Icann Says Some Isps Now Send Every DNS Query Three Times

Google has sorted out the Chromium bug that had its browser hammering the root DNS servers, but ICANN now says some ISPs with an unusual idea of resilience have become the newest source of wasted DNS traffic, most likely with good intentions.

Chrome’s Load on Root Servers Has Faded

The finding came from ICANN’s office of the Chief Technology Officer on Monday, at the organisation’s yearly conference held in Bali, Indonesia, during a session on fresh numbers from the Identifier Technology Health Indicators (ITHI) program, which tracks how secure, stable and resilient the domain name system is.

Maud Durand, a principal technologist at ICANN, started with what she called Chromioids, the strange Chrome flaw that came out of a well-intended attempt to guard against a kind of DNS hijacking, and in 2020 it meant the browser was behind roughly half of all the traffic hitting the DNS root servers. With Chrome on so many machines, nobody wanted that load. ICANN and other internet bodies asked Google to fix it, and Durand said Google got there bit by bit. ICANN has carried on measuring Chrome’s effect since then. The chart Durand showed has Chrome’s slice of root server traffic dropping to just one percent now, down from 45 percent, which she described as being “in the noise”.

Isps Asking Three Resolvers at Once

ICANN has since spotted a new kind of strange DNS traffic, and it comes from ISPs that hand every customer request to several DNS servers at the same time. Durand said each query goes out three times, once to Google, once to Cloudflare and once to a third resolver, and she called it belt-and-suspenders thinking, since one resolver is normally enough to keep things working.

Most network teams set up a primary DNS resolver and a secondary DNS resolver, so one can step in if the other stops answering. DNS resolvers rarely go down anyway, and sending the same lookup to several of them at once is not normal practice.

The ISPs that ICANN has seen send their DNS requests over both IPv6 and IPv4. For requests that are not cached they send them again within 100 milliseconds about half the time, with another 30 percent repeated within thirty seconds. Durand said ICANN can’t yet tell if the ISPs have a bug or if they set things up this way on purpose. Durand’s personal theory is that the ISPs doing this struggle to hire good network engineers, so they came up with a brute-force fix for resilience, and ICANN is still looking for the real cause. ICANN does not think the ISPs doing this are big players. But ICANN worries that the damage could be serious if more ISPs start doing the same thing, or if one big company does.

Fewer Resolvers Doing More of the Work

Busy DNS servers handling pointless traffic matter for another reason too, which is that ICANN believes the market for DNS services is getting more concentrated. ICANN tried to work out how many resolvers are visible to half of internet users when they go online, and its chart shows the number falling from close to 250 resolver farms back in 2020 to a little under 150 today. Durand is not sure why that has happened, but she suspects that fewer companies are running DNS resolvers and that the ones still visible mostly belong to big tech firms.

Subscribe to our weekly news digest.

Join now and become a part of our fast-growing community.

Invalid Email Address
Thanks for Subscribing!
We'll be sending you our best soon!
Something went wrong, please try again later