Evolution of M-Lab's Geographic and Network Annotations

In our recent roadmap post, we shared a list of milestones that the team is working on this and last quarter. Our Datatype migration and Standardized Columns milestone references the gardener service, which maintains and reprocesses M-Lab data, as well as the UUID annotator, that generates and saves per-connection metadata as annotations to user-conducted measurements. This post provides more detailed information about how these services have annotated measurements with geographic and network information in the past and present, and expands on what current work is happening now as mentioned in our roadmap post.

Evolution of Annotations

Over our history, the M-Lab platform has changed quite a bit. Our post last year, Evolution of NDT, discussed changes to the NDT measurement service over time, and touched on changes to our server instrumentation. Similarly, how we have annotated measurements has evolved over this time as well.

When you choose to run one of the tests M-Lab provides, you’re first connected to one of our available servers nearest to you, through which the measurement is conducted. That server collects the measurement, but actually not much else. As measurements are collected, the IP address is used to provide annotations that provide more context and usefulness to research and analysis. For example, a big part of what makes NDT data interesting and useful are the annotated fields allowing people to explore and aggregate by geography.

The values each of our tests collect varies depending on the measurement service, but as an example, the NDT performance test collects:

  • the packet headers collected during the test, used to calculate the measurement values
  • the measurement values you see at the end of the test
  • the IP address assigned by ISPs to routers, modems, or other on-premises devices

The table below describes the historical changes to geographic annotations from 2009 to the present, as well as changes to the two sources of TCP statistics used by M-Lab servers over our history:

Geo Annotations Dates in use TCP Statistics Source Dates in use
geo1, annotation-service 2009-01 - 2017-08 web100 2009-01 - 2019-11
geo2, annotation-service 2017-09 - 2020-02 tcpinfo 2019-11 - present
geo2, uuid-annotator 2020-03 - present    

For Geo annotations, there are three time periods of interest:

  • 2009-01 to 2017-08 - maxmind geo1 from annotation-service
  • 2017-09 to 2020-02 - maxmind geo2 from annotation-service
  • 2020-03 to present - maxmind geo2 from uuid-annotator

and for the Platform’s TCP Statistics, there are two time periods of interest:

  • 2009-01 to 2019-11 - web100 platform
  • 2019-06 to present - tcpinfo platform

geo1 annotations were provided on web100 datatypes between 2009-01 and 2017-08, using the annotation-service. This service sourced geographic annotations for IP addresses from the Maxmind GeoLite Legacy database edition, and IP address information from the Routeviews dataset. Maxmind discontinued in 2019-01. Relevant to annotations of M-Lab data, the Region field in Maxmind GeoLite Legacy used the FIPS-10-4 standard.

geo2 annotations were provided on the ndt5 datatype between 2017-09 and 2020-02, also using the annotation-service. This change replaced geo1, and uses IP geolocation data from the Maxmind GeoLite2 dataset and IP address information from the Routeviews dataset. The Maxmind GeoLite2 dataset uses the ISO 3166-2 standard for the principle subdivisions within a country. For more information, see this update from Maxmind.

In 2020-03, the uuid-annotator service was launched, providing annotations for the ndt7 datatype, using the Maxmind Geolite2 dataset, and IP address information from the Routeviews dataset. The uuid-annotator will eventually replace the annotation-service entirely. The uuid-annotator also provides previously unused fields from Maxmind Geolite2, such as the ISO 3166-2 second level subdivisions within a country, AS Name to complement AS Number, and the accuracy radius of the IP address geolocation.

Current Milestones In Support of Long Term Supported Schemas Using Standard BigQuery Columns

To enable our goal of long term schema support and consistent annotations across our archives, our team is working now to migrate the remaining datatypes (tcpinfo, ndt5, ndt.web100, traceroute1, traceroute2, sidestream) from the annotation-service to the uuid-annotator. One part of this process is reconciling or mapping FIPS-10-4 and ISO 3166-2 subdivision codes, and adding support in the gardener service to re-annotate correct ISO codes in the new standard fields in our schemas. Accurate annotations for our servers are also being added as well. For data collected prior to 2020-03, a process of exporting synthetic uuid annotations and UUIDs is being designed so that we can preserve the original and the new measurement annotations, and retire the annotation-service.

Wrap up

We hope that this post provides more detail on some items in our roadmap that the team is currently working on completing. As always, if you have questions please reach out on support@measurementlab.net.

Back to Top