Routing Protocols for Meshed Communication, Networks Targeting Communication Quality of Service (QoS) in Rural Areas

摘要

Rural areas in Africa often have poor telecommunication infrastructure. Mobile phones, if available, are frequently unaffordable to most users. Wireless mesh networks (WMNs) offer an alternative possibility of low cost voice and data communications.The focus of this research is a laboratory study of WMNs that mimic conditions found in rural areas. This work investigates routing strategies for the Mesh Potato (MP). The MP is an effective alternative communication technology that has minimal configuration requirements, low cost of deployment, low power consumption and resilience that make it an attractive choice for rural areas. The MP runs a new mesh networking algorithm called the better approach to mobile ad hoc networking (B.A.T.M.A.N or Batman). This allows a WMN to be established in which users can use plain old telephones to talk to each other using Voice over IP (VoIP).Batman daemon (Batmand) is the implementation of Batman algorithm used by the MP. Batmand is a minimalistic routing protocol which performs well in laboratory experiments. The question raised is whether adding more service specific routing metrics improve the quality of service (QoS) observed in Batmand network in practice. The research investigates delay, packet loss, throughput and jitter as performance parameters (metrics) that may serve as options to improve the simplistic Batman algorithms route selection process. These metrics are essential for QoS in voice- and data-sensitive networks. Specific focus was given to delay and it is the metric added to Batmand. In addition the research examines how well the different applications such as voice and data are supported on the Batmand network under different routing scenarios.The research approach adopted in this dissertation was experimental and an indoor testbed was created to replicate the basic scenarios encountered in the rural environment. The essential characteristics found in the Mdumbi region of the Eastern Cape, South Africa, were taken as a case study in this dissertation. The testbed was used to compare the original Batman algorithm implemented as Batmand, referred to here as O-Batmand, routing protocol and the resultant Batmand version obtained from the addition of the delay-routing metric called modified Batmand (M-Batmand).The research produced a number of findings. As the number of hops increased the per-formance of the network decreased for both protocols. O-Batmand is well suited for the task of routing packets inside a wireless network. It is designed and works for voice packets and supports data services. This is also true for the M-Batmand implementation.M-Batmand was developed as an improvement to the O-Batmand implementation at thecost of increased complexity, experienced by the protocol through modifications of its route selection process. The modification involved adding network delay values to its route selection process. This addition resulted in a protocol that is delay sensitive; however, the overall performance gains were inexistent. The main conclusions drawn from this study are that O-Batmand cannot be modified to include additional metrics and be expected to improve its performance. Second conclusion is that M-Batmand did not improve the overall performance of the O-Batmand protocol. The addition of the delay metric actually hindered O-Batmand's performance to the extent that no overall performance gains were realised. Sources of performance degradations are: increased overhead, from added delay data, in the network control packets called originator messages (OGMs). M-Batmand performs calculation which O-Batmand did not increasing CPU cycle needs. Lastly upon further internal protocol investigation it is seen that the rate of route delay data updates is slower than the original metric used by the protocol. This creates route fluctuations as route selection process will change when the updated delay values are added and change again when there are not as the network obtains the updated delay data.Both protocols support voice and data, however, the results show that the quality of the network deteriorates in the testbed with increasing hops. This affects voice more so then it does data as routes become more unstable with each increasing hop. Further Batmand is best at supporting voice and data as it outperforms M-Batmand in the laboratory experiments conducted. This dissertation argues that while there may exist one or a combination of metrics amongst the researched list (delay, packet loss, throughput and jitter) that may actually improve the performance of the protocol, it is extremely hard to realize such gains in practice.