Slice-Level Energy Model

Establishes a direct mathematical relationship between allocated virtual resources and physical power draw. The model maps Physical Resource Blocks (PRBs), CPU cores, and memory utilization to wattage consumption. This granularity enables precise energy attribution per slice rather than relying on coarse, node-level averages.

Models power consumption as a non-linear function of traffic load. A base station's power amplifier efficiency varies with load, and the model captures this dynamic. Key parameters include:

• Static power: Baseline consumption with zero traffic (cooling, control plane)
• Load-dependent power: Variable consumption scaling with PRB utilization
• Sleep mode thresholds: Load levels triggering component deactivation

Integrates Service Level Agreement (SLA) parameters as constraints on energy optimization. A URLLC slice has strict latency bounds limiting sleep mode depth, while an eMBB slice tolerates aggressive power-saving. The model quantifies the energy cost of maintaining specific KPIs like guaranteed bit rate and maximum latency.

Aggregates energy consumption across all network domains spanned by the slice:

• RAN: Power amplifiers, baseband units, MIMO antenna paths
• Transport: Optical transceivers, switches, backhaul links
• Core: UPF instances, CNF pods, data center PUE overhead The model provides a holistic, end-to-end energy footprint rather than siloed domain metrics.

Consumes streaming metrics from the Network Data Analytics Function (NWDAF) and O-RAN interfaces. Inputs include per-cell PRB utilization, per-CNF CPU consumption, and active MIMO layers. This data pipeline enables the model to reflect current operational state rather than relying on static assumptions or periodic snapshots.

Extends the model with time-series prediction to anticipate future energy consumption based on forecasted traffic patterns. This enables proactive sleep mode coordination and energy-aware slice admission control. The model can predict a slice's power draw minutes in advance, allowing the orchestrator to pre-scale resources or shift loads to greener energy windows.

A policy-driven function that steers user equipment to the most energy-efficient network slice instance available that can still satisfy the requested service requirements. This mechanism directly consumes the output of a Slice-Level Energy Model to make real-time routing decisions, minimizing the overall network power footprint without violating Slice SLA guarantees.

A centralized control strategy that synchronizes the activation of low-power states across multiple network components within a slice. The Slice-Level Energy Model provides the predictive load data necessary to determine optimal sleep durations for carriers, MIMO paths, and processing elements, maximizing energy savings without violating Guaranteed Bit Rate (GBR) commitments.

A power management technique that dynamically adjusts the clock frequency and supply voltage of a processing element in real-time. The Slice-Level Energy Model correlates computational load with energy draw, enabling the Slice Orchestrator to apply DVFS to Cloud-Native Network Functions (CNFs) during low-utilization periods.

The ability of a network slice to dynamically scale its allocated virtualized resources up or down in response to real-time workload fluctuations. The Slice-Level Energy Model quantifies the power consumption delta associated with scaling events, allowing the Slice Orchestrator to make cost-aware elasticity decisions that balance performance and energy efficiency.

A 5G core network function that collects and analyzes network data using machine learning to provide predictive analytics on slice load and performance. The NWDAF serves as the primary data pipeline feeding the Slice-Level Energy Model with real-time telemetry on traffic patterns, resource utilization, and user behavior for closed-loop optimization.

A sustainability metric that quantifies the total greenhouse gas emissions attributable to the operation of a specific network slice instance. The Slice-Level Energy Model provides the foundational power consumption data that, when combined with the carbon intensity of the local power grid, enables accurate calculation of this metric for regulatory reporting and sustainability targets.