Skip to content
Romain edited this page Nov 21, 2023 · 3 revisions

This measurement uses a DNS rendez-vous to enable "scanning" by the team of all RIPE Atlas probes towards a set of anycast prefixes. The DNS rendez-vous technique is described here: https://labs.ripe.net/author/emileaben/measuring-more-internet-with-ripe-atlas/ . The list of anycast IPs (ie. IP addresses in anycast prefixes) is generated as follows:

  • each day (at 00:00 UTC) the lists of anycasted IPv4 and IPv6 prefixes are fetched from https://github.com/bgptools/anycast-prefixes
  • For IPv4 the .1 of each prefix is put on the list , for IPv6 the ::1 of each prefix is put on the list

This measurement is potentially useful for:

  • finding IXPs close to probes (under the assumption that a decent amount of anycast is done at IXPs).
  • inspection of anycast catchment areas relative to RIPE Atlas probe deployments.
  • (as a corollary): finding which IXPs don't have RIPE Atlas probes close by

Measurement IDs:

  • IPv4: 63868161
  • IPv6: 63868190

Measurement tag: "anycast"

Considerations for measurement creation: interval is 900 (default), and the 'spread' parameter is set to 450 , to make sure the DNS rendez-vous point will be hit as distributed over time as possible (ie. avoid overload and/or all kinds of cache TTL violation effects) Paris ID is set to 1, in an attempt to capture 1 specific state between a source and dest. In case one would want to make a traceroute between the same src/dst with a Paris ID of 1, this might reproduce the same flow ID (TODO: check with probe code) The protocol was set to ICMP because limited testing revealed that for some networks (notably cloudflare), a UDP trace had multiple different IPs in the same hop, which complicates analysis.