One document matched: draft-ietf-dccp-tfrc-voip-02.ps

%!PS-Adobe-3.0
%%Creator: groff version 1.19
%%CreationDate: Tue Jul 19 11:18:32 2005
%%DocumentNeededResources: font Courier-Bold
%%+ font Times-Bold
%%+ font Times-Roman
%%+ font Courier
%%DocumentSuppliedResources: procset grops 1.19 0
%%Pages: 23
%%PageOrder: Ascend
%%DocumentMedia: Default 612 792 0 () ()
%%Orientation: Portrait
%%EndComments
%%BeginDefaults
%%PageMedia: Default
%%EndDefaults
%%BeginProlog
%%BeginResource: procset grops 1.19 0
/setpacking where{
pop
currentpacking
true setpacking
}if
/grops 120 dict dup begin
/SC 32 def
/A/show load def
/B{0 SC 3 -1 roll widthshow}bind def
/C{0 exch ashow}bind def
/D{0 exch 0 SC 5 2 roll awidthshow}bind def
/E{0 rmoveto show}bind def
/F{0 rmoveto 0 SC 3 -1 roll widthshow}bind def
/G{0 rmoveto 0 exch ashow}bind def
/H{0 rmoveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/I{0 exch rmoveto show}bind def
/J{0 exch rmoveto 0 SC 3 -1 roll widthshow}bind def
/K{0 exch rmoveto 0 exch ashow}bind def
/L{0 exch rmoveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/M{rmoveto show}bind def
/N{rmoveto 0 SC 3 -1 roll widthshow}bind def
/O{rmoveto 0 exch ashow}bind def
/P{rmoveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/Q{moveto show}bind def
/R{moveto 0 SC 3 -1 roll widthshow}bind def
/S{moveto 0 exch ashow}bind def
/T{moveto 0 exch 0 SC 5 2 roll awidthshow}bind def
/SF{
findfont exch
[exch dup 0 exch 0 exch neg 0 0]makefont
dup setfont
[exch/setfont cvx]cvx bind def
}bind def
/MF{
findfont
[5 2 roll
0 3 1 roll
neg 0 0]makefont
dup setfont
[exch/setfont cvx]cvx bind def
}bind def
/level0 0 def
/RES 0 def
/PL 0 def
/LS 0 def
/MANUAL{
statusdict begin/manualfeed true store end
}bind def
/PLG{
gsave newpath clippath pathbbox grestore
exch pop add exch pop
}bind def
/BP{
/level0 save def
1 setlinecap
1 setlinejoin
72 RES div dup scale
LS{
90 rotate
}{
0 PL translate
}ifelse
1 -1 scale
}bind def
/EP{
level0 restore
showpage
}bind def
/DA{
newpath arcn stroke
}bind def
/SN{
transform
.25 sub exch .25 sub exch
round .25 add exch round .25 add exch
itransform
}bind def
/DL{
SN
moveto
SN
lineto stroke
}bind def
/DC{
newpath 0 360 arc closepath
}bind def
/TM matrix def
/DE{
TM currentmatrix pop
translate scale newpath 0 0 .5 0 360 arc closepath
TM setmatrix
}bind def
/RC/rcurveto load def
/RL/rlineto load def
/ST/stroke load def
/MT/moveto load def
/CL/closepath load def
/Fr{
setrgbcolor fill
}bind def
/setcmykcolor where{
pop
/Fk{
setcmykcolor fill
}bind def
}if
/Fg{
setgray fill
}bind def
/FL/fill load def
/LW/setlinewidth load def
/Cr/setrgbcolor load def
/setcmykcolor where{
pop
/Ck/setcmykcolor load def
}if
/Cg/setgray load def
/RE{
findfont
dup maxlength 1 index/FontName known not{1 add}if dict begin
{
1 index/FID ne{def}{pop pop}ifelse
