Readings for Lecture Oct 2, 2008

Architecture and Evaluation of an Unplanned 802.11b Mesh Network

The Roofnet wireless mesh architecture has unplanned node placement, omni directional antennas and multi-hop routing and optimization of routing in a slowly changing network. The design goals of this network are operating without extensive planning or central management but still providing wide coverage and acceptable performance. This architecture carries risks that lack of planning might render the network’s performance unusably low.

This Roofnet is designed over an area of about four square kilometers. Most of nodes are located in three or four-storey buildings. Line-of-sight signal propagation between nodes id often obstructed. Each node is hosted by a volunteer user.

Each Roofnet node consists of a PC, an 802.11b card and a roof-mounted omni-directional antenna. The 802.11b cards operate with RTS/CTS disabled and all share the same 802.11b channel. The cards use o non-standard “pseudo-IBSS” mode , similar to standard 802.11b IBSS mode. The Pseudo-IBSS omits 802.11b’s beacons and BSSID (network ID).

Each Roofnet node runs identical turn-key software consisting Linux, routing software implemented in CLICK, a DHCP server and a webserver. The nodes are completely self-configuring, thus the software must automatically solve problems like allocating addresses, finding a gateway between Roofnet and the Internet, and choosing a good multi-hop route to that gateway.

The Roofnet carries IP packets inside its own header format and routing protocol. Each node will assign address automatically based on Ethernet address and unused class A IP address block. Only a small fraction of Roofnet users connected to Internet via wired Internet access links. The nodes connected to Internet advertise itself to Roofnet as an Internet gateway. These gateways act as a NAT for connections. Other nodes select the gateway to which it has the best route metric.

This network’s architecture favors ease of deployment by using omni-directional antennas, self-configuring software and link quality. The average throughput between nodes is 627 kbits/second. Since each internal node has multiple output gateway and can choose best current gateway, the Roofnet’s multi-hop mesh increase both connectivity and throughput.

Modeling Wireless Links for Transport Protocols

The wireless links’ intrinsic characteristics that affect the performance of transport protocols like variable bandwidth, corruption, channel allocation delays, and asymmetry. Congestion control in today’s Internet supposes that almost all packets losses result from congestion, however, packet losses on wireless links that are from corruption rather than congestion violates this assumption.

Three main types of wireless links are: cellular, WLAN and satellite. Topologies play an important role in wireless links. The number and location of wireless links in the path can affect the performance of transport protocols. Latencies, error loss rates of multiple wireless links add up, making loss recovery and congestion control more difficult.

Traffic models have significant effects on simulations results. Typically, mobile users transfer more data downlink than uplink. The performance metrics for evaluating performance include throughput, delay, fairness and dynamics. The first problem is the use of unrealistic models. The second problem occurs when the models are realistic but explore only a small corner of the parameter space.

Error losses can be modeled by dropping packets according to a per-packet, per-bit or time-based loss probability. The sudden increase in the delay can have a negative impact on transport protocols i.e. causes unnecessary retransmissions and false congestion control. Wireless links can introduce packet reordering during link-level error recovery using retransmission. They also often allocate channels based on the availability of user traffic. The bandwidth variation in wireless links can cause spurious timeouts and the asymmetry in bandwidth and latency can cause the congestion for TCP ACK. The mobility, an intrinsic property for most wireless links, presents a major challenge to transport protocols through the packet losses and delay introduced by handover.

In some cases, the wireless link level technologies have been design to take into account the performance requirement of the current TCP standard. In other cases, discussion is about the adaptation of transport protocols to the need of wireless links. The paper consider different models and parameters to be used in answering the questions and others about transport protocol performance over wireless links.