Research On Dynamic Collaborative Virtual Network Embedding Based On Resource Importance Measure

With the emergence of technologies such as cloud computing and big data,how to meet diverse business needs has become an urgent problem of the Internet.Network virtualization allows multiple heterogeneous virtual networks to be created and coexist over shared substrate network infrastructures,thereby meeting different business needs and improving resource utilization efficiency.The virtual network embedding is one of the core issues of network virtualization.Its main goal is to embed the virtual network requests with resource constraints of nodes and links from different tenants onto the substrate network,so as to build independent network environments for multiple tenants while creating as much revenue as possible for substrate network operators.Existing researches for virtual network embedding mainly focused on maximizing the probability of in-progress request to be accepted,while ignored the impact on latter virtual network requests.Such a greedy embedding strategy lacks coordination mechanism of resource allocation between different virtual network request embeddings,leading that the former requests greedily occupy too much critical node and link resources,the network resources may be scattered and difficult to use,which reduce the acceptance ratio of the latter virtual network requests,that is resource fragmentation problem.Some researchers put forward the collaborative virtual network algorithm that considers the coordination between different request embeddings.However,it only considers the local topological attributes to measure the resource importance,which cannot reflect the importance of resources in the global network,resulting in limited coordination and cannot solve the problem well.To solve the above problem,in this paper,we propose a virtual network embedding algorithm with coordination of node and link resource based on global resource importance measure.Firstly,the node importance and link importance in the global network are measured by topology-aware methods,which are used to set embedding cost of substrate resources,the more important resource may have a higher price.Then,we define the virtual network embedding as a linear programming to minimize the total embedding cost.Our proposed virtual network embedding algorithm could avoid the critical resources from being excessively used by former requests and be able to satisfy the latter ones with larger demands,which will bring about better coordination between different request embeddings so as to improve the virtual network request acceptance ratio and revenue.With the arrival and departure of virtual network requests,the problems of load imbalance and resource fragmentation could be more serious.Existing researches balanced the loads through dynamic reconfiguration,while they didnot distinct the different impact on network connectivity between the loads mapping on different substrate resources,which could cause new resource bottlenecks.In this paper,we propose a dynamic reconfiguration mechanism based on critical nodes protection:when the resource fragmentation reaches a certain level,some of the virtual nodes embedded on the critical substrate nodes where the load is more than a certain threshold will be re-embedded to the less-critical ones so as to reduce the degree of substrate network fragmentation.In this paper,we firstly analyze the current researches and the existing problems of the virtual network embedding and dynamic reconfiguration algorithms.Then the proposed collaborative virtual network embedding algorithm based on resource importance measure and the dynamic reconfiguration mechanism are described in detail.Finally,we introduce the design of simulation experiments and analysis the experimental results.Simulation results show that the proposed algorithms increase the acceptance ration and long-term average revenue.