Thanks, Benny, for your note. Actually, we did not down-select probes when picking 400 — we take all the probes that are (a) dual stack, (b) we could verify with high certainty that they use ISP resolvers by default, and (c) that are labeled as stable (I forgot what the stability criterion is exactly — Nick?). We did consider reducing the number of probes from overrepresented regions but then decided against it.
—Misha > On Apr 4, 2023, at 12:03 PM, Benno Overeinder <[email protected]> wrote: > > Dear Misha and Nick, > > Thank you for sharing the results of your (MSc?) research project. Clear and > informative graphs of your measurements. > > A suggestion. When inspecting the probe maps, Europe is over-represented, > which is no surprise given the number of probes in this region. But when you > select 400 probes, you can "steer" the selection for better global coverage. > In the probe selection, you can specify the West, North/South-Central and > North/Southeast area and the number of probes per area. Perhaps this will > allow you to better spread the available probes over the different areas for > a "balanced" global measurement. > > You already present the selection of regions on the web page, but for Africa > this results in 7 (successful) probes for January 2023. > > Cheers, > > -- Benno > > > On 23/03/2023 23:56, Michael Rabinovich wrote: >> Folks, >> FYI, Nick Kernan, my Masters student, has deployed a web portal that >> facilitates the comparison between various DNS resolvers, including >> ISP-provided and some influential publicly available resolvers (Google’s >> public DNS, OpenDNS, Quad9, and Cloudflare). The comparison is based on two >> factors: the latency of DNS resolutions and the latency of TCP handshake >> with the CDN edge server selected through a given resolver. The results >> are based on recurrent monthly measurements from RIPE Atlas probes that have >> both IPv4 and v6 connectivity, have an ISP-provided resolver available for >> comparison, and which meet certain requirements regarding probe versions and >> reliability. You can specify the desired month of measurement, a region of >> interest, and the CDNs to analyze. You can access Nick’s portal through >> https://dns-web-portal.netlify.app/ <https://dns-web-portal.netlify.app/> >> The details of the methodology are in Nick’s thesis, which awaits some >> copyright permissions before being made publicly available. In the >> meantime, Nick (cc’ed) and I would be happy to answer any questions. >> Regards, >> —Misha Rabinovich > > -- > Benno J. Overeinder > NLnet Labs > https://www.nlnetlabs.nl/ -- ripe-atlas mailing list [email protected] https://lists.ripe.net/mailman/listinfo/ripe-atlas
