Re:The Attempted Corporate Takeover of .Org
by C1sc0cat on Thursday January 23, 2020 @05:20AM
(#22126324)
I worked for a Registry (.coop) and sat next to the lead, interestingly we (poptel) also bid for .org back in the day and lost to the shysters that are selling out
Your right you have to have multiply redundant systems in four continents with everything running on HA kit with auto switch over diverse routing for power and data and I think hot standbys as well.
Re:The Attempted Corporate Takeover of .Org
by toast0 on Thursday January 23, 2020 @02:28AM
(#22125598)
I don't know if there are stats available, but the requests per second has got to be huge.
There are a ton of ISPs, and many of them run regional recursive servers. Each of those servers is going to see a ton of different domains that they need to query at the registry servers. Most registries are serving glue records with a 1-3 day TTL, but ISP caches may not have enough memory to keep the long tail names in cache for that long, and the long tail is long. There's millions of domain names to query the registry servers for.
I believe most people, when executing a name look up, will usually just end up talking to one of a battery of ISP provided DNS servers. So while the registry's name servers need to feed those servers (and probably be heavily redundant and reliable) they wouldn't actually end up handling an insane amount of traffic. I'm actually a bit curious now if there are any sorts of requirements on how much traffic registrars need to be able to support or if everyone just over-resources and still makes enough money that the question has never been deeply investigated.