Contribución a la optimización del ancho de banda en redes MPLS

Résumé

Multiprotocol Label Switching (MPLS) architecture is obtaining a remarkable success, thanks to its ability of coexist with all networks protocols, specifically for its compatibility with all networking technologies. Nowadays data traffic, specifically voice and multimedia, as well as security and redundancy mechanism needs a huge bandwidth with quality of service (QoS) guaranties. These requirements cause congestion in networks speed and bandwidth resources. The objective of present work is resolving some problems of bandwidth optimization in MPLS networks. In the research of these solutions the milestones are cost minimization and flow allocation to obtain network optimum load balancing. One drawback of MPLS is its scalability to manage high traffic flows, this drawback maybe solved using techniques from Multicommodity Network Flows (MCNF), Minimum cost theory and Maximum throughput. From the study of these techniques arise new paradigms to resolve optimization problems in networks as MPLS and GMPLS, because these techniques consider link cost as a parameter in bandwidth utilization, which allows to guarantee and to improve QoS of the network. Another way to guarantee QoS is using protection strategy of MPLS networks, being the most used pre-planned strategy because it is the best for error recovery. However, in case of none failure in network it produce a underutilization of resources, specifically in reservation of spare bandwidth, that represents a channel portion not used during transmission. This portion of bandwidth represents one of the problems to solve using use of links minimization, obtaining an optimal solution for distributed LSP (D-LSP) in MPLS networks, the majority of these problems are resource minimization with NP complete complexity. On the other hand we establish different models that permit link congestion minimization, implementing new characteristics that enhance QoS by congestion control.