Performance Issues N10-009

Slide 1 of 10  |  N10-009 Obj 5.4  |  Troubleshooting
Network Performance Issues:
Diagnosing the Congested Network
Three symptoms. One root cause. Your job is to find it.
Users complain video calls are choppy, file transfers are slow, and the VoIP phones sound robotic. Three different symptoms, one root cause: the network is congested. Bandwidth, throughput, latency, jitter, and packet loss are the five metrics that explain every performance complaint — and every N10-009 scenario question.
10 Slides N10-009 Obj 5.4 Network Performance Troubleshooting + Tools
Slide 2 of 10
Bandwidth vs Throughput
Bandwidth is the pipe's capacity. Throughput is how much actually flows. They are never the same.
BANDWIDTH = PIPE DIAMETER (Potential Capacity) 1 Gbps Bandwidth THROUGHPUT = ACTUAL FLOW (Limited by Restrictions) Congestion Queue drop ~350 Mbps actual 35% of bandwidth
Bandwidth
The theoretical maximum data rate of the link. Determined by the physical medium: 1 Gbps copper, 10 Gbps fiber, 100 Mbps cable. It is the pipe diameter — constant, unchanging, purchased from the carrier or determined by hardware. You cannot exceed it.
Throughput
The actual measured data rate through the pipe, accounting for all restrictions: congestion, packet loss, retransmissions, protocol overhead, TCP flow control. Always less than bandwidth. The delta between bandwidth and throughput is where performance problems live.
Goodput
A subset of throughput — the useful application data rate after removing protocol overhead, retransmissions, and duplicate ACKs. This is what the application actually receives. Choppy video calls = low goodput even if raw throughput is adequate.
Slide 3 of 10
Latency: The Speed of Your Connection
Bandwidth tells you how much. Latency tells you how fast. A high-bandwidth, high-latency link is still slow for interactive use.
LAN <1 ms A B RTT <1ms Same building WAN 50ms A B RTT ~50ms Cross-country SAT 600ms A B SAT RTT ~600ms GEO orbit
Sources of Latency
Propagation delay: speed of light in the medium (fundamental, irreducible).
Processing delay: time for a router/switch to process the packet header.
Queuing delay: time spent waiting in an interface queue — this is where congestion causes latency spikes.
Serialization delay: time to place all bits on the wire (significant for slow links with large packets).
Why Latency Kills VoIP and Video
G.711 VoIP: acceptable one-way latency is <150 ms (ITU G.114). Above 400 ms, conversations become unusable — people talk over each other. Video conferencing: above 150 ms RTT, the meeting experience degrades noticeably. High-bandwidth satellite links are useless for real-time comms because of 600 ms propagation delay alone.
Slide 4 of 10
Jitter: When Packets Arrive at the Wrong Time
Latency variation. Packets sent at even intervals, arriving at uneven intervals — that is jitter. VoIP phones sound robotic because of it.
SENDER (evenly spaced 20ms apart) 1 2 3 4 5 20ms 20ms 20ms 20ms RECEIVER — HIGH JITTER (erratic arrival) 1 2 3 4 5 5ms gap 92ms gap 15ms gap 115ms gap JITTER BUFFER — Smoothing Solution Erratic in: JITTER BUFFER 20-150ms depth Even out: Trade-off: buffer adds latency Adaptive buffers adjust depth dynamically
What Causes Jitter
Queue depth variations on congested links. Different routing paths per packet (asymmetric routing). CPU spikes on forwarding devices. Wireless channel contention and retransmissions. Any variable that changes the per-packet forwarding time introduces jitter.
VoIP Impact
Acceptable jitter for VoIP: <30 ms (ITU recommendation). Above 50 ms, speech becomes choppy and robotic. The phone's jitter buffer smooths arrivals but adds latency — a trade-off. When the jitter exceeds the buffer depth, packets play out of order or are dropped.
Measuring Jitter
Wireshark: RTP stream analysis shows per-packet jitter. SolarWinds VoIP Monitor: real-time jitter alerts. iPerf3: iperf3 -u (UDP mode) reports jitter and loss. Continuous testing during business hours reveals congestion correlation.
Slide 5 of 10
Packet Loss: When Data Disappears
TCP retransmits lost packets — causing slowdown. UDP silently drops them — causing degradation. The effect depends on the protocol.
Source 10.0.0.1 DROP Dest 10.0.0.2 Congested link TCP: NACK + Retransmit (adds latency, slows throughput) UDP: No retransmit — packet gone. VoIP/video degrades silently.
