Differential D20549 Diff 60209 share/man/man4/cc_dctcp.4

Changeset View

Standalone View

share/man/man4/cc_dctcp.4

	Show All 20 Lines
	.\" ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL			.\" ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
	.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS			.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
	.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)			.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT			.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
	.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY			.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
	.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF			.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
	.\" SUCH DAMAGE.			.\" SUCH DAMAGE.
	.\"			.\"
	.\" $FreeBSD$			.\" $FreeBSD$
				tuexenUnsubmitted Done Inline Actions This should not be changed. The patch should be based on head sources, not not on a release engineering branch. tuexen: This should not be changed. The patch should be based on head sources, not not on a release…
	.\"			.\"
	.Dd January 12, 2015			.Dd July 28, 2015
				tuexenUnsubmitted Done Inline Actions Please bump the date. tuexen: Please bump the date.
				tuexenUnsubmitted Not Done Inline Actions We have `2019`, not `2015` anymore. tuexen: We have `2019`, not `2015` anymore.
	.Dt CC_DCTCP 4			.Dt CC_DCTCP 4
	.Os			.Os
	.Sh NAME			.Sh NAME
	.Nm cc_dctcp			.Nm cc_dctcp
	.Nd DCTCP Congestion Control Algorithm			.Nd DCTCP Congestion Control Algorithm
	.Sh DESCRIPTION			.Sh DESCRIPTION
	The DCTCP (data center TCP) congestion control algorithm aims to maximise			The DCTCP (data center TCP) congestion control algorithm aims to maximise
	throughput and minimise latency in data center networks by utilising the			throughput and minimise latency in data center networks by utilising the
	Show All 25 Lines
	.Sx SEE ALSO			.Sx SEE ALSO
	section below.			section below.
	.Sh MIB Variables			.Sh MIB Variables
	The algorithm exposes the following tunable variables in the			The algorithm exposes the following tunable variables in the
	.Va net.inet.tcp.cc.dctcp			.Va net.inet.tcp.cc.dctcp
	branch of the			branch of the
	.Xr sysctl 3			.Xr sysctl 3
	MIB:			MIB:
	.Bl -tag -width ".Va alpha"			.Bl -tag -width ".Va dctcp_shift_g"
				tuexenUnsubmitted Not Done Inline Actions This should read .Bl -tag -width ".Va slowstart" tuexen: This should read ``` .Bl -tag -width ".Va slowstart" ```
	.It Va alpha			.It Va alpha
	An initial estimator of the congestion on the link.			The initial value to estimate the congestion on the link.
				The valid range is from 0 to 1024, where 1024 reduces the congestion
				window to half, if a CE is observed in the first window and
				.Va alpha
				could not yet adjust to the congestion level on that path.
				bcrUnsubmitted Done Inline Actions Looks like trailing whitespace after the sentence stop. bcr: Looks like trailing whitespace after the sentence stop.
	Default is 0.			Default is 0.
	.It Va dctcp_shift_g			.It Va dctcp_shift_g
				tuexenUnsubmitted Done Inline Actions The name of the variable is `shift_g`, not `dctcp_shift_g`. (Sorry for not catching it earier. tuexen: The name of the variable is `shift_g`, not `dctcp_shift_g`. (Sorry for not catching it earier.
	An estimation gain in the alpha calculation.			An estimation gain in the
	Default is 16.			.Va alpha
				calculation.
				bcrUnsubmitted Done Inline Actions You need to make a line break after each sentence stop. bcr: You need to make a line break after each sentence stop.
				This influences the responsiveness when adjusting alpha
				to the most recent observed window.
				tuexenUnsubmitted Done Inline Actions I think you need to macros around dctcp_shift_g. I'll leave this to Benedict. tuexen: I think you need to macros around dctcp_shift_g. I'll leave this to Benedict.
				Valid range from 0 to 10, the default is 4, resulting in an effective
				gain of 1 / ( 2 ^
				.Va dctcp_shift_g
				tuexenUnsubmitted Done Inline Actions `shift_g` instead of `dctcp_shift_g`. tuexen: `shift_g` instead of `dctcp_shift_g`.
				), or 1/16th.
	.It Va slowstart			.It Va slowstart
	A trigger to halve congestion window after slow start.			A flag if the congestion window should be reduced by one half after slow start.
	Default does nothing to halve window.			Valid settings 0 and 1, default 0.
	.El			.El
	.Sh SEE ALSO			.Sh SEE ALSO
	.Xr cc_chd 4 ,			.Xr cc_chd 4 ,
	.Xr cc_cubic 4 ,			.Xr cc_cubic 4 ,
	.Xr cc_hd 4 ,			.Xr cc_hd 4 ,
	.Xr cc_htcp 4 ,			.Xr cc_htcp 4 ,
	.Xr cc_newreno 4 ,			.Xr cc_newreno 4 ,
	.Xr cc_vegas 4 ,			.Xr cc_vegas 4 ,
	Show All 25 Lines
	.Sh HISTORY			.Sh HISTORY
	The			The
	.Nm			.Nm
	congestion control module first appeared in			congestion control module first appeared in
	.Fx 11.0 .			.Fx 11.0 .
	.Pp			.Pp
	The module was first released in 2014 by Midori Kato studying at Keio			The module was first released in 2014 by Midori Kato studying at Keio
	University, Japan.			University, Japan.
	.Sh AUTHORS			.Sh AUTHORS
				tuexenUnsubmitted Done Inline Actions Is this correct? tuexen: Is this correct?
				rscheffAuthorUnsubmitted Done Inline Actions No, spurious leftover from learing fraction of nroff markup rscheff: No, spurious leftover from learing fraction of nroff markup
	.An -nosplit			.An -nosplit
	The			The
	.Nm			.Nm
	congestion control module and this manual page were written by			congestion control module and this manual page were written by
	.An Midori Kato Mt katoon@sfc.wide.ad.jp			.An Midori Kato Mt katoon@sfc.wide.ad.jp
	and			and
	.An Lars Eggert Mt lars@netapp.com			.An Lars Eggert Mt lars@netapp.com
	with help and modifications from			with help and modifications from
	.An Hiren Panchasara Mt hiren@FreeBSD.org			.An Hiren Panchasara Mt hiren@FreeBSD.org