Virtual Network Function Selection and Chaining based on Deep Learning in SDN and NFV-Enabled Networks

Software-Defined Network (SDN) and Network Function Virtualization (NFV) make Service Function Chain (SFC) become a popular service paradigm. For an SFC Request (SFCR), the flow needs to be routed to traverse a series of specified Virtual Network Functions (VNFs) in predefined order. The key challenge is how to make an optimal VNF selection and chaining approach for an SFCR in multi-instance environment. Moreover, as the SDN controller should achieve centralized network control over the state of data plane elements, it is an urgent need to design a high-performance routing strategy to reduce the routing computation time. In the paper, we formulate the VNF selection and chaining problem as a Binary Integer Programming (BIP) model with the aim to minimize end-to-end delay. Then, a novel deep learning-based strategy is proposed to solve the problem by intelligent routing learning and prediction. Performance evaluation shows that the deep learning-based strategy can obtain high network performance in SFCR acceptance rate and end-to-end delay and enhance the time efficiency of routing computation compared with existing approach.