netdev - Re: [RFC PATCH net-next] tcp: add CDG congestion control

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CA++eYdvtH22M-7t=-aO4mZHRuNRYBvgSXkuHYRRtPN_bjCpzVw@mail.gmail.com>
Date:	Thu, 4 Jun 2015 17:37:14 +0200
From:	Kenneth Klette Jonassen <kennetkl@....uio.no>
To:	Stephen Hemminger <stephen@...workplumber.org>
Cc:	Kenneth Klette Jonassen <kennetkl@....uio.no>,
	netdev <netdev@...r.kernel.org>,
	David Hayes <davihay@....uio.no>,
	Yuchung Cheng <ycheng@...gle.com>,
	Andreas Petlund <apetlund@...ula.no>,
	Dave Taht <dave.taht@...il.com>
Subject: Re: [RFC PATCH net-next] tcp: add CDG congestion control

On Tue, May 19, 2015 at 12:16 AM, Stephen Hemminger
<stephen@...workplumber.org> wrote:
>
> I am really happy to see more congestion control development.

:)

>
> Module parameters make for rather ugly tuning.
> They are ok for testing but difficult to use in real life.

CDG's throughput-delay tradeoff is affected by the parameters window,
backoff_beta, and backoff_factor. Most users will likely use the
default parameters, but others might want to tune them for less
queueing delay, more throughput, slower backoff response, etc.

window is a difficult parameter to get right when tuning; it is
intrinsically RTT-dependent, but also influences CDG's RTT-independent
backoff probability..  For the record, CDG as a congestion control is
still RTT-dependent, e.g., it uses NewReno to grow cwnd.

The upcoming patch renames ineffective_thresh to detect_ineff, and
removes ineffective_hold. Grepping experiment logs, I never once saw
the hold function kick in. A threshold of 5 is also overly
conservative when limiting the shadow window to 2*cwnd; a threshold of
1 is a better starting point.

Thanks to Eric, Yuchung and Stephen for your feedback!
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html