1 of 45

Measurement Lab

​

Pavlos Sermpezis pavlos@measurementlab.net

​

@measurementlab

​

OMG 3 convening (2025-06-25)

​

​

@

2 of 45

Agenda

  • Introduction to M-Lab’s platform & data
  • The Giga/UNICEF use-case�Using the M-Lab platform to monitor school connectivity worldwide
  • The IP Route Survey (IPRS)�A new dataset by Sorbonne University published via M-Lab
  • Host Managed Deployment�Evolving M-lab’s platform: the new type of M-lab servers
  • Internet Quality Barometer (IQB)�A framework to assess and quantify Internet quality (beyond speed tests)

2

@

3 of 45

3

Introduction to M-Lab

4 of 45

M-Lab

  • Who we are?
    • M-Lab is an open source project
    • Founded in 2009
    • Fiscally Sponsored Project of Code for Science & Society
  • What we do?
    • Measure the Internet.
    • Save the data.
    • Make it universally accessible and useful.

​

4

@

5 of 45

M-Lab’s platform

5

@

6 of 45

M-Lab’s platform

6

@

7 of 45

M-Lab’s platform

  • 500+ servers globally
  • 40+ countries / 100+ metros,
  • 50+ transit providers

7

@

8 of 45

M-Lab measurements

8

@

  • On the M-Lab platform, we host the server-side of “experiments” or “measurement services”.
  • When clients run these measurements, they test against M-Lab servers.
  • Every measurement is publicly archived and published in BigQuery.

9 of 45

M-Lab measurements

9

@

  • Integrated into Google Search → majority of measurements

10 of 45

M-Lab measurements

10

@

11 of 45

M-Lab measurements

11

@

  • Other integrators: Giga/UNICEF, AirBnB, etc.

12 of 45

M-Lab measurements

12

@

  • Murakami: M-Lab’s On-premise Measurement Client
    • https://www.measurementlab.net/blog/murakami/
    • “Murakami is a software program that runs automated measurements from a computer attached to a network.”
    • Have your own fleet of devices (e.g., Raspberry Pi) or computers/users/infrastructure? �→ Use Murakami to monitor internet connectivity and performance with the regional focus, detail, and scale you need!

13 of 45

Accessing M-Lab Data

13

@

14 of 45

Accessing M-Lab Data

14

@

15 of 45

Accessing M-Lab Data

  • BigQuery examples

15

@

SELECT

a.TestTime as time,

a.MeanThroughputMbps as download

FROM `measurement-lab.ndt.ndt7`

WHERE

date BETWEEN "2021-01-01" and "2021-01-31"

AND client.Geo.CountryName = "Uganda"

ORDER BY time ASC

SELECT

a.TestTime,

a.MeanThroughputMbps,

a.MinRTT,

a.LossRate,

client.Geo.city as ClientCity,

client.Network.ASNumber as ClientASNumber,

client.Network.ASName as ClientASName,

server.Geo.City as ServerCity,

server.Network.ASNumber as ServerASNumber,

server.Network.ASName as ServerASName

FROM

`measurement-lab.ndt.unified_downloads`

WHERE

date >= "2022-07-01" AND date <= "2022-07-01"� AND client.Geo.city = "Baltimore"

AND client.Geo.CountryCode = "US"

ORDER BY a.TestTime

16 of 45

16

The Giga/UNICEF use-case

17 of 45

Giga

  • Giga https://giga.global/
    • Giga is an initiative launched by UNICEF and ITU in September 2019 to connect every school to the Internet and every young person to information, opportunity and choice
  • Giga Meter https://meter.giga.global/
    • Giga Meter is a powerful desktop application that allows school administrators to monitor and visualize the quality of their institution's connectivity

17

@

18 of 45

Giga maps

  • Giga maps https://maps.giga.global/map
    • School locations and basic connectivity information (global map)
    • Real-time connectivity status (live map)

18

@

19 of 45

Giga Meter

  • Window OS application that uses minimal data to automatically perform up to 4 daily internet measurements in schools.
    • First test runs within 15mins of device being on.
    • Schools can manually trigger tests to measure real-time internet speed at their school.
  • Data is collected over time and synced directly into Giga Maps global connectivity map.

19

@

20 of 45

Giga Meter data (@ M-Lab)

  • Regular/periodic tests - monitoring Internet performance
  • 1.5M+ measurements since 2024
  • From schools/locations in 27 countries �(including underrepresented regions!)
  • To M-Lab servers in 70+ metros

20

@

21 of 45

21

New dataset: The IP Route Survey (IPRS)

22 of 45

IPRS dataset

  • Who?�Dioptra research group at Sorbonne University, LIP6 CS Lab
  • What?�Periodic traceroute-style measurements
  • Goal �Continuously monitor IP-level routing across the Internet
  • Where?�BiqQuery @ M-Lab → https://www.measurementlab.net/tests/iprs/

