People:
Nicholas Kernan (Code author): nicholas.kernan@gmail.com, Michael Rabinovich: michael.rabinovich@case.edu
This work was supported in part by NSF through grant CNS-2219736.
Every month, we schedule measurements from ~400 Ripe-Atlas probes globally to evaluate the performance of ISP resolvers and four popular public resolvers in terms of their DNS latency and quality of CDN mappings. To view specific results choose the measurement month, CDN, IP version, and region from the dropdowns below. Clicking "submit" will display 4 plots (or 16, if all CDNs are selected): the median and mean CDFs, for DNS latency and CDN mapping (TCP) latency corresponding to your chosen options.
Probes attempt measurements to 5 websites accelerated by the chosen CDN (or 10 in the case of Akamai). CDF curves for a given
resolver are obtained by taking the mean (or median) twice: once to obtain the mean (median) reading for each probe*website combination,
then again to obtain the mean (median) latency for each probe on that CDN. Probes are sorted according to this value before plotting. Thus, a point of
(10, 100) on a plot would indicate that 100 probes had mean (median) latencies of 10 ms or better for that resolver, while the rest had a higher latency.
We can take can the mean (median) one final time across all probes to obtain the overall mean (median) latency of that resolver. Below each plot we display the
worst and best mean (median) latencies obtained this way.