You say that like only 117 million records of very fixed-format, fixed-fields, heavily-defined data with archaic length and content constraints is a lot... Any modern RDBMS can handle that in a single instance on a decent not-too-beefy machine. Heck, a modern file system could probably do it with flat files and a directory-sharding technique without batting an eye.
Now try doing it on tech from when the internet was first invented.... Maybe that's why?
> Now try doing it on tech from when the internet was first invented.... Maybe that's why?
You could store the entirety of .com (and all other TLDs) on a midrange server. There might well be such a server that serves as the Source of Truth for the database.
The much harder part is serving those records authoritatively for billions of requests, which is why we have Root Nameservers. There are 13 of them (by DNS name) distributed around the world, all synchronized from the same SoT. The total number is actually much higher (500?), because they are redundant locally and Anycasted globally.
Also, FWIW: Domain name registration was free from inception in 1983 until 1995, at which point .com registrations were $50/yr, minimum 2 years for first registration, with only one vendor (Network Solutions).
1995 was the year of Netscape and the Internet on the cover of Time magazine. NetSol was the monopoly vendor, and 30% of that registration fee went to the U.S. Government (National Science Foundation), which was the majority funder of the post-ARPANET internet. This fee was to establish an "Internet Intellectual Infrastructure Fund". After a lawsuit, the fee was scrapped (illegal tax) and instead the entire remaining $70 fee went to the granted-monopoly private corp NetSol.
Prior to NetSol, domain registration was handled by one guy working at USC, Jon Postel. Over email. Even then, there were almost a dozen (early 90s tech) Root Nameservers to handle the query load.
Now try doing it on tech from when the internet was first invented.... Maybe that's why?