xref: /freebsd/share/man/man4/cc_dctcp.4 (revision 6fa42b91ca3f481912af98c4d49c44507eb1b8e1)
1.\"
2.\" Copyright (c) 2014 Midori Kato <katoon@sfc.wide.ad.jp>
3.\" Copyright (c) 2014 The FreeBSD Foundation
4.\" All rights reserved.
5.\"
6.\" Portions of this documentation were written at Keio University, Japan.
7.\"
8.\" Redistribution and use in source and binary forms, with or without
9.\" modification, are permitted provided that the following conditions
10.\" are met:
11.\" 1. Redistributions of source code must retain the above copyright
12.\"    notice, this list of conditions and the following disclaimer.
13.\" 2. Redistributions in binary form must reproduce the above copyright
14.\"    notice, this list of conditions and the following disclaimer in the
15.\"    documentation and/or other materials provided with the distribution.
16.\"
17.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
21.\" ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27.\" SUCH DAMAGE.
28.\"
29.Dd November 8, 2022
30.Dt CC_DCTCP 4
31.Os
32.Sh NAME
33.Nm cc_dctcp
34.Nd DCTCP Congestion Control Algorithm
35.Sh DESCRIPTION
36The DCTCP (data center TCP) congestion control algorithm aims to maximise
37throughput and minimise latency in data center networks by utilising the
38proportion of Explicit Congestion Notification (ECN) marks received from capable
39hardware as a congestion signal.
40.Pp
41DCTCP uses fraction of ECN marked packets to update congestion window.
42The window reduction ratio is always <= 1/2.
43Only when all of the packets are
44marked, congestion window is halved.
45.Pp
46In order to keep the accuracy of the ECN marked fraction, a DCTCP receiver
47mirrors back incoming (or missing) CE marks by setting (or clearing) ECE marks.
48This feedback methodology is also adopted when the receiver uses delayed ACK.
49.Pp
50The
51.Fx
52DCTCP implementation includes two minor modifications for the one-sided
53deployment.
54Considering the situation that DCTCP is used as sender and classic
55ECN is used as receiver, DCTCP sets the CWR flag as the reaction to the ECE
56flag.
57In addition, when classic ECN is used as sender and DCTCP is used as
58receiver, DCTCP avoids to mirror back ACKs only when the CWR flag is
59set in the incoming packet.
60.Pp
61The other specifications are based on the paper and the RFC referenced
62in the
63.Sx SEE ALSO
64section below.
65.Sh MIB Variables
66The algorithm exposes the following tunable variables in the
67.Va net.inet.tcp.cc.dctcp
68branch of the
69.Xr sysctl 3
70MIB:
71.Bl -tag -width ".Va slowstart"
72.It Va alpha
73The initial value to estimate the congestion on the link.
74The valid range is from 0 to 1024, where 1024 reduces the congestion
75window to half, if a CE is observed in the first window and
76.Va alpha
77could not yet adjust to the congestion level on that path.
78Default is 1024.
79.It Va shift_g
80An estimation gain in the
81.Va alpha
82calculation.
83This influences the responsiveness when adjusting alpha
84to the most recent observed window.
85Valid range from 0 to 10, the default is 4, resulting in an effective
86gain of 1 / ( 2 ^
87.Va shift_g
88), or 1/16th.
89.It Va slowstart
90A flag if the congestion window should be reduced by one half after slow start.
91Valid settings 0 and 1, default 0.
92.It Va ect1
93Controls if a DCTCP session should use IP ECT(0) marking when sending out
94segments (default), or ECT(1) marking making use of L4S infrastructure.
95Changes to this setting will only affect new sessions, existing sessions will
96retain their previous marking value.
97.El
98.Sh SEE ALSO
99.Xr cc_cdg 4 ,
100.Xr cc_chd 4 ,
101.Xr cc_cubic 4 ,
102.Xr cc_hd 4 ,
103.Xr cc_htcp 4 ,
104.Xr cc_newreno 4 ,
105.Xr cc_vegas 4 ,
106.Xr mod_cc 4 ,
107.Xr tcp 4 ,
108.Xr mod_cc 9
109.Rs
110.%A "Mohammad Alizadeh"
111.%A "Albert Greenberg"
112.%A "David A. Maltz"
113.%A "Jitendra Padhye"
114.%A "Parveen Patel"
115.%A "Balaji Prabhakar"
116.%A "Sudipta Sengupta"
117.%A "Murari Sridharan"
118.%T "Data Center TCP (DCTCP)"
119.%U "http://research.microsoft.com/pubs/121386/dctcp-public.pdf"
120.%J "ACM SIGCOMM 2010"
121.%D "July 2010"
122.%P "63-74"
123.Re
124.Rs
125.%A "Stephen Bensley"
126.%A "Dave Thaler"
127.%A "Praveen Balasubramanian"
128.%A "Lars Eggert"
129.%A "Glenn Judd"
130.%T "Data Center TCP (DCTCP): TCP Congestion Control for Data Centers"
131.%U "https://tools.ietf.org/html/rfc8257"
132.Re
133.Sh HISTORY
134The
135.Nm
136congestion control module first appeared in
137.Fx 11.0 .
138.Pp
139The module was first released in 2014 by Midori Kato studying at Keio
140University, Japan.
141.Sh AUTHORS
142.An -nosplit
143The
144.Nm
145congestion control module and this manual page were written by
146.An Midori Kato Mt katoon@sfc.wide.ad.jp
147and
148.An Lars Eggert Mt lars@netapp.com
149with help and modifications from
150.An Hiren Panchasara Mt hiren@FreeBSD.org
151