URLLC Delay in a 5G MPLS Backhaul
A 5G non-standalone transport network uses 7750 SR pre-aggregation routers and an MPLS core. A gNB connects through a 1 Gb/s backhaul SAP. The mobile operator defines three traffic classes: PTP/sync highest, URLLC user plane DSCP EF at 10 Mb/s, and eMBB DSCP AF41 up to 700 Mb/s. The gNB marks DSCP correctly, but the transport network QoS policy is the factory default; all packets are mapped to EXP 0 and a single network queue. During busy hour, URLLC one-way delay rises from 2 ms to 18 ms and PTP packet delay variation exceeds 5 microseconds. The RAN team says the radio is healthy. The transport team has 40% average utilization on the core links. There is no admission control on the backhaul SAP, and eMBB bursts fill the queue.
1. Explain why DSCP marking at the gNB is not enough to protect URLLC across the MPLS core.
Read the worked answer
DSCP is carried in the IP header; in an MPLS core the per-hop behavior is driven by EXP or traffic-class bits and local forwarding class. With the factory network QoS policy, every packet gets EXP 0 and one network queue, so the core cannot distinguish URLLC from eMBB even though the gNB marked DSCP. The 40% average utilization does not preclude microburst congestion; a 700 Mb/s eMBB burst can fill the queue and add delay. PTP is especially sensitive: it needs separate high-priority treatment and consistent per-hop behavior. The fault is not the radio; it is missing mapping and queuing at the transport boundary.
2. Design the per-node QoS mapping and queuing you would apply.
Read the worked answer
At the PE network ingress, classify: DSCP EF to FC EF with profile in and EXP 5; AF41 to AF with EXP 3; PTP to NC or a dedicated high-priority FC and queue with EXP 5 or 6 and the highest scheduler priority. Configure network queues with strict priority for sync and URLLC and weighted or shaped service for eMBB; set CIR, PIR, CBS, MBS, and WRED so eMBB cannot starve latency traffic. Apply the same network QoS policy on every router in the path. Add admission control or shaping on the backhaul SAP to cap eMBB below the 1 Gb/s SAP rate. Verify with per-queue counters and delay measurements.
3. What monitoring would prove the transport is meeting the URLLC and sync objectives?
Read the worked answer
Monitor per-EXP and per-FC counters, queue drops, and queue delay, not just link utilization; use streaming telemetry at sub-minute intervals to catch microbursts. Check PTP PDV at the far end, one-way delay for URLLC, and whether any node re-marks EXP. A mirror or capture at ingress and egress can confirm EXP values. Also verify scheduler configuration consistency and that the SAP shaper is not dropping bursts. Compare busy-hour and off-peak baselines so a transient improvement is not mistaken for a fix.