M-Lab Knowledge Base Data Access
17 articles
M-Lab's packet-headers service captures TCP packet headers for every connection on the M-Lab platform, stored as per-flow .pcap files indexed by UUID.
TCP INFO collects detailed TCP socket statistics from the Linux kernel for every connection on the M-Lab platform, capturing congestion control state, RTT, loss, and dozens of other metrics at regular intervals.
A guide for ISP engineers, network operators, and IXP staff who want to understand M-Lab data from their networks, host a measurement node, or use M-Lab for network diagnostics and performance benchmarking.
M-Lab's Traceroute core service automatically runs a network path measurement from every M-Lab server back to the client for every TCP connection, building a continuous record of Internet routing.
How to access M-Lab's raw NDT7 data in Google Cloud Storage using gsutil or gcloud without permission errors.
How to get free access to M-Lab's BigQuery datasets, run your first queries, understand the data structure, and work efficiently with large tables.
A curated starting path for academic researchers and data scientists using M-Lab data — covering data access, key datasets, measurement methodology, and common research patterns.
How M-Lab annotates measurements with geographic and network metadata, the accuracy and limitations of each annotation type, and how to use them correctly in analysis.
MSAK is M-Lab's configurable multi-stream throughput and UDP latency measurement service — designed for researchers and developers who need more control than single-stream NDT provides.
NDT is M-Lab's flagship speed test — it measures single-stream download and upload throughput, latency, and provides TCP-level diagnostics for every test.
Methods, tools, and best practices for using M-Lab's open datasets in academic research, policy analysis, and community broadband advocacy.
DASH emulates a video streaming player to measure how well a network connection supports adaptive video streaming, without relying on a real video platform.
Reverse Traceroute reconstructs the network path from an M-Lab server back to a client — the direction that standard traceroute cannot see — using a distributed system of vantage points and spoofed probes.
WeHe tests whether your ISP is treating traffic from specific apps (like YouTube or Spotify) differently from other traffic — detecting throttling or blocking on a per-application basis.
What the Monthly Stats dataset is, how it's derived from NDT measurements, what the parquet files contain, and how to access them without BigQuery.
How to interpret percentile columns in M-Lab Monthly Stats data, including the counterintuitive polarity of latency and loss, and what p50 vs p95 tells you.
A practical guide to loading, filtering, and visualising M-Lab Monthly Stats parquet files in Python using pandas, without needing BigQuery or a Google Cloud account.