sfi

NDP

ENTERPRISE IRELAND

NCNRC

Communications Network Research Institute

Associated Overhead of WMN Routing

Brian Keegan

Abstract

Wireless Mesh Networks (WMNs) are a type of radio-based network which require minimal configuration and infrastructure. They can be built using relatively low cost radios and computing platforms. The increased capacity of WMNs means that they are now capable of providing backhaul services traditionally provided by wired networks. The attraction of WMNs is their ease of deployment and ability to self organize, self configure and self heal. In order to successfully achieve this objective careful consideration must be applied when constructing a WMN. In order to optimize the network, data should traverse the network by means of the most efficient route. This may mean an optimization of routing algorithms, improved characterization and path selection or alternatively enhancements to the MAC mechanism. These methods and others will involve some form of a trade off between the positive and negative impacts on the network performance. It is our intention to investigate the performance of the various WMN methods in order to reduce the associated overhead so as to maximize the bandwidth usage.

Overview

Routing protocols can be divided into two categories; active and reactive. Active routing protocols generally work by periodically broadcasting and maintaining routing tables. This type of protocol produces a significant amount of overhead and consequently consumes a large fraction of the available bandwidth. Reactive routing protocols attempt to minimize the impact on the network by gathering information only when necessary. However, the characterization of a network is carried out by obtaining performance metrics by means of active or passive techniques. With active probing there will be an overhead associated with the broadcast of probe requests. Broadcasting probe requests, however, will have both a direct and indirect overhead. It is our intention to compare these techniques and show how one can optimize a wireless mesh network by carefully selecting the appropriate technique.

Multicasting and by extension, broadcasting are services used to transmit data to multiple nodes. However, the 802.11 WLAN standard defines these services as unreliable as well as making use of a single transmission rate. Multicasting can be used as an efficient means of communication when utilized in a wireless mesh as a one hop broadcast to multiple receivers. However, this is not the case when used in a multihop network due to unnecessary traffic being transmitted to multiple receivers.

Investigation

Our investigation of WMNs includes a number of different test bed scenarios, architectures and topologies. Initial performance testing was carried out on a simple linear mesh network (i.e. A basic multihop network with IP forwarding) using single radio wireless enabled Soekris boxes. This was then extended to a distributed WMN throughout our building again using the Soekris boxes but this time including the Click Modular Router software. Our current configuration involves 16 desktop PCs configured as wireless mesh nodes contained within a single laboratory room. Utilizing this laboratory set up we were able to conduct a series of experiments to compare the overhead experienced using various routing algorithms which are required to probe the network. The results from these experiments serve to demonstrate the advantages of using a passive monitoring technique to probe the network by showing a much reduced overhead penalty.

The next step in our investigation of WMN performance involves the construction of a 25 node (i.e. 5 x 5 grid) WMN using RF cabling. This will allow us to eliminate external sources of interference as well as affording us greater flexibility and control over the signal propagation environment. For example, using a wired mesh network in this way will allow us to control the interference between different links through the use of inline attenuators.

To complement this work we will also carry out network simulations. By using a basic abstract network simulation it will be possible to analyses the basic factors which may influence network performance. This should prove to be particularly useful when investigating multicasting techniques. By simulating multicast trees we can analyses the effects of broadcast latency with relation to the distance between links. We can also test the efficiency of different multicast tree topologies and investigate how they perform when using multi-rate broadcasting. These techniques may be used to develop a more efficient routing technique.

WMN