NS3 Queueing Theory Simulation

Network Simulation & Queueing Analysis

---

Overview

This project uses NS-3 to simulate queueing theory concepts in a real network topology. Three clients send UDP traffic through a shared bottleneck link to a server, allowing observation of how packets queue, wait, get delayed, or get dropped under different network conditions.

The project covers three distinct queueing theory scenarios:

Scenario	Concept	Key Parameter
1 — Light Load	Stable queue (under-utilised M/M/1)	rate = 300Kbps, bottleneck = 2Mbps
2 — Heavy Congestion	Saturated queue (overloaded M/M/1)	rate = 2Mbps, bottleneck = 1Mbps
3 — Discipline Comparison	Priority vs Fair Queuing	qdisc = Prio then FqCodel

---

Project Structure

ns-3-dev/
│
├── scratch/
│   └── queueing/
│       └── mm1-queueing.cc        # Main NS3 simulation
│
├── results/                       # Auto-created by simulation
│   ├── scenario1_lightLoad_stable/
│   │   ├── queue-size.tr          # Queue depth trace
│   │   ├── flowmon.xml            # Delay, throughput, loss data
│   │   ├── queueing.xml           # NetAnim animation file
│   │   ├── plot_queue.png         # Queue graph (generated by Python)
│   │   └── delay_analysis.png     # Stats graph (generated by Python)
│   ├── scenario2_heavyLoad_congestion/
│   ├── scenario3_Prio/
│   ├── scenario3_FqCodel/
│   ├── plot_queue_comparison.png  # All scenarios overlaid
│   └── plot_delay_comparison.png  # All scenarios side by side
│
├── analysis/
│   ├── plot_queue.py              # Queue length plots
│   └── plot_delay.py              # Delay / throughput / loss plots
│
└── README.md

---

Prerequisites

1. NS-3 (already installed at ~/ns-3-dev)

Verify it works:

cd ~/ns-3-dev
./ns3 --version

2. Python dependencies

pip install matplotlib numpy lxml --break-system-packages

3. NetAnim (optional, for packet animation)

sudo apt install netanim

---

Setup

Step 1 — Copy the simulation file

cp mm1-queueing.cc ~/ns-3-dev/scratch/queueing/mm1-queueing.cc

Step 2 — Copy the analysis scripts

cp plot_queue.py ~/ns-3-dev/analysis/plot_queue.py
cp plot_delay.py ~/ns-3-dev/analysis/plot_delay.py

Step 3 — Create the results directory

mkdir -p ~/ns-3-dev/results

Step 4 — Build the simulation

cd ~/ns-3-dev
./ns3 build

Expected output (no errors):

[  0%] Building CXX object scratch/...
Finished executing the following commands:
/usr/bin/cmake --build ...

---

Running the Experiments

All commands are run from ~/ns-3-dev. Each run saves its output into its own subfolder inside results/ so runs never overwrite each other.

---

Scenario 1 — Light Load: Stable Queue

Concept: Arrival rate is well below the bottleneck capacity. Queueing theory (M/M/1) predicts the queue stays near zero, delay is minimal, and no packets are lost.

./ns3 run scratch/queueing/mm1-queueing -- \
  --rate=300Kbps \
  --bottleneck=2Mbps \
  --qdisc=Fifo \
  --simTime=30 \
  --tag=scenario1_lightLoad_stable

What each flag does:

Flag	Value	Meaning
--rate	300Kbps	Each client sends at 300 Kbps (900 Kbps total)
--bottleneck	2Mbps	Bottleneck link can carry 2 Mbps — more than enough
--qdisc	Fifo	Simple first-in-first-out queue, no prioritisation
--simTime	30	Run for 30 seconds
--tag	scenario1_lightLoad_stable	Saves results to results/scenario1_lightLoad_stable/

---

Scenario 2 — Heavy Load: Congestion and Packet Loss

Concept: Arrival rate far exceeds bottleneck capacity. Queueing theory predicts the queue fills completely, packets are dropped, delay spikes, and throughput collapses.

./ns3 run scratch/queueing/mm1-queueing -- \
  --rate=2Mbps \
  --bottleneck=1Mbps \
  --qdisc=Fifo \
  --simTime=30 \
  --tag=scenario2_heavyLoad_congestion

What each flag does:

Flag	Value	Meaning
--rate	2Mbps	Each client sends at 2 Mbps (6 Mbps total)
--bottleneck	1Mbps	Bottleneck can only carry 1 Mbps — severe overload
--qdisc	Fifo	FIFO queue — no protection, fills and drops blindly
--simTime	30	Run for 30 seconds to observe sustained congestion
--tag	scenario2_heavyLoad_congestion	Saves to results/scenario2_heavyLoad_congestion/

---

Scenario 3a — Queue Discipline: Priority Scheduling

Concept: Moderate load (arrival ≈ capacity). Shows how the PrioQueueDisc scheduler serves smaller packets first, giving lower delay to Client 0 (512B) vs Client 2 (2048B).