}forall
/Encoding exch def
dup/FontName exch def
currentdict end definefont pop
}bind def
/DEFS 0 def
/EBEGIN{
moveto
DEFS begin
}bind def
/EEND/end load def
/CNT 0 def
/level1 0 def
/PBEGIN{
/level1 save def
translate
div 3 1 roll div exch scale
neg exch neg exch translate
0 setgray
0 setlinecap
1 setlinewidth
0 setlinejoin
10 setmiterlimit
[]0 setdash
/setstrokeadjust where{
pop
false setstrokeadjust
}if
/setoverprint where{
pop
false setoverprint
}if
newpath
/CNT countdictstack def
userdict begin
/showpage{}def
/setpagedevice{}def
}bind def
/PEND{
clear
countdictstack CNT sub{end}repeat
level1 restore
}bind def
end def
/setpacking where{
pop
setpacking
}if
%%EndResource
%%BeginFeature: *PageSize Default
<< /PageSize [ 612 792 ] /ImagingBBox null >> setpagedevice
%%EndFeature
%%IncludeResource: font Courier-Bold
%%IncludeResource: font Times-Bold
%%IncludeResource: font Times-Roman
%%IncludeResource: font Courier
grops begin/DEFS 1 dict def DEFS begin/u{.001 mul}bind def end/RES 72
def/PL 792 def/LS false def/ENC0[/asciicircum/asciitilde/Scaron/Zcaron
/scaron/zcaron/Ydieresis/trademark/quotesingle/Euro/.notdef/.notdef
/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef
/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef/.notdef
/.notdef/.notdef/space/exclam/quotedbl/numbersign/dollar/percent
/ampersand/quoteright/parenleft/parenright/asterisk/plus/comma/hyphen
/period/slash/zero/one/two/three/four/five/six/seven/eight/nine/colon
/semicolon/less/equal/greater/question/at/A/B/C/D/E/F/G/H/I/J/K/L/M/N/O
/P/Q/R/S/T/U/V/W/X/Y/Z/bracketleft/backslash/bracketright/circumflex
/underscore/quoteleft/a/b/c/d/e/f/g/h/i/j/k/l/m/n/o/p/q/r/s/t/u/v/w/x/y
/z/braceleft/bar/braceright/tilde/.notdef/quotesinglbase/guillemotleft
/guillemotright/bullet/florin/fraction/perthousand/dagger/daggerdbl
/endash/emdash/ff/fi/fl/ffi/ffl/dotlessi/dotlessj/grave/hungarumlaut
/dotaccent/breve/caron/ring/ogonek/quotedblleft/quotedblright/oe/lslash
/quotedblbase/OE/Lslash/.notdef/exclamdown/cent/sterling/currency/yen
/brokenbar/section/dieresis/copyright/ordfeminine/guilsinglleft
/logicalnot/minus/registered/macron/degree/plusminus/twosuperior
/threesuperior/acute/mu/paragraph/periodcentered/cedilla/onesuperior
/ordmasculine/guilsinglright/onequarter/onehalf/threequarters
/questiondown/Agrave/Aacute/Acircumflex/Atilde/Adieresis/Aring/AE
/Ccedilla/Egrave/Eacute/Ecircumflex/Edieresis/Igrave/Iacute/Icircumflex
/Idieresis/Eth/Ntilde/Ograve/Oacute/Ocircumflex/Otilde/Odieresis
/multiply/Oslash/Ugrave/Uacute/Ucircumflex/Udieresis/Yacute/Thorn
/germandbls/agrave/aacute/acircumflex/atilde/adieresis/aring/ae/ccedilla
/egrave/eacute/ecircumflex/edieresis/igrave/iacute/icircumflex/idieresis
/eth/ntilde/ograve/oacute/ocircumflex/otilde/odieresis/divide/oslash
