Research on Deployment of Service Function Chains under Hybrid Network Functions
Jiahui Xu, Hefei Hu · 2024
With the diversification of network users and services, network architectures have gradually become rigid. Network Function Virtualization (NFV) technology decouples network function (NF) from network devices to reduce excessive network expenses caused by frequent equipment replacements. However, in the short term, achieving complete NFV by removing all dedicated devices from the underlying network architecture remains impractical. The coexistence of dedicated hardware devices and commercial servers is an inevitable trend. In this paper, we investigate the deployment of service function chains (SFC) for hybrid network functions, where physical devices and commercial servers collaborate to provide services. The challenge in such a hybrid network lies in minimizing SFC deployment costs while meeting user demands, leveraging the different characteristics of dedicated hardware devices and servers. We find that this problem has not been satisfactorily addressed to date. Therefore, we formulate the optimization problem as an integer linear programming (ILP) problem and designs an enhanced Particle Swarm Optimization Algorithm (PSOA). In the initialization phase of the particles, the algorithm employs resource prediction to determine the types of nodes to which network functions should be mapped. Furthermore, it distinguishes between different types of nodes by utilizing updates from various domains. Experimental validation demonstrates that the proposed algorithm effectively reduces the complexity of the hybrid network SFC deployment problem and establishes its superiority.