Adaptive-VNE: A flexible resource allocation for virtual network embedding algorithm

In this paper, we propose a new dynamic adaptive virtual network resource allocation strategy named Adaptive-VNE to deal with the complexity and the inefficiency of resource allocation. The proposal coordinates virtual node and virtual link mapping stages. The main idea behind the proposal is take advantage of unused bandwidth with respect to the occupancy rate of embedded virtual links. Hence, the unused bandwidth will be reassigned to incoming virtual network requests. To do so, Adaptive-VNE adopts the “divide and conquer” strategy. It divides the virtual network request topology into many star topologies. Then, the mapping of each piece within the whole topology is formulated as a K-supplier problem and resolved by an approximation bottleneck algorithm. To generate the global virtual network topology, Adaptive-VNE uses a backtracking algorithm in order to minimise the global mapping cost. Note that the proposal forecasts usage rate of virtual links and adapts their bandwidth reservation. Adaptive-VNE was validated by simulations and compared to the related strategies found in literature. The results obtained show that, contrarily to static bandwidth allocation approaches, the adaptive strategy maximises substrate bandwidth usage while the virtual links' bottleneck rate is minimised. Moreover, the congestion periods are minimised and during the bottleneck the bandwidth satisfaction is maximised. Finally, Adaptive-VNE improves performances in terms of acceptance rate of virtual networks and revenue of infrastructure providers.