Slice-Aware Scheduling

The scheduler maintains separate logical queues for each active network slice instance. Packets are not scheduled solely on channel quality; the scheduler first applies a slice-level weight derived from the slice's SLA parameters. A URLLC slice queue receives preemptive priority over an eMBB slice queue, ensuring latency budgets are met even under heavy load. This is implemented through a hierarchical scheduling structure where inter-slice resource partitioning occurs before intra-slice user selection.

Physical resource blocks are partitioned into slice-specific resource pools to enforce hard isolation. The scheduler uses a token bucket or proportional fair share mechanism at the slice level to prevent a single greedy slice from starving others. Key enforcement mechanisms include:

• Minimum PRB reservation: A guaranteed floor of resources for GBR slices
• Maximum PRB cap: A ceiling to limit resource consumption by non-GBR slices
• Dynamic borrowing: Idle resources from one slice pool can be temporarily lent to another, with preemption rights retained

Within each slice, the scheduler maps 5QI (5G QoS Identifier) values to specific scheduling policies. Each QoS flow carries a resource type (GBR, Delay-Critical GBR, or Non-GBR) that dictates the scheduling discipline:

• Delay-Critical GBR: Earliest deadline first with short transmission time intervals
• GBR: Proportional fair with guaranteed bit rate floor
• Non-GBR: Best-effort with weighted fair queuing This granular mapping ensures that the slice's internal traffic mix is handled according to its composite SLA definition.

Slice-aware scheduling does not ignore channel conditions; it integrates them as a secondary optimization dimension. After slice-level resource allocation, the scheduler selects the optimal user within each slice based on frequency-selective scheduling metrics. For a URLLC slice, this may prioritize the user with the most reliable channel to minimize retransmissions. For an eMBB slice, it may select the user experiencing peak spectral efficiency to maximize aggregate throughput, exploiting multi-user diversity within the slice's allocated PRB pool.

An advanced variant of slice-aware scheduling that consolidates active transmissions into the minimum number of PRBs and OFDM symbols. By packing slice traffic tightly in the time-frequency grid, the scheduler creates extended idle periods where power amplifiers and RF chains can enter micro-sleep states. This is particularly effective for mMTC slices with sporadic, small-payload traffic. The scheduler may deliberately delay non-latency-sensitive slice transmissions by a few subframes to align them into a burst transmission window, maximizing the duration of subsequent sleep intervals.

The scheduler dynamically selects the Modulation and Coding Scheme (MCS) per user based on both channel quality and slice reliability requirements. For a URLLC slice, the scheduler applies a conservative MCS backoff—selecting a lower-order modulation than channel conditions would permit—to achieve a block error rate target of 10^-5 or lower. For an eMBB slice, it targets a 10% BLER to maximize spectral efficiency. This slice-differentiated outer loop link adaptation ensures that the physical layer transmission parameters align with the logical slice's error tolerance.

The foundational identifier that makes slice-aware scheduling possible. NSSAI is a collection of Single NSSAI (S-NSSAI) parameters sent by user equipment during registration. Each S-NSSAI contains a Slice/Service Type (SST) —a standardized value indicating expected behavior like eMBB, URLLC, or MIoT—and an optional Slice Differentiator (SD) for tenant-specific instances. The MAC scheduler reads the NSSAI associated with each user's data radio bearer to map traffic to the correct scheduling policy, ensuring a URLLC slice gets immediate grant priority while an eMBB slice receives best-effort treatment.

A data-driven analytical model that quantifies the power consumption of a specific network slice instance as a function of its allocated Physical Resource Blocks (PRBs) , traffic load, and configured SLA parameters. The slice-aware scheduler uses this model to predict the energy impact of its allocation decisions in real time. For example, the model might reveal that consolidating non-URLLC traffic onto fewer PRBs and activating Cell Discontinuous Transmission (Cell DTX) on idle carriers saves 15% power without violating latency budgets. This transforms the scheduler from a pure performance optimizer into an energy-aware decision engine.

The fundamental scheduling dichotomy. A GBR slice is configured with dedicated network resources and a fixed bandwidth commitment, suitable for constant-throughput services like real-time video or industrial automation. The scheduler must pre-allocate PRBs to meet the GBR contract before serving other traffic. A Non-GBR slice receives no minimum bandwidth guarantee and is served on a best-effort basis. Slice-aware scheduling logic must enforce strict priority: first satisfy all GBR slice commitments, then distribute remaining PRBs among Non-GBR slices using proportional fairness or weighted round-robin algorithms.

An automation framework where a policy-driven controller continuously monitors slice KPIs, analyzes deviations from the desired state using AI, and automatically executes corrective reconfiguration actions without human intervention. The Network Data Analytics Function (NWDAF) feeds predictive load and UE mobility analytics to the controller, which then adjusts the slice-aware scheduler's weights and resource partitioning ratios. If a URLLC slice's latency drifts above its SLA threshold, the closed-loop system can instantly increase its scheduling priority and preempt resources from lower-priority slices, restoring compliance autonomously.

A centralized control strategy that synchronizes the activation of low-power states across multiple network components—carriers, MIMO paths, and baseband processing units—within a slice to maximize energy savings without violating service guarantees. The slice-aware scheduler plays a critical role by deliberately packing delay-tolerant traffic into specific time slots and carriers, creating extended idle periods where components can enter deep sleep. Coordination with Wake-Up Signal (WUS) mechanisms ensures UEs in sleep states are alerted only when their slice has pending data, avoiding unnecessary wake-ups that waste device battery.

The process of dynamically reassigning an active user equipment session from one network slice instance to another to optimize for changing service requirements, load conditions, or energy efficiency targets. The slice-aware scheduler triggers remapping when it detects a mismatch between a UE's actual traffic pattern and its assigned slice's profile—for example, moving a UE that was assigned to a high-power URLLC slice but is only sending background data to a lower-energy eMBB slice. This requires tight integration with the Slice Orchestrator to ensure session continuity and state transfer during the transition.