Summary

  • RIPE NCC’s new Makati AuthDNS node handled more than 726 million queries on 1 July 2025, about 8,400 per second on average. Source addresses announced by AS15169 accounted for 96.5% of that day’s queries.
  • Rene Wilhelm’s measurements from 50 RIPE Atlas probes in the Philippines showed some paths moving to Makati, while other probes still reached distant AuthDNS instances, including Paris near the end of June.

The server was in Makati. The routing decision was somewhere else. Rene Wilhelm, a RIPE NCC research engineer, used the AuthDNS node’s own query ledger and RIPE Atlas measurements to show how much distance remained after a new authoritative DNS instance appeared at Maharlika IX. On 1 July 2025 it answered more than 726 million queries, averaging roughly 8,400 a second. That is a busy node, not proof that every Philippine resolver found it.

The first surprise was who filled the ledger. RIPE NCC joined its node-log source addresses to origin AS information from its own RIS collectors. Networks announced by AS15169 accounted for 96.5% of queries to the new node that day. The number identifies the source-address network, not a country, a count of users or a share of Philippine DNS demand. It also means a large QPS headline can mostly describe one network’s routing choice.

Wilhelm and contributor Qasim Lone checked the other side with RIPE Atlas. From 50 probes located in the Philippines, they asked which AuthDNS instance answered. Some began reaching the Makati node. Others still went to distant instances; Paris reappeared for some toward late June. Anycast gives multiple servers the same service address. RIPE NCC’s AuthDNS description says BGP announcements from the sites let each provider’s route selection decide which one a client reaches. Installing a node at an exchange supplies a route; it does not write every provider’s policy.

The original study also found that source IPs using Makati on 1 July still sent queries to Tokyo and other instances that same day. A client can meet different paths as routing or its own behaviour changes. Counting a query at one node says where that query was answered. It cannot, by itself, assign every resolver a permanent “local” badge or turn a before-and-after total into a clean migration count.

RIPE’s hosted-node register still listed the Maharlika IX AuthDNS instance as Operational at the November 2026 check. That status proves the node remains on the public service map; it supplies no new query-rate measurement. Wilhelm’s 2025 ledger remains the concrete test of the installation: a new physical option, a huge load from one origin AS, and Philippine vantage points taking different routes.

The next useful record is a per-network map of the BGP routes and peering choices that kept local probes on distant instances. Without it, the hardware location is visible while the decision that steers each DNS request stays with the network carrying it.