GEOVA: A Genetic Algorithm for Energy-Efficient VNF Placement in O-RAN Networks

This paper proposes GEOVA (Genetic Energy-aware O-RAN VNF Allocator), a batch Genetic Algorithm for the joint placement of access functions - virtualized Distributed Unit (vDU) and virtualized Centralized Unit (vCU) - and service chains in Open Radio Access Network (O-RAN) deployments. The algorithm minimizes the energy consumption of the computational nodes hosting the VNFs, subject to Service Level Agreement (SLA) constraints on latency, capacity, and isolation. Each chromosome jointly represents the placement of all scenario demands, which allows the genetic operators to consolidate VNFs on a reduced number of active nodes. The evaluation considers two topologies, POLSKA with 12 nodes and POPS with 27 nodes, under different load intensities. Under low load, GEOVA achieves up to 13.7 percentage points of additional acceptance and reduces the energy per accepted demand by up to 16.5%. Under saturation, it cuts total energy consumption by 1.3% at the cost of approximately 2 percentage points in acceptance, exposing an explicit trade-off between energy efficiency and admission rate.