/ugrave/uacute/ucircumflex/udieresis/yacute/thorn/ydieresis]def
/Courier@0 ENC0/Courier RE/Times-Roman@0 ENC0/Times-Roman RE
/Times-Bold@0 ENC0/Times-Bold RE/Courier-Bold@0 ENC0/Courier-Bold RE
%%EndProlog
%%Page: 1 1
%%BeginPageSetup
BP
%%EndPageSetup
/F0 10/Courier-Bold@0 SF(Internet Engineering Task Force)72 85 Q
(Sally Floyd)179.998 E 317.998(INTERNET-DRAFT ICIR)72 98 R 167.998
(draft-ietf-dccp-tfrc-voip-02.ps Eddie)72 111 R(Kohler)6 E
(Expires: January 2006)72 124 Q(UCLA)281.998 E(19 July 2005)431.998 137
Q/F1 14/Times-Bold@0 SF(TCP Friendly Rate Contr)149.806 162 Q
(ol \(TFRC\) f)-.252 E(or V)-.35 E(oice:)-1.4 E -1.4(Vo)248.709 175 S
(IP V)1.4 E(ariant)-1.288 E/F2 11/Times-Bold@0 SF(Status of this Memo)72
220 Q/F3 11/Times-Roman@0 SF
(This document is an Internet-Draft and is subject to all pro)100.8
236.6 Q(visions of section 3 of RFC)-.165 E 2.75(3667. By)100.8 249.6 R
(submitting this Internet-Draft, each author represents that an)2.75 E
2.75(ya)-.165 G(pplicable patent)-2.75 E
(or other IPR claims of which he or she is a)100.8 262.6 Q -.11(wa)-.165
G(re ha).11 E .33 -.165(ve b)-.22 H(een or will be disclosed, and an)
.165 E 2.75(yo)-.165 G(f)-2.75 E(which he or she becomes a)100.8 275.6 Q
-.11(wa)-.165 G
(re will be disclosed, in accordance with Section 6 of BCP 79.).11 E
(Internet-Drafts are w)100.8 292.2 Q
(orking documents of the Internet Engineering T)-.11 E(ask F)-.88 E
(orce \(IETF\), its)-.165 E(areas, and its w)100.8 305.2 Q
(orking groups.)-.11 E(Note that other groups may also distrib)5.5 E
(ute w)-.22 E(orking)-.11 E(documents as Internet-Drafts.)100.8 318.2 Q
(Internet-Drafts are draft documents v)100.8 334.8 Q
(alid for a maximum of six months and may be)-.275 E
(updated, replaced, or obsoleted by other documents at an)100.8 347.8 Q
2.75(yt)-.165 G 2.75(ime. It)-2.75 F(is inappropriate to use)2.75 E
(Internet-Drafts as reference material or to cite them other than as "w)
100.8 360.8 Q(ork in progress.")-.11 E
(The list of current Internet-Drafts can be accessed at http://www)100.8
377.4 Q(.ietf.or)-.715 E(g/ietf/1id-)-.198 E(abstracts.txt.)100.8 390.4
Q(The list of Internet-Draft Shado)100.8 407 Q 2.75(wD)-.275 G
(irectories can be accessed at)-2.75 E(http://www)100.8 420 Q(.ietf.or)
-.715 E(g/shado)-.198 E -.715(w.)-.275 G(html.).715 E
(This Internet-Draft will e)100.8 436.6 Q(xpire on January 2006.)-.165 E
F2(Copyright Notice)72 462.6 Q F3(Cop)100.8 479.2 Q
(yright \(C\) The Internet Society \(2005\). All Rights Reserv)-.11 E
(ed.)-.165 E F2(Abstract)72 505.2 Q F3(TCP-Friendly Rate Control \(TFRC\
\) is a congestion control mechanism for unicast \215o)100.8 521.8 Q(ws)
-.275 E(operating in a best-ef)100.8 534.8 Q(fort Internet en)-.275 E
(vironment [RFC 3448]. This document proposes a)-.44 E -1.419(Vo)100.8
