Online VNF Placement using Deep Reinforcement Learning and Reward Constrained Policy Optimization

Appropriate deployment of Virtual Network Functions (VNFs) within the network infrastructure to optimize resource utilization, while fulfilling performance criteria, is critical to Network Function Virtualization (NFV). However, traditional optimization methods frequently struggle to deal with the dynamic and complex nature of the VNF placement problem. Therefore, this study proposes an online VNF placement approach that addresses these problems by leveraging Deep Reinforcement Learning (DRL) and Reward Constrained Policy Optimization (RCPO). This technique combines DRL’s adaptability with RCPO’s constraint incorporation capabilities, ensuring that the learned policies satisfy performance and resource restrictions while optimizing the overall VNF placement. Hence, the VNF placement problem is formulated as a constrained optimization problem and adapted to the DRL setting. We demonstrate that our proposed framework outperforms certain state-of-the-art VNF placement techniques when it comes to resource utilization and constraint satisfaction.