Re: [tcpm] About Cubic when slow start exits with no loss

I also agree. I believe it won't affect our consensus on the draft as it is
a small additional tip.
Although the doc already has been requested for publication several days
ago, I prefer adding it to the draft if there's no problem for the
procedure.
I hope we still have a little time before IESG starts reviewing it.

Thanks,
--
Yoshi

On Fri, Sep 15, 2017 at 12:44 AM, Michael Welzl <michawe@ifi.uio.no> wrote:
> Hi,
>
> I give a +1 to this request: I remember searching an answer to the same
> question, long (actually a few years) ago - I looked at the (old) I-D,
found
> nothing, and gave up after a bit of Linux code reading, at that time.
>
> I should have remembered to request this being added when the Cubic draft
> was picked up again - indeed this is an important thing to add, IMO.
>
> Cheers,
> Michael
>
>
>
> On Sep 15, 2017, at 8:57 AM, Le Boudec Jean-Yves
> <jean-yves.leboudec@epfl.ch> wrote:
>
> Dear Lisong
>
> thanks for the pointer. Shouldn't that be also described in the ID ?
>
> Best, JY
>
>
> On 14.09.2017 20:10, Lisong Xu wrote:
>
> Dear Jean-Yves,
>
> Yes, if the first slow start ends with no loss, Cubic enters congestion
> avoidance. In this case, both K (i.e., bic_K) and W_max (i.e.,
> bic_last_max_cwnd) are still their initial values (i.e., 0), and the
origin
> of the cubic function (i.e.,bic_origin_point) is set to the current
> congestion window size (i.e., cwnd). In addition, in this case, Cubic
> maintains a minimum cwnd increase speed of 5% per RTT (line 308-309 at
> http://elixir.free-electrons.com/linux/latest/source/net/ipv4/tcp_cubic.c
).
>
> Thanks
> Lisong
>
> On 9/14/2017 11:23 AM, Le Boudec Jean-Yves wrote:
>
> Hi,
>
> in draft-ietf-tcpm-cubic-05, I see that slow start is assumed, but I did
not
> find what Cubic does if slow start ends with no loss. I guess congestion
> avoidance is entered, but then how does Cubic computes W_max and K when
> congestion avoidance is entered following slow start with no loss ? It
seems
> reasonable to have K=0 and W_max set to a very large value, right ?  Did I
> miss anything ?
>
> JY
>
>
>
> _______________________________________________
> tcpm mailing list
> tcpm@ietf.org
> https://www.ietf.org/mailman/listinfo/tcpm
>
>
>
> _______________________________________________
> tcpm mailing list
> tcpm@ietf.org
> https://www.ietf.org/mailman/listinfo/tcpm
>
>
>
> _______________________________________________
> tcpm mailing list
> tcpm@ietf.org
> https://www.ietf.org/mailman/listinfo/tcpm
>