No abstract

Modern large-scale networked services, such as video streaming, are typically deployed at multiple locations in the network to provide redundancy and load balancing. Different techniques are used to provide performance monitoring information so that client nodes can select the best service instance. One of them is collaborative sensing, where clients share measurement results on the observed service performance to build a common ground of knowledge with low overhead. Clients can then use this common ground to select the most suitable service provider. However, collaborative algorithms are susceptible to false measurements sent by malfunctioning or malicious nodes, which decreases the accuracy of the performance sensing process. We propose Sense-Share, a simple light-weight and resilient collaborative sensing framework based on the similarity of the client nodes' perception of service performance. Our experimental evaluation in different topologies shows that service performance sensing using Sense-Share achieves, on average, 94% similarity to non-collaborative brute force performance sensing, tolerating faulty nodes. Furthermore, our approach effectively distributes the service monitoring requests over the service nodes and exploits direct inter-node communication to share measurements, resulting in reduced monitoring overhead.

Community wireless mesh networks have emerged as cooperative initiatives to provide Internet Access in areas where traditional ISP costs are not affordable for the population. It is common in wireless mesh networks sharing several capacity limited Internet gateways to provide Internet access. As routing does not handle capacity planning, end-users have to select gateways in such a way that the overall capacity of all gateways could be used effectively. An efficient gateway selection should minimize the processing logic and measurements over the mesh network. Selecting a high performance gateway can also ensure that the overall network load is balanced. This paper presents RIMO, a standalone best-effort algorithm for client nodes to select their preferred gateway without interacting with other client nodes. RIMO-based selection matches the gateway performance of the reference brute-force and omniscient algorithms for 60% of the test duration while reducing the gateway performance measurement cost from a factor of n to 2. With a reduced overhead and high efficiency, the RIMO algorithm automates the aggregation of multiple Internet gateways in wireless mesh networks, which results in robust last mile Internet connectivity to people in vulnerable situation.

No abstract

Internet access is still unavailable to one-third of the world population due to the lack of infrastructure, high cost, and the digital divide. Many access-limited communities opt for shared Internet access where they build common network infrastructures to mitigate the cost. Internet connectivity in such infrastructures is typically provided by several limited, sometimes non-dedicated, gateways. Client nodes, i.e., the end-user hosts, use one gateway and switch to another when the first fails. In this scheme, the gateway configuration is done manually on the end-user side. This form of Internet connectivity is widespread and has the advantage that no central control is required, but it is also unreliable and inefficient due to several factors, such as unbalanced traffic load across the gateways. There is no doubt that the network would benefit from a gateway selection mechanism that can provide good connectivity to the client node as well as balanced load distribution and a dynamic adaptation to the current network state. However, providing such a dynamic gateway selection is complicated: since the perceived performance of the gateways changes frequently and might depend on the location of the client node in the network, and optimal selection would require the continuous monitoring of the gateway performance by the client node. The cost of such network-wide performance monitoring is high in large-scale networks and can outweigh the benefits of the dynamic gateway selection. The thesis's goal is to design a low-cost, distributed mechanism that provides an efficient and dynamic gateway selection while considering the overall balanced gateway selection distribution. To this end, we have split the problem of gateway selection into different sub-problems. First, we focus on reducing the cost of gateway performance monitoring. We propose an approach to reduce the number of monitoring requests generated by each node and analyze its effect on the gateway selection. Then, we present a collaborative monitoring method that allows neighbor nodes to share the load of the gateway monitoring. We show that every node can carry out the necessary tasks: performance monitoring, collaboration with its neighbors, and fault tolerance measures, with little computation and communication overhead. Second, to improve the gateway selection, we focus on making a selection decision that fulfills the individual performance requirements of the client nodes as well as global load balancing requirements. The solutions developed by us for the different sub-problems are embedded into a general and extensible, layered framework for gateway selection that we have called the Sense-Share-Select framework. Experimental validation and comparison with existing methods show that our framework provides accurate collaborative performance monitoring, improves the QoE for the nodes, and distributes the client nodes over the gateways in a balanced manner. The simplicity and flexibility of the framework make it adaptable to other network domains such as IoT networks and other scenarios where resource monitoring and load balancing are required. El acceso a Internet aún no está disponible para un tercio de la población mundial debido a la falta de infraestructuras, el alto costo y la brecha digital. Muchas comunidades con acceso limitado optan por el acceso compartido a Internet donde construyen infraestructuras de red comunitaria para mitigar el costo. La conectividad a Internet en dichas infraestructuras suele estar a cargo de varias puerta de enlaces limitadas en recursos, y a veces no dedicadas. Los nodos de cliente, es decir, los hosts de usuario final, utilizan una puerta de enlace y cambian a otra cuando falla la primera. En este esquema, la configuración de la puerta de enlace se realiza manualmente en el lado del usuario final. Esta forma de conectividad a Internet está muy extendida y tiene la ventaja de que no se requiere un control central, pero tampoco es confiable y eficiente debido a varios factores, como una carga desequilibrada de tráfico a través de las puertas de enlace. No hay duda de que la red se beneficiaría de un mecanismo de selección de pasarela que pueda proporcionar una buena conectividad al nodo cliente, así como una distribución equilibrada de la carga y una adaptación dinámica al estado actual de la red. Sin embargo, proporcionar una selección de puerta de enlace tan dinámica es complicado: dado que el rendimiento percibido de las puertas de enlace cambia con frecuencia y podría depender de la ubicación del nodo cliente en la red, y la selección óptima requeriría la supervisión continua del rendimiento de la puerta de enlace por parte del nodo cliente. El costo de dicha supervisión del rendimiento en toda la red es muy alto en redes de gran escala y puede superar los beneficios de la selección de puerta de enlace dinámica. El objetivo de la tesis es diseñar un mecanismo distribuido de bajo costo que proporcione una selección de puerta de enlace dinámica y eficiente al tiempo que considera la distribución general de selección de puerta de enlace equilibrada. Con este fin, hemos dividido el problema de la selección de la puerta de enlace en diferentes subproblemas. Primero, nos enfocamos en reducir el coste del monitoreo del rendimiento de la puerta de enlace. Proponemos un enfoque para reducir la cantidad de solicitudes de monitoreo generadas por cada nodo y analizar su efecto en la selección de la puerta de enlace. Luego, presentamos un método de monitoreo colaborativo que permite a los nodos vecinos compartir la carga del monitoreo de la puerta de enlace. Demostramos que cada nodo puede realizar las tareas necesarias: monitoreo del rendimiento, colaboración con sus vecinos y medidas de tolerancia a fallas, con poca sobrecarga de cómputo y comunicación. En segundo lugar, para mejorar la selección de la puerta de enlace, nos centramos en tomar una decisión de selección que cumpla con los requisitos de rendimiento individuales de los nodos del cliente, así como con los requisitos de equilibrio de carga global. Las soluciones desarrolladas por nosotros para los diferentes subproblemas están integradas en un marco general y extensible en capas para la selección de puertas de enlace que hemos llamado el marco Sense-Share-Select. La validación experimental y la comparación con los métodos existentes muestran que nuestro marco proporciona un monitoreo de rendimiento colaborativo preciso, mejora la QoE para los nodos y distribuye los nodos del cliente a través de las puertas de enlace de manera equilibrada. La simplicidad y flexibilidad del marco lo hacen adaptable a otros dominios de red, como las redes de IoT y otros escenarios donde se requiere monitoreo de recursos y equilibrio de carga.