Troubleshooting performance issues
Collisions
Two hosts transmit at once and the signals merge into garbage. Prevent them with smaller collision domains: every switch port is its own domain.
Deferred counters show carrier-sense retries: normal on hubs, a red flag on switched networks.
Late collisions (after 5.12 microseconds) point to wrong cables, bad
NICs, or too many hubs. Excess collisions (the retransmit limit hit)
point to full duplex on a shared segment, a broken NIC, or an
overcrowded domain; show controller ethernet counts them. The
fixes: disable autonegotiation, hard-code a lower speed, or drop to
half duplex on the shared segment.
Broadcast storms
Continuous broadcast and multicast traffic drowning the network (L2 FF:FF:FF:FF:FF:FF, L3 255.255.255.255). The causes: oversized broadcast domains, DHCP request floods, switching loops.
Prevent them by breaking up domains with Layer 3 devices, limiting MACs per port, and STP/BPDU loop protection. Spot them by packet counters spiking past baseline; confirm with Wireshark or tcpdump.
VoIP issues
Latency: keep it under 50-100 ms for calls. Past 100-200 ms you get echoes and talk-over (satellite adds 150-250 ms by physics alone).
Jitter is delay variation: robotic, static-y audio. Trouble starts when latency swings by 30-50 ms.
QoS prioritizes voice inside your network, but your ISP won’t honor it across the internet.
Packet loss
The symptoms: unexplained slowdowns, jitter on calls, dropped streams. The causes: congestion, faulty router configs, bad cables, hardware failures.
Troubleshoot with ping (reachability) and traceroute (path), then monitoring tools for patterns. Mitigate with more bandwidth, better layout, QoS, cable replacement, firmware updates, config verification, and continuous monitoring.
General performance issues
High CPU on network devices (upgrade or simplify load), high bandwidth usage (add capacity, NetFlow analysis to find the hogs), poor physical connectivity (test from the demarc inward: cable tester for copper, light meter for fiber), device misconfigurations, and DNS latency.
Other culprits
- Low optical link budget: expected fiber losses (distance, bends, splices, patches) exceed the power budget. Measure with an OTDR; about 0.25 dB/km is normal. Total link budget = power budget minus losses.
- Certificate issues: untrusted signer, expired, or missing. Buy from a trusted CA, renew, install properly.
- License feature errors: wrong license tier for the features you need.
- BYOD: device sprawl to support and segregation to enforce.
- Hardware failures: isolate the component, replace it, keep spares and backups.
Exam quick facts
N10-009 domains by weight: Network Troubleshooting (24%), Networking Concepts (23%), Network Implementation (20%), Network Operations (19%), Network Security (14%).
No guessing penalty. Brain-dump a cheat sheet at the start, and save simulations for after the multiple choice.