Apache Druid vs DNSimple

DNSimple's latest post reports that its origin name servers use 12% less CPU per query and handle 14% more queries per second after rewriting hot paths in dns_erlang, the open-source library that decodes and encodes every query, without adding C. Name compression for long names is more than 20x faster, and the techniques made it upstream into Erlang/OTP and Elixir. It follows the August statement that a three-year migration finished and every query now runs on infrastructure DNSimple owns. Above that layer sit the official CLI launched in May, browser login from June, and Domain Trustee for restricted country-code TLDs.

The through-line is removing prerequisites — a token, a local presence, a third party in the query path — rather than adding product surface. The edge migration is the same instinct applied to the company's own dependencies: a registrar whose resolution ran partly on someone else's anycast could not fully own its availability story, and now it does. Automation remains the consistent audience for everything above the infrastructure line, with the CLI, structured output and the async Rust client all pointing at DNS managed from pipelines.

Owning the edge is the kind of investment that gets monetized next: expect DNSimple to start attaching claims it could not previously make — latency, availability, or region-level routing control — to the resolution service. What the post does not say is whether any of that reaches the product as configurable behaviour rather than background improvement.