Troubleshooting QoS Marking: Configuration, Compatibility & Error Resolving

Overview & Thematic Scope

This technical FAQ addresses the most common challenges encountered when configuring Quality of Service (QoS) marking—specifically DSCP and 802.1p priority tags—and mapping them to GEM (GPON Encapsulation Method) ports on B2B telecom hardware. Aimed at network engineers and support staff, this guide covers pre-sales design questions, post-sales configuration troubleshooting, compatibility pitfalls, and verification steps to ensure carrier-grade traffic prioritization.

Frequently Asked Questions

Q1: What is the exact CLI command to map a specific DSCP value to a GEM port for upstream traffic?

The primary command sequence on most major OLT platforms is ‘traffic-class’ and ‘gemport’ mapping. Specifically, use ‘configure qos dscp-mapping to traffic-class ‘ followed by ‘configure gemport set traffic-class ‘. This establishes the hardware-level priority queue mapping. Verification involves executing ‘show qos dscp-map’ and ‘show gemport detail’ to confirm the association is active and the expected packet counters are incrementing.

Q2: Why are my 802.1p tagged packets being ignored or re-marked by the ONU after mapping to a GEM port?

This is typically caused by ‘trust mode’ misconfiguration or a service profile overriding the tag. By default, many ONUs are set to ‘untrusted’, which strips or ignores the incoming 802.1p value. The definitive fix is to enable ‘trust cos’ on the ONU’s upstream interface and ensure the VLAN profile is set to ‘preserve-tag’. Additionally, verify that the GEM port’s ‘priority-queue’ setting matches the intended 802.1p priority to prevent reclassification.

Q3: What are the QoS mapping capacity limits (maximum rules/queues) per GEM port and per OLT chassis?

Maximum mapping capacity is typically hardware-constrained to 8 priority queues per GEM port, aligning with standard 802.1p COS values. Per OLT chassis, you can define up to 2048 distinct DSCP-to-queue mapping entries depending on the ASIC, though the active rule-set is usually capped at 1024. Forwarding capacity remains wire-speed (up to 10 Gbps per port) without performance degradation, as the mapping is resolved in the hardware TCAM, not the CPU.

Q4: How can I troubleshoot packet drops specifically related to QoS mapping on a GEM port?

The definitive troubleshooting methodology uses hardware counters. First, issue ‘show qos statistics interface | include drop’. Next, check the ‘WRED’ drop counters for the specific traffic-class bound to the GEM port. Common errors include incorrect ‘queue-depth’ thresholds or a mismatch between the ‘shaping-rate’ and actual traffic. If drops are seen, increase the queue-depth or adjust the shaper rate via ‘configure qos queue depth ‘ and monitor the counters in real-time.

Q5: Are there compatibility issues with third-party ONUs when using DSCP-to-GEM mapping for VoIP traffic?

Yes, proprietary QoS parsing is a frequent issue. Third-party ONUs may not honor standard DSCP values unless explicitly mapped via the OMCI (ONU Management and Control Interface). Verify the ONU’s MIB supports the ‘traffic-class’ attribute. As a standard workaround, map the DSCP to 802.1p at the ONU edge using ‘configure bridge priority dscp-to-cos’ to ensure universal compatibility. Always confirm interoperability via the hardware’s certified partner list.

Q6: What is the best practice for dynamically marking SIP signaling vs. RTP traffic to ensure they map to different GEM port queues?

Best practice dictates using a Layer-7 policy map on the upstream router to mark SIP (signaling) with DSCP 24 (AF31) and RTP (media) with DSCP 46 (EF). The OLT then maps these distinct DSCP values to separate GEM ports (or separate priority queues within a single GEM) via the ‘dscp-mapping’ rule. This physically isolates media from signaling, ensuring that RTP gets strict priority (PQ) while signaling is placed in a lower-bandwidth queue (WFQ) to prevent starvation.

Q7: How do I verify that QoS marking is working end-to-end after a firmware upgrade that reset my configurations?

After a firmware upgrade, definitively verify by generating test traffic with a BERT (Bit Error Rate Tester) or using ‘ping with ToS’ values. Check the egress counters on the GEM port using ‘show interface gem accounting’. If counters are absent, restore the ‘startup-config’ via ‘copy startup-config running-config’. The root cause is often that upgrade scripts reset the ‘qos default’ settings, requiring a reload of the mapping XML file or running ‘configure qos apply-profile’.