How RootViz works

This is the documentation on how we generate the graphs for https://rootviz.sidnlabs.nl

How we generate these graphs

We measure the reachability of each Root Server Letter (A--M) by doing the following:

1. Download every 30min all results from the RIPE Atlas Root DNS Measurements:
   • All probes (10k+) send chaos TXT hostname.bind to each Root Server Letter direclty, bypassing their resolvers
   • For instance, see K-ROOT IPv6 measurement. Other Root Sever Letter IDs can be found here
     • (they have been used in multiple studies, such as this one
2. For each DNS response obtained, we look into their result variable (see RIPE Atlas documentation):

   "result" -- variable content, depending on type of response (array)
   Each element is an associative array consisting of:
   Case: Timeout
   "x" -- "*" (string)
   Case: Error
   "error" -- description of error (string)
   Case: Reply
   "rtt" -- round-trip-time in milliseconds (float)
   "ttl" -- [optional] time-to-live reply if different from ttl in first reply (int)
   "dup" -- [optional] signals that the reply is a duplicate (int)

3. We count, for each measurement ID, how many unique prb_id (Atlas Probe unique ID) had an timeout:
   • Timeout means the probe was active by somehow it could not reach the designated IP address
4. We show the results in a time series

Details and caveats:

• We know that a single probe is not reliable -- after all, they can be interecepted, hijacked, and so forth
• However, what we do here is a bit of crowd sourcing: if many probes report an issue, therefore it must be something with the server and not the clients
• We were able to spot and confirm with two different Root Sever Letter operators events that happened and last ~ 30mins
  • We can look at the catchment of these probes before an event and see which anycast site likely had issue
• We are working on other metrics, and we will post them soon as other public dashboards
• Contact: giovane . moura @ sidn .nl