Order name servers by SRTT

[nhurley3]

Calculates the smoothed round-trip time (SRTT) using an exponentially weighted moving average and orders the name servers by the value. The implemented algorithm is similar to what was discussed in #1702.

Fixes #1702 (originally filed by @peterthejohnston)

[nhurley3]

@bluejekyll and @djc, here's a draft PR that implements a version of the algorithm mentioned in #1702 (comment). Let me know what you think. Thanks!

[bluejekyll]

Ok, so with these changes, we’ll be ignoring the success and failures from the algorithm and only paying attention to request time. I think I’m ok with this, but it’s worth a discussion before landing.

[nhurley3]

Ok, so with these changes, we’ll be ignoring the success and failures from the algorithm and only paying attention to request time. I think I’m ok with this, but it’s worth a discussion before landing.

This algorithm doesn't explicitly track the number of successes/failures, but they are still incorporated in some fashion. Successes now correspond to recorded latencies. Similarly, failures result in a penalty being applied to the SRTT value, which serves as a backoff.

Coupling the failure state with the SRTT value is somewhat of an arbitrary decision that simplifies things. One potentially interesting alternative could be to effectively remove failed servers from the rotation and probe them in the background (somewhat similar to what Unbound does). Of course, the tradeoff is increased complexity and additional resource cost. Happy to consider other alternatives as well. There's certainly more than 1 reasonable approach here.

[codecov]

Codecov Report

Merging #1784 (7bf15a6) into main (a88a7a5) will increase coverage by 0.09%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #1784      +/-   ##
==========================================
+ Coverage   80.03%   80.12%   +0.09%     
==========================================
  Files         177      177              
  Lines       18108    18196      +88     
==========================================
+ Hits        14492    14580      +88     
  Misses       3616     3616              

Impacted Files	Coverage Δ
crates/resolver/src/name_server/name_server.rs	96.89% <100.00%> (-0.80%)	⬇️
...ates/resolver/src/name_server/name_server_stats.rs	99.24% <100.00%> (+3.89%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a88a7a5...7bf15a6. Read the comment docs.

[bluejekyll]

Thanks for the work and getting this completed. It looks like a really good improvement. Merging.