On Tue, 15 Jul 2025 09:23:35 -0000, Andrew McConachie via rssac-caucus wrote:
On 15 Jul 2025, at 09:27, James Olorundare <olorundarek@gmail.com> wrote:
Distinguished colleagues,
I am preparing a presentation for an upcoming regional event, and a crucial component is a comprehensive overview of the DNS Root Zone. This includes publicly verifiable, up-to-date details on its operational intricacies, particularly within Africa. I need specific information on the operators of Root servers on the continent, their locations, deployment dates, and other operational specifics, all of which are good for public dissemination.
I aim to present a clear, concise, and highly technical perspective on this foundational internet infrastructure. I will delve into how these critical components enhance local internet resilience and reduce latency across Africa, highlighting the strategic importance of their presence.
Your assistance in compiling this precise data will be invaluable for a compelling and informative presentation (URLs of relevant docs will be appreciated, as well as any other relevant info).
Hi James,
The map here may be useful for showing instances in Africa.
https://rssac002.root-servers.org/site_detail_map.html
It's important to emphasize in your talk that, to the extent that latency even matters, geographic proximity will not necessarily result in reduced latencies to instances. Routing and peering policies matter more than geography when determining hop-count to root server instances.
Sounds like an interesting talk, James. I like that you're considering both resilience and latency. To add to what Andrew said about latency, with some references (since you asked for technical). This paper compares end-user-visible latency for both the Root Server System and CDNs: Thomas Koch, Ke Li, Calvin Ardi, Ethan Katz-Bassett, Matt Calder, and John Heidemann. Anycast in Context: A Tale of Two Systems. In _Proceedings of the ACM SIGCOMM Conference _, Virtual, ACM. August, 2021. <https://doi.org/10.1145/3452296.3472891>, <https://ant.isi.edu/%7ejohnh/PAPERS/Koch21a.html>. It shows that latency to the root server system should not affect visible performance for normal web use, although it definitely matters for CDNs. The paper is fairly technical, to support that claim, but the TL;DR version is: once you've cached .com and handful of other TLDs that you use frequently, you only need to go to the root for typos. Hopefully an extra 50ms to generate an error on a typo does not ruin your day. This short paper shows that most DNS recursives will select a low-latency site when given a choice, although they sometimes probe other sites: Moritz Mueller, Giovane C. M. Moura, Ricardo de O. Schmidt, and John Heidemann. Recursives in the Wild: Engineering Authoritative DNS Servers. In _Proceedings of the ACM Internet Measurement Conference_, pp. 489-495. London, UK, ACM. 2017. <https://doi.org/10.1145/3131365.3131366>, <https://ant.isi.edu/%7ejohnh/PAPERS/Mueller17b.html>. (Such selection is important in the Root Server System---with 13 different root IPs, if one is closer than others you want to prefer it. Fortunately most recursive resolvers will.) This paper expands on site placement and latency (it pre-dates the Koch paper): Ricardo de O. Schmidt, John Heidemann, and Jan Harm Kuipers. Anycast Latency: How Many Sites Are Enough?. In _Proceedings of the Passive and Active Measurement Conference_, pp. 188-200. Sydney, Australia, Springer. March, 2017. <https://ant.isi.edu/%7ejohnh/PAPERS/Schmidt17a.html>. Hopefully these papers cover the latency question pretty well. Resilience is a much bigger issue, with DDoS and cable cuts the biggest threats. I'll leave those references to another time. -John