A Node Probability-based Reinforcement Learning Framework for Virtual Network Embedding

At present, the traditional heuristic method to solve the problem of virtual network embedding (VNE) is still the mainstream. In the environment of network virtualization (NV), a more efficient VNE algorithm is needed to serve the construction of smart city. Using heuristic algorithm to solve the problem of VNE does not meet its development requirements. In this paper, a VNE algorithm based on node probability is proposed by using reinforcement learning (RL) algorithm. The algorithm extracts three attributes of each substrate node to form a feature matrix, which is used as the input of the policy network to train the agent. The purpose is to deduce the mapping probability of each node and rank the base nodes according to this probability, then embed the virtual nodes in this order. Finally, the breadth first search (BFS) strategy is used to map the links. Simulation results show that our algorithm is superior to a representative algorithm based on node ranking in terms of the acceptance rate of virtual network requests (VNR), long-term revenue consumption ratio and long-term average revenue.