RTT Under Load — Network Parameter Analysis

Computer Networks (BCSE308L) — Digital Assignment 3 VIT Chennai | Slot: F1+TF1 | Dr. SubbuLakshmi T

Naman Bhukar · 24BCE5047 · B.Tech CSE

---

Overview

This repository contains the full artifacts for my DA3 on network parameter analysis. The parameter studied is Round Trip Time (RTT), measured across 30 iperf3 tests run against iperf.he.net (Hurricane Electric's public iperf3 endpoint) under three traffic classes:

• Low (L1–L10) — 1 stream, light load
• Medium (M1–M10) — 2–3 parallel streams, rate caps, longer durations
• High (H1–H10) — 5–10 parallel streams or UDP floods up to 200 Mbps

Packets were captured with Wireshark and RTT samples were extracted from TCP timestamps using a Python script (Scapy + matplotlib).

---

Repository Structure

. ├── rtt.py # Main analysis script ├── pcaps/ # All 30 packet captures (L1–L10, M1–M10, H1–H10) ├── rtt_graphs/ │ ├── per_file/ # Per-test RTT trend graphs (PNG) │ └── summary/ # Summary bar chart, boxplot, and CSV └── README.md

---

How It Works

1. Capture — Each iperf3 run was captured in Wireshark and saved as a .pcapng file, named by traffic class (e.g. M5.pcapng, H10.pcapng).

2. Parse — rtt.py reads every pcap in the folder. For each outgoing TCP data segment it records (sequence_end, timestamp). When the matching ACK arrives from the server, RTT is computed as the time difference. If a capture has very little data traffic, the script falls back to SYN → SYN-ACK handshake timing.

3. Plot — Per-file RTT trend graphs + summary charts (average RTT bar chart, traffic-class boxplot) + a CSV of all metrics.

---

Running the Script

Requirements:

pip install scapy matplotlib

Run:

# Place rtt.py in the same folder as your .pcapng files
python rtt.py

Output goes to rtt_graphs/per_file/ (30 trend plots) and rtt_graphs/summary/ (bar chart, boxplot, CSV).

---

iperf3 Commands Used

Examples from each traffic class:

# Low traffic
iperf3 -c iperf.he.net -t 10
iperf3 -c iperf.he.net -t 10 -b 5M
iperf3 -c iperf.he.net -t 10 -l 256

# Medium traffic
iperf3 -c iperf.he.net -P 2 -t 30
iperf3 -c iperf.he.net -P 2 -t 30 -b 10M
iperf3 -c iperf.he.net -P 2 -t 45 -b 15M     # worst average RTT: 6886 ms

# High traffic
iperf3 -c iperf.he.net -P 10 -t 30
iperf3 -c iperf.he.net -u -b 100M -t 30
iperf3 -c iperf.he.net -P 10 -t 60 -b 10M -l 8K

All 30 commands are documented in the blog.

---

Key Findings

Traffic class	Typical avg RTT	Worst run
Low	258–423 ms	L9: 423 ms
Medium	346–6886 ms	M9: 6886 ms
High	1143–5416 ms	H2: 5416 ms

• Baseline RTT: ~250 ms (physical floor of the link)
• Bufferbloat is real and visible — M5, M6, M8, M9 all show the slow-ramp signature
• M9 (medium traffic) beat every high-traffic test on average RTT — traffic shape matters more than volume
• More streams can produce lower RTT than fewer (H4: 10 streams → 1143 ms; H1: 5 streams → 4617 ms)
• UDP without rate control produces distinctive triangular RTT curves

---

Links

• 📝 Blog: https://rrt-analysis.blogspot.com/2026/04/rtt-under-load-beginners-tour-through.html
• 🎥 Video: https://youtu.be/kE-ID_D7Yg0

---

Tools

• iperf3 — traffic generation
• Wireshark — packet capture
• Python 3 — Scapy (pcap parsing), matplotlib (plotting)

---

Acknowledgements

• Dr. SubbuLakshmi T — course instructor, BCSE308L
• VIT Chennai and SCOPE
• The Wireshark, Scapy, and iperf3 open-source communities

---

© 2026 Naman Bhukar · VIT Chennai