> One of the things bugging me lately is that we actually have a lot of
> forms of "slow start"
> on the table - HyStart, Initial Spreading, reno vs cubic, dctcp, IW2,
> IW4, IW10, TSO offloads, the effect of GRO on it, etc. I dont know
> what is in QUIC, either.
>
> I would love a comprehensive guide to exactly the behaviors of "slow
> start" in every tcp known to man and some sane way to refer to them
> all in a cross reference and a spreadsheet.
That reminded me of a loose end from something I noticed a year or so
ago... that hystart is sometimes problematic, and the magic numbers
in hystart vary in different versions of linux. It is problematic in
that sometimes hystart limits the throughput rather dramtically
(compared to what can be obtained when switching linux's
tcp_congestion_control algorithm to "reno") and it is also problematic
that different versions of linux have different versions of hystart
(with different magic numbers compiled in).
The situation I recall was where hystart was limiting the congestion
window in situations where one of its two mechanisms was being
triggered prematurely over a short (low RTT) path that involved
802.11n or 802.11ac wireless hop with high capacity which was the
bottleneck.
And this reminds me of Bob's other observations on this list yesterday
regarding what he found in Linux code. (But in that case, it was, in
linux PIE which is not used by default anywhere AFAIK, but both of
hystart's mechanisms seem to be on by default in linux's default TCP
congestion control algorithm these days. Oops.)
-Tim Shepard
[email protected]
_______________________________________________
aqm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/aqm
