Overview
Some non-DNS test types, including Web and API, do generate DNS requests and capture DNS resolution time as a metric. With the exception of the DNS Experience test, all Catchpoint tests which perform DNS resolution use a commercial DNS resolver housed in the same box as the agent. This resolver honors TTLs and will retry with the next NS server on the same level if there is a NS server failure.
For DNS Experience tests, requests are not retried unless specified using the option shown below. We generally recommend leaving this setting turned off in order to help identify live issues that might otherwise be missed. The DNS Experience test can be set to honor TTLs with the setting Cache TLD Query Levels. Additionally, the agent has the option to favor NS servers with lower round trip time as well as send EDNS Client Subnet for geo-targeting.
Favoring Lower RTT Servers
Different DNS servers have different implementations of a feature that allows them to prefer NS servers with lower RTT (round trip time). The DNS server keeps track of the RTT to each NS server and prefers servers that have lower RTT for a percentage of the queries. One potential misconception is that it will do this 100% of the time, but that is not true; the server will still send a percentage of queries to the other NS servers in order to measure their RTT and determine whether one of them is now faster, and should be prefered (a type of A/B testing).
Another factor impacting the feature is how often the resolver sees a particular NS server. The more frequently it sees it - the more reliable the historical RTT data. This feature works best for domains that are accessed frequently by end-users, as the NS servers will have a larger sampling of RTT data for those domains.
In our case, the Favor Fastest Round-Trip Name Server option tells the agent to send 80% of requests to the server with the current fastest Round Trip Time.