Cooperation in open, decentralized, and heterogeneous computer networks

Community Networks (CN) are naturally open and decentralized structures, that grow organically with the addition of heterogeneous network devices, contributed and configured as needed by their participants. The continuous growth in popularity and dissemination of CNs in recent years has raised the perception of a mature and sustainable model for the provisioning of networking services. However, because such infrastructures include uncontrolled entities with non delimited responsibilities, every single network entity does indeed represent a potential single-point of failure that can stop the entire network from working, and that no other entity can prevent or even circumvent. Given the open and decentralized nature of CNs, that brings together individuals and organizations with different and even conflicting economic, political, and technical interests, the achievement of no more than basic consensus on the correctness of all network nodes is challenging. In such environment, the lack of self-determination for CN participants in terms of control and security of routing can be regarded as an obstacle for growth or even as a risk of collapse. To address this problem we first consider deployments of existing Wireless CN and we analyze their technology, characteristics, and performance. We perform an experimental evaluation of a production 802.11an Wireless CN, and compare to studies of other Wireless CN deployments in the literature. We compare experimentally obtained throughput traces with path-capacity calculations based on well-known conflict graph models. We observe that in the majority of cases the path chosen by the employed BMX6 routing protocol corresponds with the best identified path in our model. We analyze monitoring and interaction shortcomings of CNs and address these with Network Characterization Tool (NCT), a novel tool that allows users to assess network state and performance, and improve their quality of experience by individually modifying the routing parameters of their devices. We also evaluate performance outcomes when different routing policies are in use. Routing protocols provide self-management mechanisms that allow the continuous operation of a Community Mesh Network (CMN). We focus on three widely used proactive mesh routing protocols and their implementations: BMX6, OLSR, and Babel. We describe the core idea behind these protocols and study the implications of these in terms of scalability, performance, and stability by exposing them to typical but challenging network topologies and scenarios. Our results show the relative merits, costs, and limitations of the three protocols. Built upon the studied characteristics of typical CN deployments, their requirements on open and decentralized cooperation, and the potential controversy on the trustiness of particular components of a network infrastructure, we propose and evaluate SEMTOR, a novel routing-protocol that can satisfy these demands. SEMTOR allows the verifiable and undeniable definition and distributed application of individually trusted topologies for routing traffic towards each node. One unique advantage of SEMTOR is that it does not require a global consensus on the trustiness of any node and thus preserves cooperation among nodes with even oppositional defined trust specification. This gives each node admin the freedom to individually define the subset, and the resulting sub-topology, from the whole set of participating nodes that he considers sufficiently trustworthy to meet their security, data-delivery objectives and concerns. The proposed mechanisms have been realized as a usable and open-source implementation called BMX7, as successor of BMX6. We have evaluated its scalability, contributed robustness, and security. These results show that the usage of SEMTOR for securing trusted routing topologies is feasible, even when executed on real and very cheap (10 Euro, Linux SoC) routers as commonly used in Community Mesh Networks.