QoS, Traffic Management, and Monitoring
QoS questions test the pipeline: classification → marking → queuing → scheduling → shaping/policing. Know DSCP (6 bits) and 802.1p (3 bits) markings, trust boundaries, and mapping tables. Comware uses SP, WRR, WFQ, and hybrid scheduling; GTS shapes, CAR/policing drops or remarks excess. WRED provides congestion avoidance. Traps: confusing shaping (buffers) with policing (drops); assuming remarking happens before classification; and forgetting that queue scheduling only matters under congestion.
Monitoring covers SNMP (v2c/v3), syslog (severity levels 0-7), NetStream/sFlow, and NTP. Know syslog severity: 0 emergency through 7 debugging - lower number is more severe, a frequent trap. SNMPv3 adds authentication and privacy. Traps also include log buffer vs. log host configuration and time zone/NTP effects on timestamps.
Worked example: voice traffic is starved during congestion. Classify voice with an ACL or NBAR-like match, mark EF (DSCP 46), assign to a high-priority queue with SP scheduling and a bandwidth guarantee, and police excess to protect the queue. Verify with display qos queue-statistics interface and confirm the trust boundary is set at the access edge.
Rule of thumb: mark at the edge, trust only known markings, and reserve strict priority for latency-sensitive traffic with a policer to prevent starvation of other queues.