./ns3 run scratch/queueing/mm1-queueing -- \
  --rate=1Mbps \
  --bottleneck=1Mbps \
  --qdisc=Prio \
  --pkt0=512 \
  --pkt1=1024 \
  --pkt2=2048 \
  --simTime=30 \
  --tag=scenario3_Prio

What each flag does:

Flag	Value	Meaning
--rate	1Mbps	Each client sends at 1 Mbps (3 Mbps total)
--bottleneck	1Mbps	Bottleneck at 1 Mbps — moderate pressure
--qdisc	Prio	Priority queue: smaller/higher-priority packets go first
--pkt0/1/2	512/1024/2048	Different packet sizes per client to make priority visible
--simTime	30	30 seconds
--tag	scenario3_Prio	Saves to results/scenario3_Prio/

---

Scenario 3b — Queue Discipline: Fair Queuing with AQM

Concept: Same load as Scenario 3a but using FqCoDelQueueDisc. FqCoDel gives each flow its own queue (fairness) and actively manages delay (AQM). Compare results directly with Scenario 3a to see the discipline effect.

./ns3 run scratch/queueing/mm1-queueing -- \
  --rate=1Mbps \
  --bottleneck=1Mbps \
  --qdisc=FqCodel \
  --pkt0=512 \
  --pkt1=1024 \
  --pkt2=2048 \
  --simTime=30 \
  --tag=scenario3_FqCodel

What each flag does:

Flag	Value	Meaning
--rate	1Mbps	Same as 3a — identical load for fair comparison
--bottleneck	1Mbps	Same bottleneck
--qdisc	FqCodel	Fair Queuing + CoDel AQM — reduces bufferbloat
--pkt0/1/2	512/1024/2048	Same packet sizes as 3a
--simTime	30	30 seconds
--tag	scenario3_FqCodel	Saves to results/scenario3_FqCodel/

---

Verifying Simulation Output

After each run, check that the three output files were created:

ls -lh ~/ns-3-dev/results/scenario1_lightLoad_stable/
ls -lh ~/ns-3-dev/results/scenario2_heavyLoad_congestion/
ls -lh ~/ns-3-dev/results/scenario3_Prio/
ls -lh ~/ns-3-dev/results/scenario3_FqCodel/

Each folder should contain:

queue-size.tr      # timestamped queue depth events
flowmon.xml        # per-flow delay, throughput, loss statistics
queueing.xml       # NetAnim packet animation file

---

Generating Analysis Graphs

All analysis commands are run from the analysis/ directory:

cd ~/ns-3-dev/analysis

---

Queue Length Graphs

Single scenario — detailed two-panel chart (smooth line + raw events):

python3 plot_queue.py --tag scenario1_lightLoad_stable
python3 plot_queue.py --tag scenario2_heavyLoad_congestion
python3 plot_queue.py --tag scenario3_Prio
python3 plot_queue.py --tag scenario3_FqCodel

Each saves a PNG to its own results subfolder:

results/scenario1_lightLoad_stable/plot_queue.png
results/scenario2_heavyLoad_congestion/plot_queue.png
results/scenario3_Prio/plot_queue.png
results/scenario3_FqCodel/plot_queue.png

All scenarios overlaid — comparison chart:

python3 plot_queue.py --compare

Saves to:

results/plot_queue_comparison.png

---

Delay / Throughput / Loss Graphs

Single scenario — 4-panel chart (delay, throughput, loss, summary table):

python3 plot_delay.py --tag scenario1_lightLoad_stable
python3 plot_delay.py --tag scenario2_heavyLoad_congestion
python3 plot_delay.py --tag scenario3_Prio
python3 plot_delay.py --tag scenario3_FqCodel

Each saves to its own subfolder:

results/scenario1_lightLoad_stable/delay_analysis.png
results/scenario2_heavyLoad_congestion/delay_analysis.png
results/scenario3_Prio/delay_analysis.png
results/scenario3_FqCodel/delay_analysis.png

All scenarios side by side — grouped bar comparison:

python3 plot_delay.py --compare

Saves to:

results/plot_delay_comparison.png

---

Run Everything at Once

To run all four simulations and generate all graphs in one go:

cd ~/ns-3-dev

./ns3 run scratch/queueing/mm1-queueing -- --rate=300Kbps --bottleneck=2Mbps --qdisc=Fifo --simTime=30 --tag=scenario1_lightLoad_stable

