Performance analysis of mesh networks in indoor and outdoor wireless testbeds

Abstract

Physical indoor wireless network testbeds as well as outdoor wireless testbeds have the potential to accelerate the pace of research in the field of wireless ad hoc and mesh networking. They form part of a critical chain of steps needed to develop and test ad hoc networking protocols from concept to eventual uptake by industry. Current research in this area makes use of simulations or mathematical models which oversimplify the physical and Medium Access Control layer. In Africa specifically, wireless mesh networking has the potential to make a substantial impact on the lack of telecommunications infrastructure across the continent. A combination of good theoretical analysis, indoor test facilities and rural testbeds forms a perfect suite to carry out meaningful research in the field. A 7x7 wireless grid of closely spaced computers was constructed, making use of highly attenuated 802.11 radios running in ad hoc mode. Modelling and analysis revealed that a suitably attenuated environment was created with variation in signal strength between node pairs following a Gaussian distribution. This emulates a real outdoor network with normal signal propagation issues such as multi-path fading and lack of Fresnel zone clearance. This testbed was then used to evaluate 3 popular MANET ad hoc routing protocols, namely AODV, DYMO and OLSR. OLSR was tested with the standard hysteresis routing metric as well as the ETX routing metric. OLSR showed the best performance in terms of average throughput and packet loss for a medium size (21 node) and large (49 node) mesh network, with the hysteresis routing metric performing best in large networks and ETX performing best in medium sized networks. DYMO also performed very well, considering its low routing overhead, exhibiting the least amount of delay in a large mesh network (49 nodes). The AODV protocol showed the weakest performance in the grid with close to 60% of possible link pairs achieving no route in a 49-node grid. However, it did present the least amount of routing overhead compared with other routing protocols. Finally, a medium-sized rural mesh network testbed consisting of 9 nodes was built in a mountainous area of about 15 square kilometers around an AIDS clinic using the OLSR routing protocol with ETX as the routing metric. The network provided a good service to the satellite-based Internet with throughput rates ranging between 300 kbps for 4 hops and 11000 kbps for 1 hop and an average throughput rate of 2324 kbps. To encourage fair sharing of Internet connectivity, features were installed to limit each user to 40 MB/month of free Internet traffic. A local web server offers cached pages of Wikipedia and Linux repositories to reduce the need for Internet access. VoIP services were also installed between clinic infrastructure to reduce the the need for making expensive GSM calls. It was shown that a mesh network of this size provides a very satisfactory level of broadband service for users accessing a satellite-based Internet facility as well as local VoIP services.