Progressive virtual topology embedding in OpenFlow networks

Future internet would provide a flexible and simpler architectural design by combining novel low level clean-slate techniques such as OpenFlow with high level design principles such as network virtualization. However, previous research outlined that at the core of network virtualization stands a new set of challenges for network resources allocation. In this work we focus on one such challenge, namely the problem of virtual topology embedding. In this context users need to leverage the infrastructure substrate by accomodating logical topologies with high degree of flexibility. The network provider, on the other hand, aims at maximizing its revenue in term of size and number of topologies accepted while minimizing costs accounting for the substrate network resources used. To this aim, we present VT-Planner a novel virtual network embedding technique with reduced computational cost and very efficient over substrate topologies encountered in practice. Extensive numerical simulations are provided comparing this technique with state-of-the-art solutions: our results show that VT-Planner is able to achieve a good balance in terms of complexity and performance.