./ns3 run scratch/queueing/mm1-queueing -- --rate=2Mbps --bottleneck=1Mbps --qdisc=Fifo --simTime=30 --tag=scenario2_heavyLoad_congestion

./ns3 run scratch/queueing/mm1-queueing -- --rate=1Mbps --bottleneck=1Mbps --qdisc=Prio --pkt0=512 --pkt1=1024 --pkt2=2048 --simTime=30 --tag=scenario3_Prio

./ns3 run scratch/queueing/mm1-queueing -- --rate=1Mbps --bottleneck=1Mbps --qdisc=FqCodel --pkt0=512 --pkt1=1024 --pkt2=2048 --simTime=30 --tag=scenario3_FqCodel

cd analysis

python3 plot_queue.py --tag scenario1_lightLoad_stable
python3 plot_queue.py --tag scenario2_heavyLoad_congestion
python3 plot_queue.py --tag scenario3_Prio
python3 plot_queue.py --tag scenario3_FqCodel
python3 plot_queue.py --compare

python3 plot_delay.py --tag scenario1_lightLoad_stable
python3 plot_delay.py --tag scenario2_heavyLoad_congestion
python3 plot_delay.py --tag scenario3_Prio
python3 plot_delay.py --tag scenario3_FqCodel
python3 plot_delay.py --compare

---

Viewing Results

Open individual PNGs:

xdg-open ~/ns-3-dev/results/scenario1_lightLoad_stable/plot_queue.png
xdg-open ~/ns-3-dev/results/plot_queue_comparison.png
xdg-open ~/ns-3-dev/results/plot_delay_comparison.png

View the packet animation in NetAnim:

netanim ~/ns-3-dev/results/scenario1_lightLoad_stable/queueing.xml

List all generated graphs:

find ~/ns-3-dev/results -name "*.png"

---

All Available Simulation Parameters

Parameter	Default	Description
--simTime	15	Total simulation duration in seconds
--rate	500Kbps	Data rate for each client OnOff application
--bottleneck	1Mbps	Bottleneck link capacity (router → server)
--accessRate	100Mbps	Client access link capacity
--accessDelay	2ms	Client access link one-way delay
--bnDelay	10ms	Bottleneck link one-way delay
--qdisc	Prio	Queue discipline: Prio, FqCodel, TBF, Fifo
--pkt0	512	Packet size for Client 0 in bytes
--pkt1	1024	Packet size for Client 1 in bytes
--pkt2	2048	Packet size for Client 2 in bytes
--tag	auto	Name for this experiment run (used as output folder name)

If --tag is not provided, an automatic tag is generated from the parameters, e.g. rate500Kbps_bn1Mbps_Prio.

---

Network Topology

Client 0 (512B)  ─┐
                   │  100Mbps/2ms     1Mbps/10ms
Client 1 (1024B) ──┤──────────────► Router ──────────────► Server
                   │   access links   [queue disc]   bottleneck
Client 2 (2048B) ─┘

• Clients send UDP traffic using an OnOff application (alternating on/off bursts)
• The router applies the selected queue discipline on the bottleneck link
• All three flows share the single 1Mbps (default) bottleneck — this is where queuing occurs
• The FlowMonitor measures per-flow statistics end-to-end

---

Troubleshooting

Clock skew warning during build:

gmake[3]: warning: Clock skew detected.

This is harmless. Fix it permanently with:

sudo hwclock --hctosys

FigureCanvasAgg is non-interactive warning from Python: This is expected in headless/SSH terminals. The scripts use the Agg backend which writes directly to PNG — no display server needed. The PNG files are saved correctly.

lxml not found:

pip install lxml --break-system-packages

No experiment subfolders found: Make sure you ran the NS3 simulation first and that the results/ folder is inside ~/ns-3-dev/, not somewhere else.

Build fails with MobilityHelper not declared: Make sure your mm1-queueing.cc has this include at the top:

#include "ns3/mobility-module.h"

---

What the Output Files Mean

File	Generated by	Contains
queue-size.tr	NS3 trace	Timestamped queue depth in packets — one row per change event
flowmon.xml	NS3 FlowMonitor	Per-flow: packets sent/received/lost, total bytes, delay sum, timestamps
queueing.xml	NS3 NetAnim	Full packet animation — open in the NetAnim GUI
plot_queue.png	plot_queue.py	Queue length over time — smooth line + raw event scatter
delay_analysis.png	plot_delay.py	Avg delay, throughput, packet loss per flow + summary table
plot_queue_comparison.png	plot_queue.py --compare	All scenarios overlaid on one chart
plot_delay_comparison.png	plot_delay.py --compare	All scenarios side by side grouped bars