveness.
Establish security and dependability for ubiquitous networks.
Low total cost of purchase, installation, and maintenance to provide complete coverage despite difficult and changing RF environments.
Easy and spectrally efficient network scaling.
Assured and reliable service for broadband subscriber data flows with an end-to-end communication quality that matches or exceeds that available over wires.

This work addresses two issues, QoS support and self-organization, in the context of the ZAP (Zone Access Points) project at NCIT. Mesh networks will carry a variety of traffic to support a range of applications, from video and voice to more "traditional" data services such e-mail, file transfer, telnet, and WWW traffic. To support these applications sufficiently, the network has to be able to differentiate between data packets and offer appropriate network services in terms of throughput, delay, jitter, and loss rate, i.e., we need QoS support. QoS has been heavily research in wired networks, starting with ATM, but more recently with efforts in the IETF as well. Intserv, Diffserv, and MPLS are all solutions to provide service differentiation in previously only best-effort IP networks. These approaches have been heavily researched, their relative strengths and weaknesses are well understood, and increasingly there is experience with their deployment in production networks. It is therefore tempting to apply these solutions to the new mesh networks as well.

However, it is the central assumption of our work that solutions for the wired domain are not appropriate in the new wireless domain. This core assumption is based on a number of reasons. For starters, the IETF has explicitly recognized that for mobile multihop wireless networks (MANETs), traditional Internet routing protocols are inefficient and chartered the MANET working group, to design, evaluate, and propose new routing protocols for standardization. The routing protocols under development (such as AODV, DSR, or OLSR) address many of the specific multihop wireless issues. However, driven by insights from wired routing protocols such as RIB, they all attempt to find the shortest possible path between sender and receiver. Evaluation of these protocols is heavily based on simulations (using NS2 or Opnet), which by default use a very simple radio propagation model. In recent work, we have identified that the use of a more realistic physical link layer model results in very poor performance of these protocols and suggested enhancements to AODV and DSR that address wireless link characteristics such as unstable and fluctuating signal propagation.

Besides routing, QoS provisioning is also substantially impacted by the wireless multihop nature of mesh networks. The basic problem is obvious in the presence of mobility. Traditional wired networks that provide QoS support achieve this by managing the network resources and applying admission control to new flows. Once a flow is admitted, appropriate network resources can be set aside to meet the QoS requirements of the flow. This requires, among other things, that the data path is known. In the presence of mobility, the data path will change over time, requiring new QoS reservations at the least, and may lead to QoS violations (since user mobility is not under the control of the network, users - and their associated flows - may congregate in one network area and overload it). Appropriate and efficient QoS solutions have to take mobility into account from the start, rather than treating mobility-related changes as the exception.

In addition, in wireless networks, nodes that are not in direct range of each other can still interfere with each other. Published data for IEEE 802.11 radios, for example, shows that the interference range is typically about twice the communication range. This is another inherent property of wireless networks that does not exist in the wired domain. One major consequence is that resource reservation schemes to manage, for example, link bandwidth among all nodes "sharing" the media will result in either extremely high overhead (in essence flooding control information over multiple hops) or are by definition inaccurate (coordinating link access only between nodes in direct reach, ignoring interfering nodes). Again, this argues for a different treatment of QoS from wired networks, where careful resource management at routers is a key building block. This is true even in a completely static mesh network. Some published data shown that link bandwidth between nodes in static network varies over time. This is due to a number of reasons: with signal propagation subject to fading and reflection, the physical environment has a significant impact on wireless communication. In addition, since most radio technologies such as IEEE 802.11 and Bluetooth operate in unlicensed bands, interference from other users of the band will lead to significant changes in link characteristics.

Another advantage of solutions proposed for multihop wireless networks is their self-organizing nature. Recognizing that changes in network topology and link conditions are the norm, rather than the exception, proposed protocols are almost always self-configuring, which addresses the need for low cost operation and reducing the amount of human intervention.