22

@

23 of 45

IPRS data details

  • Regular collection of traceroute-style measurements
  • From 10 vantage points
  • To all routable IPv4 prefixes
  • Use cases
    • Get a broader picture of Internet routes
    • Correlate with M-Lab Internet measurements at a given time
      • E.g., to examine anomalies (relate to a test or as part of a broader phenomenon)
    • Compare with / complement BGP data (different vantage points, more fine-grained)
      • E.g., to see significant or unusual routing changes
    • Any ideas to leverage these data for Internet freedom initiatives?
  • Learn more:

23

@

24 of 45

24

Host Managed Deployment

25 of 45

M-Lab’s platform

  • 500+ servers globally
  • 40+ countries / 100+ metros,
  • 50+ transit providers

25

@

26 of 45

Types of M-Lab Sites

26

@

M-Lab Managed

Virtual

Physical

Minimal Site

Full Site

Host-Managed

Physical

Sponsor

Virtual

Cloud Credits

For more details: https://www.measurementlab.net/contribute/#host-or-sponsor-an-m-lab-measurement-site

27 of 45

Host Managed Deployments

  • Host Managed Deployments
  • Single server, running M-Lab software, managed by the host organization
  • Interested? → Complete the Infrastructure Contribution Form

27

@

  • Organization registers
  • Data sharing agreement
  • M-Lab published software
  • Host managed servers
  • Physical or virtual server
  • One IPv4 & IPv6 addr
  • Minimum 1 Gbps uplink
  • Minimum 4GB & 4 Core Intel
  • RNP pilot Q4 2024
  • Goal: GA 2025-2026
  • Easier to contribute
  • Access & edge networks
  • Scale hardware support

Why

​

How

​

What

​

When

28 of 45

Why have �Host Managed Deployments?�

28

@

29 of 45

Why have �Host Managed Deployments?�

There are regions (e.g., Middle East and North Africa) that are poorly covered by M-Lab’s infrastructure

29

@

30 of 45

Why have �Host Managed Deployments?�

  • Our mission:
    • Since 2008, M-Lab has worked to align efforts in industry, research, and public interest projects to measure the Internet openly
    • In March 2024, we announced: M-Lab is committed to evolving with the Internet
  • Our goal: Expand M-Lab’s platform to measure more of the Internet from new network locations and while increasing resource efficiency
    • Diversify platform servers and costs
    • Larger geographical footprint
    • Decrease operational costs

30

@

31 of 45

Host Managed Deployments

  • Current status
    • Phase 2 of experimental deployment
    • 40+ operating servers
    • Collaboration with RNP (Brazil) - 27 servers
    • Currently support only for ndt7 (i.e., no support for WeHe, reverse traceroute, DASH)
  • Next steps
    • Make the program generally available to the public
    • Data publicly available in BigQuery

31

@

32 of 45

32

Internet Quality Barometer (IQB)

33 of 45

IQB main goal

→ Redefine Internet quality beyond “speed” ←

​

33

@

16 month project funded by the Internet Society Foundation Research Grant Program

​

34 of 45

Quality Internet ≠ “High-speed”

  • Internet quality depends on the use case �& it’s not always about “speed”

​

  • E.g.,�Web browsing → latency is important�vs. �Online gaming → latency & download/upload throughput�vs. �Online backup → download/upload throughput

34

@

35 of 45

IQB

35

@

If decision-makers want to make data-driven decisions about improving the quality of the Internet, what metric(s) should they use?

�→ welcome to the IQB framework!

36 of 45

IQB score

  • The goal of IQB framework is to calculate a score �(i.e., a composite index) that characterizes the Internet Quality
  • IQB score → from 0 (low quality) to 1 (high quality)

36

@

37 of 45

IQB framework

What should we take into account?

  • Multiple use cases (video streaming, web browsing, etc.)
  • Multiple network requirements (throughput, latency, etc.)
  • Multiple measurements/datasets

37

@

38 of 45

IQB framework

  • Network requirements contribute to use cases
  • Use cases contribute to the IQB score
  • Different weights per requirement and use case

38

@

39 of 45

Designing the IQB framework

  • With the help of experts & stakeholders�(research, industry, policy-making, etc.)
  • >60 experts
  • Interviews & Workshops

39

@

40 of 45

Designing the IQB framework

  • Network requirement thresholds for minimum & high quality per use case

40

@

41 of 45

Designing the IQB framework

  • Network requirement weights per use case

41

@

42 of 45

IQB score

42

@

43 of 45

IQB goals

  • Shift the conversation around Internet Quality beyond speed
  • Help decision-makers make sense of the data
  • Empower users to make more informed decisions about their Internet
  • Use existing, openly available datasets as complementary sources
  • Advocate for the collection of more nuanced metrics

43

@

44 of 45

IQB next steps

  • IQB design report just published
  • Creation of the IQB tool
  • Assessment of findings �& iterations on framework �& new measurement methods
  • Reach out to hello@measurementlab.net if you’re interested!

44

@

→ Idea: What about extending/adapting IQB to create the IFB? � (IFB = Internet Freedom Barometer)

45 of 45

45

Thank you!

@