<p>The rapid proliferation of 5G networks, edge computing, and IoT applications raises stringent requirements on end-to-end latency, energy efficiency, and the operational stability of service deployments. This paper addresses a core network and systems management task: the automated orchestration of Virtual Network Functions (VNFs) that compose Service Function Chains (SFCs) over distributed edge infrastructures. We propose a constrained optimization model that supports different management goals through alternative objective functions (latency-aware or power-aware) while enforcing operational constraints, including node capacities, slice-specific latency bounds, and explicit limits on VNF migrations/relocations between scheduling periods. We develop two complementary solution approaches: (i) an exact Integer Linear Programming (ILP) formulation to compute optimal placements, and (ii) a Reinforcement Learning (RL) framework as a learning-based alternative for fast online decision-making. To improve responsiveness under time-varying demand, we incorporate predictive intelligence for workload forecasting and proactive resource estimation. Using a realistic 5G Core Network slicing use case driven by real traffic traces and detailed latency/power models for edge nodes, our experiments show that goal-driven strategies substantially outperform baseline placements under practical constraints, achieving energy savings of up to 18% while limiting service latency.</p>