Research On Virtual Network Mapping Based On SDN

With the expansion of the scale of the Internet,now the network architecture is difficult to meet people's needs.Software defined network(SDN)is designed to replace traditional networks.The most notable feature of SDN is openness and programmability.SDN is developed and applied in the fields of network virtualization,data center network,wireless local area network,and cloud computing.Network virtualization technology can allow multiple heterogeneous networks to share in the same infrastructure network,which are independent of each other and not interfered.One of the biggest challenges of network virtualization technology is the virtual network embedding problem,that is,the virtual request of the service demanded by the user is efficiently embedded in the physical node and the physical path which satisfy the request resource constraint.At present,most researches don't consider the time dimension for virtual network mappings,and only make decisions about virtual network mappings based on the resource load of the physical network at present time.In practical applications,virtual network requirements of users have time constraints,the underlying network to allocate resources for virtual request according to the time of arrival.Due to the lifetime of request is uncertainty,the physical network will releases resources anytime in the future,which it makes the underlying physical resources dynamic changes.It leading to the underlying network resources load unbalance and affecting the embedding rate of virtual network requests.Therefore,we should consider the time factor in the study of virtual network mapping.When introducing the time dimension,we should give full consideration to mutual restrains of time and resources.We propose a virtual network embedding algorithm based on the load balancing of the underlying network resources.This paper proposes a two-dimensional discrete weighted model time and resource,and establishes a mathematical programming model for minimizing the degree of two-dimensional load balancing.We improve the two stage virtual network mapping algorithm which iterative processing after completed node and link mapping.It is easy to find that the algorithm can improve the acceptance rate of virtual request mapping and the load balance of the underlying network through a large number of simulation experimental data.Aiming at the problem that the path can be split in the virtual network mapping process which virtual network request is divided into smaller requests map to the physical network.Multipath mapping problem can be transformed into a multi commodity flow model.This paper proposes a multi-path virtual network mapping algorithm based on the spatio-temporal association with the aim of load balancing of the underlying physical resources when considering the time attribute of the virtual network embedding process.The model of multi-commodity flow based on time is solved by finding the effective path set on the underlying network,and satisfy the demand of constraint,such as bandwidth,node and time.Then the corresponding algorithm is proposed to solve the problem of virtual network mapping based on the model.Finally,simulation results show that the algorithm can maintain the load balance of the whole network,and has better receiving rate.