From_GEO_to_LEO_IMC2025

This repository contains the dataset and analysis code used for the paper "From GEO to LEO: First Look Into Starlink In-Flight Connectivity" from the ACM IMC 2025 conference.

Overview

This dataset provides comprehensive network performance measurements comparing traditional Geostationary Orbit (GEO) satellite networks with Low Earth Orbit (LEO) satellite networks, specifically Starlink, in the context of in-flight connectivity. The dataset includes comparative metrics between 5 different GEO Satellite Network Providers (Intelsat, Inmarsat, Panasonic, SITA, Viasat) and Starlink (LEO SNO). All measurements were conducted using Amigo testbed.

Repository Structure

├── data/                           # Raw measurement data
│   ├── GEO_v_LEO/                 # Comparative GEO vs LEO measurements
│   │   ├── cdn/                   # CDN performance measurements
│   │   ├── dns/                   # DNS resolver 
│   │   ├── mtr/                   # Network path analysis (myTraceroute)
│   │   └── speedtest/             # Ookla speedtest results
│   └── Starlink_Extension/        # Detailed Starlink-specific diagnostics
│       ├── flight_positions/      # Fine-grained flight position data
│       ├── irtt/                  # High-frequency UDP ping measurements
│       └── tcp_file_transfer/     # TCP file transfer performance
└── code/                         
    └── *.py                       # Python scripts for figures

Data Description

1. GEO_v_LEO Dataset

This folder contains comparative measurements between GEO satellite providers and Starlink (LEO).

Network Path Analysis (data/GEO_v_LEO/mtr/)

• My Traceroute (MTR) results to DNS and service providers
• Targets:
  • DNS providers: Cloudflare DNS, Google DNS
  • Service providers: google.com, facebook.com
• File Naming Convention: <client_asn>_<timestamp>_<target>_<ip_version>_<PoP>.txt.gz
• mtr_ip_lookup.csv: IP address details (geolocation, ASN, hostname, etc) from ipinfo

Speed Test Results (data/GEO_v_LEO/speedtest/)

• Content: Ookla speedtest measurements
• File Naming Convention: <client_asn>_<timestamp>.json

CDN Performance (data/GEO_v_LEO/cdn/)

• Content: Results from downloading jQuery library (v3.6.0) from 5 CDN providers
• CDN Providers: Cloudflare, Google, jQuery, jsDelivr, Microsoft Ajax
• Data Types:
  • Download statistics (from curl)
  • HTTP headers collected during download
• File Naming Convention:

  • stats_<client_asn>_<timestamp>_<cdn_provider>_<PoP>.gz

  • headers_<client_asn>_<timestamp>_<cdn_provider>_<PoP>.gz

  • PoP: Point of Presence (Starlink AS14593 clients) or "GEO" for GEO flights

DNS resolvers (data/GEO_v_LEO/dns/)

• Content: DNS resolver in use, obtained via NextDNS
• File Naming Convention: <client_asn>_<timestamp>_<PoP>.txt

2. Starlink Extension Dataset

This folder contains detailed diagnostics collected during Starlink flights.

Flight Positions (data/Starlink_Extension/flight_positions/)

• Content: Fine-grained flight position data across 6 Starlink flights
• Routes:
  • DOH-JFK (Doha to New York JFK) - 2 flights
  • DOH-LHR (Doha to London Heathrow) - 1 flight
  • JFK-DOH (New York JFK to Doha) - 2 flights
  • LHR-DOH (London Heathrow to Doha) - 1 flight

High-Frequency UDP Ping (data/Starlink_Extension/irtt/)

• Content: High-frequency UDP ping results using IRTT
• Coverage: 2 Starlink flights (DOH-LHR, LHR-DOH)
• File Naming Convention: <PoP>_<aws_server>_<timestamp>
  • aws_server: AWS cloud region hosting the measurement endpoint
• AWS Regions: me-central-1 (UAE), eu-central-1 (Frankfurt), eu-west-2 (London), eu-south-1 (Milan)

TCP File Transfer Performance (data/Starlink_Extension/tcp_file_transfer/)

• Content: TCP file transfer test results with socket-level statistics
• Coverage: 2 Starlink flights (DOH-LHR, LHR-DOH)
• Data Types:
  • Socket-level statistics (from ss command): <timestamp>_<PoP>_<aws_server>_<TCP_CCA>_client-socket-stats.log
  • Retransmission data (per 100ms): <timestamp>_<PoP>_<aws_server>_<TCP_CCA>_retransmission.csv
• TCP Congestion Control Algorithms: BBR v1, Cubic, Vegas

Analysis Code

The code/ folder contains Python scripts used to generate figures in the research paper:

Script	Figure
mtr_analysis.py	Figures 4, 5
speedtest_analysis.py	Figure 6
cdn_analysis.py	Figure 7
irtt_analysis.py	Figure 8
delivery_rate_analysis.py	Figure 9
retransmission_analysis.py	Figure 10

Citation

If you use this dataset in your research, please cite our paper:

Contact

For questions about this dataset or the research, please contact daniel.jang@u.northwestern.edu.