In the C@R EU financed FP6 project we are developing a mobile communication infrastructure form rural areas. This communication framework will be able to provide high level services staring with localization and ending with SIP proxy. A Wireless Mesh Network will be formed from the partcipating devices. We will compose a software boundle from already exiting applications (mobile IPv6, SIP proxy, VoIP client...) and application developed by us for the Maemo and OpenWRT platforms. With the help of this software package the Nokia770 will be able to take part in the WMN as a full featured participant.
To validate the feasibility of the WMN based on Nokia770 and OpenWRT based devices we made several measurements:
Based on these measurement a Nokia770 and an OpenWT based WMN seems to be feasible. There will be two mode of the planned WMN protocol:
A sample topology is shown in the Figure 1. The WMN will be divided into zones. At the central point of each zone there is a gateway node. The gateway node has high speed out-of-band connectivity to other gateway nodes and to the Internet. In our picture the medium-sized nodes are the gateway nodes (like Z1). In our case we will use the classical W-ISP solution to provide the backbone. Some P2MP nodes will also be installed in high buildings to provide connectivity for gateway nodes (like C1). We will use our current 802.11h-based infrastructure for this but in the future it can be replaced by WiMax. Each circle in the picture represents a zone (In reality the zones are not circles, but we did this only to simplify the picture). Inside the zones there is a gateway node (like Z1) and several simple nodes (like N1)
Figure 1.
To ensure scalability we will use two-level routing. The zone level routing will be a link state routing based algorithm. The inter-zone routing will be based on a BGP-like approach.
Figure 2.
In the Figure 2 there is a the high level scheme of our WMN solution. In the physical layer all but the C type node (which has sectorized antennas) all other nodes have Omni directional antennas. As is widely known, directional antennas have a higher gain than Omni directional antennas and they produce less interference, but they are susceptible to wind damage and the user should mount them with care. As we intend to involve volunteers without any special knowledge we will use Omni directional antennas.
To be able to cooperate with other devices without special drivers we will use the 802.11g and 802.11b protocols. With this solution each WiFi speaking device will be able to utilise the services of our network. The Wireless Mesh software will be based on MIT’s GNU licensed Roofnet code base.
We will use time division and frequency division multiplexing for medium access control among nodes inside a zone. The central node here has a traffic matrix and a topology matrix. It will schedule the resource allocation based on these information sources. The routing inside the zone will be a special class of source-based routing. For each communication traffic flow the centre node will schedule the time/frequency/modulation/power level values for each hop. One can find simple TDD in WiMax, Tropos, or something similar. One novelty of our solution lies in dynamically tuned time quantum. Depending on the actual traffic situation, the central node may change the length of the resource occupation period. If there are many QoS sensitive traffic flows then this period will be shorter; if there are few QoS sensitive flows this period will be longer. To be able to properly time the actions the central stations will be equipped with GPS receivers. Using this solution we can also minimise interflow interference. The links among nodes are characterised by the measured bandwidth and the number of nodes in the interference range. With this novel metric we will be able to fine tune the transmit power level/modulation/frequency triple to minimize the interference and maximize the global bandwidth. Inside a zone each node will have a matrix showing all the links among nodes inside the zone. Each packet contains for each hop a pointer to a row of this table. Based on this pointer and the routing table the node can select the time/modulation/frequency/power level values to listen or to transmit data packets.
The membership of the zones will be based on the distance of the node from the zone central nodes. The central nodes submit a beacon periodically containing the most important information for the zone. This information is relayed by the zone members selected by the central node. The membership change should be initiated by the centre node as this is the only node in a zone with out-of-zone information.
In the WMN there will be a distributed certificate authority and only registered users can participate in the network. Users will be authenticated with the help of 802.1x. The central node will also schedule flows for authenticated users only.
To maintain the inter-zone routing table a distance path based algorithm will be used. Most of the inter-zone traffic will go through out-of-band channels but the central node may decide to use in band communication channels (especially when the destination zone is a neighbour). The addressing scheme of the WMN will be hierarchical, and the GPS coordinates will provide a top level aggregator for central nodes. The zone centrals will find each other with the help of dedicated multicast channels. The simple nodes themselves will generate an IP address from the prefix received from the central node.
Above this layer we will use the Mobile IP to handle the mobility of the nodes. The correspondent node functionality is supported by most operating systems (Linux, Windows Vista,…). The whole mesh network will act as a distributed home agent with the help of a group communication sub layer.
The infrastructure outlined above can be based on fixed and mobile nodes as well. One might use fixed nodes with high gain antennas as backbone but this is not compulsory. The whole protocol stack will be implemented/ported for OpenWRT and Maemo as well.