Multiagent Deep Reinforcement Learning for Cost- and Delay-Sensitive Virtual Network Function Placement and Routing

This paper proposes an effective and novel multi-agent deep reinforcement learning (MADRL)-based method for solving the joint virtual network function (VNF) placement routing (P&R;), where multiple service requests with differentiated demands are delivered at same time. The of reflected by their delay- cost-sensitive factors. We first construct a VNF P&R; problem to jointly minimize weighted sum delay resource consumption cost, which is NP-complete. Then, decoupled into two iterative subtasks: subtask subtask. Each consists concurrent parallel sequential decision processes. By invoking deterministic policy gradient technique, MADRL-P&R; framework designed perform subtasks. new joint reward internal rewards mechanism proposed match goals constraints also propose parameter migration-based model-retraining deal changing topologies. Corroborated experiments, superior its alternatives in terms cost delay, offers higher flexibility personalized demands. can efficiently accelerate convergence under moderate topology changes.