M-Lab Knowledge Base MSAK (Measurement Swiss-Army Knife) is a measurement service hosted by M-Lab that implements two measurement protocols: a configurable WebSocket-based throughput protocol and a UDP-based latency protocol. Where NDT is a standardized single-stream bulk transport test, MSAK is designed for cases where researchers need to tune measurement parameters.
MSAK is also the engine behind M-Lab’s official speed test at speed.measurementlab.net.
How MSAK Works
Throughput Protocol
MSAK’s throughput protocol is WebSocket-based and supports multi-stream tests — running several parallel TCP connections simultaneously. This allows it to measure the aggregate throughput that a connection can sustain across multiple streams, which is closer to how many real applications (browsers, download managers, video platforms) behave.
Configurable parameters include:
- Number of streams — how many parallel TCP connections to open
- Congestion control algorithm — e.g., BBR or Cubic
- Test duration — how long to run the measurement
- Per-stream byte limit — cap the data transferred per stream
For most users, the defaults are appropriate. The configurability is intended for researchers designing specific measurement studies.
UDP Latency Protocol
MSAK also implements a UDP-based latency measurement protocol, distinct from the RTT measurements derived from TCP in NDT. UDP latency can reveal queuing and delay characteristics that TCP-based measurements may mask (since TCP’s flow control adapts to congestion).
What MSAK Measures
Each MSAK throughput test records:
- Per-stream throughput over time, including start/end timestamps per stream
- Aggregate throughput across all streams
- TCP socket statistics (via tcp-info sidecar)
- Packet captures (via pcap sidecar)
- Client IP address and server site
UDP latency tests record round-trip times and packet loss for a series of UDP probes.
Privacy and Data Collection
When you run MSAK, your IP address is collected along with measurement results and published publicly. See M-Lab’s Privacy Policy.
Accessing MSAK Data
BigQuery
MSAK data is available in BigQuery for free. See Setting Up Free BigQuery Access.
Data is in two datasets:
| Dataset | Contents |
|---|---|
measurement-lab.msak | Processed/annotated MSAK measurements |
measurement-lab.msak_raw | Raw parsed measurements |
Raw Data in Google Cloud Storage
Raw MSAK data is available in GCS:
- Throughput and latency results: gs://archive-measurement-lab/autoload/v1/msak
- Sidecar data (tcpinfo snapshots and PCAPs): gs://archive-measurement-lab/msak
See Accessing Data in GCS for how to download and work with raw archives.
Running an MSAK Test
- Browser: Visit speed.measurementlab.net — it uses MSAK under the hood.
- Command line: A standalone Go client is available in the MSAK GitHub repository.
How People Use MSAK Data
Multi-stream throughput research — comparing multi-stream throughput against single-stream NDT results can reveal how much a connection’s apparent capacity depends on parallelism — relevant for understanding real-world application performance.
Congestion control comparison — MSAK’s ability to configure the CCA (BBR vs. Cubic) allows researchers to study how different algorithms perform across network types.
Latency under load — combining MSAK’s throughput test with UDP latency measurements allows researchers to characterize bufferbloat and latency degradation under load.
Platform development — because MSAK is the test powering M-Lab’s official speed test, its data represents the current operational measurement load on the platform.