For those of you that have missed work happening in TSVWG – there are recent 
L4S and NQB docs that specify a method to achieve low latency via a 2nd network 
queue at bottleneck links. That depends on a client marking traffic using ECN – 
via ECT(1) or CE – and/or DSCP-45.

I am in the midst of leading a field trial of this technology in Comcast’s 
network and we’ve done some very limited testing of an on-network recursive 
resolver using DNS traffic marked on the client-side as DSCP-45 (we’d use ECN 
but have not yet implemented a supporting congestion control algorithm on the 
server). Initial results from a very small scale test suggest a 20% reduction 
in QRT for cached responses.

We’re now looking for ideas on larger scale, longer-running DNS tests that 
average end users would be able to run – either by using a small probe that 
they’d install (a la RIPE Atlas) or app/script they could let run in the 
background on their computer.

If you have suggestions on good ways to expand this low latency DNS testing or 
would like us to test something you developed – please let me know off-list.

Thanks!!
Jason

More info:
https://datatracker.ietf.org/doc/html/draft-ietf-tsvwg-nqb
https://www.rfc-editor.org/rfc/rfc9330.html
https://www.rfc-editor.org/rfc/rfc9331.html
https://www.rfc-editor.org/rfc/rfc9332.html
https://corporate.comcast.com/stories/comcast-kicks-off-industrys-first-low-latency-docsis-field-trials
https://github.com/jlivingood/IETF-L4S-Deployment/blob/main/Week5-PartB.md
_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop

Reply via email to