Research On Key Technologies For Virtual Network Embedding In Dynamic Environments

The core idea of network virtualization lies in constructing multiple virtual heterogeneous networks on the same physical network infrastructure.As one of the core techniques in network virtualization,virtual network embedding has attracted lots of attention from both academic and industry.Virtual networking embedding is the technique to allocate corresponding physical network resource for virtual networks given various resource constraints on nodes and links.Most of the recent studies focus on the embedding scheme in static scenario,while the dynamic characteristics has been neglected.Thus,stemming from the dynamic characteristics concerning the requests' arrival,the electricity supply market and the optimization,this thesis studies the virtual network embedding problem in the three scenarios.The main goal is to leverage these dynamic characteristics to further improve the utility and reduce the energy consumption of the physical network provider.The main contributions of this work are listed below:· Proposing an artificial bee colony based virtual network embedding algorithm to solve the virtual network embedding problem given the condition of dynamic virtual network's arrival.It leverages the iteration and intelligence of artificial bee colony to achieve a better virtual network embedding solution.Specifically,it redefines the parameters and operations in artificial bee colony algorithm to adapt to our addressed problem.Two strategies are promoted to further optimize the performance of the artificial bee colony based VN embedding algorithm.The preferential exchange strategy accelerates the speed of convergence.The pheromone and sensitivity model strategy avoids local optimum limitation.Extensive simulations show that the proposed algorithm can obtain nearly 35.4%more revenue than the state-of-the-art algorithm.·Proposing a clustering based low electricity cost virtual networkembedding algorithm given the condition of dynamic electricity price.It takes both of the electricity price in inter-domain virtual network embedding and energy consumption in intra-domain virtual network embedding into consideration to optimize the energy cost.Specifically,based on the topology of the VN it first generates a weighted graph for capturing the affinity between any two virtual nodes on the aspects of the electricity price and topology.It then employs local clustering algorithm to cluster the virtual nodes based on the affinity graph and performs the inter-domain VN embedding based on such clusters.It embeds each cluster to the physical network in one domain.It finally leverages the energy-aware nodes and link embedding algorithm to perform the intra-domain embedding.Extensive simulations demonstrate that our algorithm saves up to 14%energy cost for the physical network than the existing algorithm while obtaining at attractive revenues.·Proposing a genetic algorithm based virtual network migration algorithm given the condition of dynamic re-optimization.First the virtual network migration model is established,which analyzes the virtual network migration,quantifies the communication traffic cost of the physical network brought by the virtual network migration,investigates the effect of the migration process on the virtual network service,and establishes the total overhead model caused by the virtual network migration.Then,based on the above model,it proposes a genetic algorithm based virtual network migration algorithm.Bearing the advantages of the genetic algorithm,certain personalized strategies such as encoding,population initialization,crossover and mutation,are introduced to better leverage the genetic algorithm in virtual network migration.The experimental results show that the virtual network migration approach based on genetic algorithm can further reduce the energy consumption of physical network providers.