Topologías para la distribución de contenidos en redes de comunicación

The peer-to-peer (P2P) overlay networks are distribution topologies that take advantage of the ability that their nodes have to forward the information that they receive. For the particular case of video transmission using P2P networks, known as streaming P2P, there are implementations based on BitTorrent protocol data dissemination for files downloading. These networks use a pseudorandom topology which makes difficult to determine both the size of the network -defined as the maximum number of nodes which can receive information with a given quality of service- and the distribution delay. Despite this lack of definition, in the current thesis we present a structured network in which the nodes establish both incoming and outgoing connections according to a deterministic joining algorithm. That topology allows to calculate exactly the number of nodes that may join the network and that in stable conditions may get the flow with guaranteed quality of service, and also one very fitted bound for the total distribution delay. In our case, we have defined networks which contain a single source node with a bandwidth weater than the bandwidth of the information flow, and also a set of nodes with a bandwidth lower than the bandwidth of the flow. This constraint is used to distinguish the high capacity nodes from the terminal nodes which either have physical limitations due to their access link bandwidth or do not consider to allocate all their bandwidth to the forwarding of the information flow. In fact, high capacity nodes can be considered sources either because they produce the flow physically or because, receiving it from another distribution network, they forward it as source to the low bandwidth nodes that belong to the P2P network. Once the distribution topology is described, we define the mathematical model for its analysis. With this purpose, we normalize the flow bandwidth and we define a network joining algorithm with the aim of optimizing distribution delay. Following, the network topology is characterized by means of three parameters: the number of nodes that can be served directly by the source, the number of download connections that must arrive to a node to guarantee the reception at unitary rate, and the maximum number of upload connections that a node can establish. Likewise, the information flow is segmented into generic information units called objects, which are transmitted into the network by means of a certain set of dissemination trees. The number of dissemination trees corresponds to the number of download connections of a node. With this background material, we calculate the maximum number of peers which can be connected, under stable conditions, to the network, and also a very fitted bound for the transmission delay. Besides, we present a simulation that allows to evaluate the performance of the described topology in the field of VANET networks. Finally, in the last chapter we present a proposal for application-layer multicast distribution network which can be used to send the information flow from the node that initially produces it to the source nodes of the P2P distribution network that we have described.