SFC Mapping Algorithm Based on Prediction and Node Resource Ability Awareness
Jiahong Lai, Fan Yang, Wenchao Song, Chaoran Ying · 2023
Service function chaining (SFC) is an important application in network function virtualization (NFV). In this paper, we propose a SFC mapping algorithm based on prediction and node resource ability awareness (SMPRA) to solve the load balancing problem in dynamic SFC mapping. In order to prioritize the deployment of virtual network functions (VNFs) in resource-rich nodes, we adopt three approaches. First, we price the usage cost of a node based on its remaining resources. The less remaining resources a node has, the higher its usage cost. Second, we use deep learning to predict the node's resource status after a number of time steps, and calculate the remaining resources of a node taking into account not only the current node's resource status but also the remaining resources after a number of time steps. Third, we employ the PageRank method, by which the remaining resources of nodes are combined with the importance of nodes in the network topology, to evaluate of node resource more accurately. For our proposed algorithm to be applicable more widely, we support the dependency between the bandwidth change factor and the VNF in the deployment of the SFC. Simulation results show that the proposed SMPRA algorithm can reduce mapping costs and load imbalance effectively compared with the conventional MRACM algorithm.