Causes of Packet Loss
Interface queue overflow (congestion). Physical errors — bad cable, damaged SFP, signal degradation. Memory exhaustion on switch or router. Wireless retransmission failure. Tail drop (packets dropped at tail of full queue). RED (Random Early Detection) — intentional pre-emptive drops to signal TCP to slow down.
TCP vs UDP Behavior
TCP detects loss via ACK timeout or NACK, reduces congestion window (CWND), retransmits. Throughput drops but data arrives. Noticeable as slowness in file transfers.

UDP has no recovery mechanism. Lost packet = lost data. VoIP call drops a 20ms frame. Video call shows artifacts. File download over UDP would silently corrupt data.
Acceptable Loss Thresholds
Data transfer: 0% ideally. Even 0.1% causes TCP slowdown.
VoIP: <1% acceptable per ITU. Above 5%, intelligibility degrades significantly.
Video conference: <0.5% for HD quality.
Measure with ping -n 100 or pathping.
Slide 6 of 10
Congestion & Bottlenecks
Multiple wide flows converging on one narrow link. The queue builds. Then it overflows. Then packets die.
1 Gbps link 1 Gbps link 1 Gbps link 100 Mbps bottleneck 3 Gbps in, 100 Mbps out Queue building Packets dropped Congested output Dest Sees loss+jitter FIXES: Upgrade link Add QoS Traffic shaping
Identifying the Bottleneck
Use pathping or traceroute to find which hop introduces latency. SNMP monitoring: interface utilization. The hop where utilization approaches 100% and latency spikes is the bottleneck. Check both directions — congestion is often asymmetric.
Solutions
Link upgrade: permanent fix. 100 Mbps to 1 Gbps eliminates the current bottleneck.
QoS: prioritize VoIP and video over bulk data. Does not add bandwidth but ensures critical traffic gets through first.
Traffic shaping / policing: rate-limit bulk transfers (backups, updates) during business hours.
Slide 7 of 10
Wireless Performance Issues
Wireless performance problems are often invisible until users complain. Signal, interference, and roaming misconfiguration are the usual suspects.
CHANNEL INTERFERENCE AP1 Ch 6 AP2 Ch 6 ! CCI Zone Non-overlapping 2.4 GHz: Ch 1, 6, 11 5 GHz: 24+ non-overlapping 6 GHz (Wi-Fi 6E): 59 channels
Co-channel Interference (CCI)
Two APs on the same channel in overlapping coverage areas. Devices must wait for the channel to be clear. Throughput halved or worse. Fix: channel planning — assign non-overlapping channels to adjacent APs.
Adjacent Channel Interference
APs on channels 1 and 3 partially overlap in frequency. Signal bleeds between channels. Worse than CCI in some cases. Solution: use only channels 1, 6, 11 on 2.4 GHz. Use 5 GHz for dense environments.
Signal Degradation
RSSI below -70 dBm: connectivity degrades. Below -80 dBm: connection drops. Physical obstacles: concrete walls attenuate 2.4 GHz by 10-15 dB. Microwave ovens, Bluetooth, baby monitors all cause 2.4 GHz interference.
Client Disassociation
Client refuses to roam from weak AP to strong AP ("sticky client"). Common when minimum RSSI threshold not configured on AP. Result: client stays connected at low data rates when a better AP is available nearby.
Roaming Misconfiguration
802.11r (Fast BSS Transition): reduces roaming handoff time from 200-400 ms to <50 ms. 802.11k: neighbor AP reports so clients know where to roam. 802.11v: AP can steer clients. All three together = enterprise-grade seamless roaming.
Slide 8 of 10
QoS: Priority Queuing & DSCP
Quality of Service does not add bandwidth. It ensures that when bandwidth is scarce, the right traffic wins.
VoIP (EF DSCP 46) Video (AF41 DSCP 34) Data (CS0 DSCP 0) Bulk/Backup (CS1) PRIORITY QUEUES Q1: Voice (EF) Q2: Video (AF4) Q3: Data (BE) Q4: Bulk (CS1) Scheduler: strict priority for Q1, WFQ for Q2-Q4 DSCP QUICK REFERENCE EF (46) — Expedited Fwd — VoIP AF41(34) — Assured Fwd — Video AF21(18) — Assured Fwd — Data CS3 (24) — Class Select — Signal BE (0) — Best Effort — Default CS1 (8) — Low Priority — Bulk DSCP value in IP header ToS field Routers and switches read and honor
How QoS Fixes the Scenario
Without QoS: a 500 MB file upload during a VoIP call competes equally for bandwidth. The large TCP transfer causes queue depth to grow, adding latency and jitter to voice packets. With QoS: VoIP packets are always served first from the priority queue, regardless of what else is in the queue.
