Attendees:  rscheff@, thj@

====== Discussion ======

Cubic:

Latest base stack cubic observations. Fast Convergence should be kicking in the same way during initial slow-start, as well as after-idle using the heuristic in the RFC.

  * [[https://reviews.freebsd.org/D24657|D24657]]: Only updated comment, committed.
  * [[https://reviews.freebsd.org/D25746|D25746]]: Some more logic oversights in handling cubic variable updates. Needs review. Testing data [[https://github.com/ccui1000/freebsd-tcp/blob/FreeBSD_head/images/cwnd_ssthresh_chart_5202_beforePatch_D25746.pdf|before this Patch]], showing odd cwnd jumps after loss recovery, and [[https://github.com/ccui1000/freebsd-tcp/blob/FreeBSD_head/images/cwnd_ssthresh_chart_5202_afterPatch_D25746.pdf|after D25746 updated the behavior]]

  * [[https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=241958|PR241958]]: Not sure what happens there. tuexen will try to reproduce it on stable/12 in VBox. thj@ has reproduced this.
High Interrupt rate with FBSD12, with HPTS - but none of the stacks using it have MFC'd to -12, so why would anyone compile in HPTS in the first place.

  * Can the TCP_STATS framework be turned on and sample no connections? Yes. Still on thj@ 's list. 

  * [[https://reviews.freebsd.org/D22438|D22438]]: Upcoming: Performance improvement for after idle -> need to ask TCP afflianados to validate this change conforms with RFC2861.

Next meeting: 13 August @ 1400 UTC