DSCP Marking vs Trusting
Edge devices (phones, cameras) mark their own DSCP values. QoS policy on switches and routers either trust those marks (internal endpoints) or reclassify/police them (external or untrusted). Phones typically mark EF. Rogue or misconfigured devices can abuse markings — police at the network edge.
Slide 9 of 10
Diagnostic Commands
The right tool for each metric. Know the command, know what it measures, know how to read the output.
Latency — ping
C:\> ping -n 100 8.8.8.8
Reply: bytes=32 time=8ms TTL=117
Reply: bytes=32 time=9ms TTL=117
Packets: Sent=100, Received=98, Lost=2 (2%)
Min=7ms, Max=142ms, Avg=9ms
High Max vs Avg = latency spikes under load
Hop-by-hop Loss — pathping
C:\> pathping 10.5.0.1
Hop 1: 192.168.1.1    0% loss   2ms
Hop 2: 10.1.0.1      12% loss  5ms
Hop 3: 10.2.0.1      0% loss   6ms
Loss at hop 2 = problem on that link or device
Throughput — iperf3
$ iperf3 -c 10.5.0.100 -t 30
[ 5] 0.00-30.00 sec  3.58 GBytes
     Bandwidth: 1.00 Gbits/sec
$ iperf3 -c 10.5.0.100 -u -b 1G
Jitter: 4.231 ms   Lost: 2.1%
Interface Statistics — show interface
SW1# show interface Gi0/1
GigabitEthernet0/1 is up
Input queue: 0/75/850/0 (drops)
Output drops: 1240
5 min input rate: 998,432,000 bits/sec
Output drops + 99.8% utilization = congestion
Wireless Signal — netsh (Windows)
C:\> netsh wlan show interfaces
SSID         : CorpWiFi
Signal       : 42%
Receive rate : 24 Mbps
Transmit rate: 24 Mbps
Low signal + low rate = sticky client or weak AP
Wireshark for Jitter Analysis
Capture RTP stream. Statistics > VoIP Calls > Stream Analysis. Shows per-packet jitter graph. Any spike above 30 ms correlates to robotic audio. Filter: rtp && ip.src == 10.1.0.50
Slide 10 of 10  |  Summary
Performance Issues: Scenario Resolved
1Bandwidth = pipe capacity (potential). Throughput = actual measured flow. Goodput = useful application data after overhead. Never the same number.
2Latency = propagation + processing + queuing + serialization delay. VoIP requires <150 ms one-way. Satellite latency (600 ms) is fundamentally incompatible with real-time voice.
3Jitter = latency variation between consecutive packets. Causes robotic VoIP audio. Fixed by jitter buffer (adds latency, removes variation). Keep jitter <30 ms for voice.
4Packet loss: TCP retransmits (slowdown), UDP discards (degradation). VoIP: <1% acceptable. Use pathping to identify which hop drops packets.
5Congestion: multiple flows converging on a narrow link. Queue builds, packets drop, latency spikes. Fixes: link upgrade, QoS, traffic shaping.
6Wireless: CCI (same channel), ACI (adjacent channel), low RSSI, sticky clients, missing 802.11r/k/v for roaming. Use 5 GHz / 6 GHz for dense deployments.
7QoS prioritizes traffic via DSCP: EF(46) = VoIP, AF41(34) = Video, CS0 = best effort. Strict priority queue ensures voice packets skip the line.
Scenario resolved. pathping showed 18% packet loss at the WAN edge router. show interface on that router revealed 99.3% utilization and 4,200 output drops. The cause: a backup job scheduled to run at 9am was saturating the 100 Mbps WAN link.

Fix applied: QoS policy marking backup traffic CS1, VoIP EF. Backup job rescheduled to midnight. WAN link upgraded from 100 Mbps to 500 Mbps. Video calls, file transfers, and VoIP now operate without interference.
N10-009 Obj 5.4 Coverage
Bandwidth, throughput, goodput, latency (propagation/queuing/processing), jitter, jitter buffer, packet loss (TCP vs UDP), congestion, bottleneck, wireless interference (CCI/ACI), signal degradation, sticky clients, roaming (802.11r/k/v), QoS, DSCP, EF/AF/CS markings, diagnostic tools (ping, pathping, iperf3, Wireshark, show interface).