547.8 S(IP v)1.419 E(ariant to TFRC.)-.275 E(TFRC w)5.5 E
(as intended for applications that use a \214x)-.11 E(ed pack)-.165 E
(et size,)-.11 E(and w)100.8 560.8 Q(as designed to be reasonably f)-.11
E(air when competing for bandwidth with TCP)-.11 E
(connections using the same pack)100.8 573.8 Q(et size.)-.11 E(The V)5.5
E(oIP v)-1.419 E(ariant of TFRC is designed for)-.275 E
(applications that send small pack)100.8 586.8 Q
(ets, where the design goal is to achie)-.11 E .33 -.165(ve t)-.275 H
(he same).165 E(bandwidth in bps as a TCP \215o)100.8 599.8 Q 2.75(wu)
-.275 G(sing 1500-byte data pack)-2.75 E 2.75(ets. This)-.11 F -.275(va)
2.75 G(riant is referred to in).275 E
(RFC 3448 as TFRC-PS, for applications that might v)100.8 612.8 Q
(ary their pack)-.275 E(et size in response to)-.11 E 2.75
(congestion. The)100.8 625.8 R -1.419(Vo)2.75 G(IP v)1.419 E
(ariant of TFRC enforces a Min Interv)-.275 E(al of 10 ms between data)
-.275 E(pack)100.8 638.8 Q(ets, to pre)-.11 E -.165(ve)-.275 G
(nt a single \215o).165 E 2.75(wf)-.275 G(rom sending small pack)-2.75 E
(ets arbitrarily frequently)-.11 E(.)-.715 E(Flo)100.8 664.8 Q
(ws using the V)-.275 E(oIP v)-1.419 E
(ariant of TFRC compete reasonably f)-.275 E(airly with lar)-.11 E
(ge-pack)-.198 E(et TCP)-.11 E(and TFRC \215o)100.8 677.8 Q(ws in en)
-.275 E(vironments where lar)-.44 E(ge-pack)-.198 E(et \215o)-.11 E
(ws and small-pack)-.275 E(et \215o)-.11 E(ws)-.275 E -.165(ex)100.8
690.8 S(perience similar pack).165 E(et drop rates.)-.11 E(Ho)5.5 E(we)
-.275 E -.165(ve)-.275 G .88 -.44(r, i).165 H 2.75(ne).44 G -.44(nv)
-2.75 G(ironments where small-pack).44 E(et \215o)-.11 E(ws)-.275 E
-.165(ex)100.8 703.8 S(perience lo).165 E(wer pack)-.275 E
(et drop rates than lar)-.11 E(ge-pack)-.198 E(et \215o)-.11 E
(ws \(e.g., with DropT)-.275 E(ail queues in)-.88 E
(units of bytes\), the current V)100.8 716.8 Q(oIP v)-1.419 E
(ariant of TFRC can recei)-.275 E .33 -.165(ve c)-.275 H
(onsiderably more than its).165 E(Flo)72 769 Q(yd/K)-.11 E 333.671
(ohler [P)-.385 F(age 1])-.165 E 0 Cg EP
%%Page: 2 2
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E(share of the bandwidth.)100.8
85 Q(\(W)5.5 E 2.75(en)-.88 G(ote ho)-2.75 E(we)-.275 E -.165(ve)-.275 G
2.75(rt).165 G(hat in all scenarios the V)-2.75 E(oIP v)-1.419 E
(ariant of TFRC is)-.275 E(better)100.8 98 Q 2.75(,i)-.44 G 2.75(nt)
-2.75 G(erms of congestion in the netw)-2.75 E
(ork, than the same application in the absence of)-.11 E
(congestion control\).)100.8 111 Q(Flo)72 769 Q(yd/K)-.11 E 333.671
(ohler [P)-.385 F(age 2])-.165 E 0 Cg EP
%%Page: 3 3
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E .396 -.198(TO B)100.8 85 T 2.75
(ED).198 G(ELETED BY THE RFC EDIT)-2.75 E(OR UPON PUBLICA)-.198 E(TION:)
-1.221 E(Changes from draft-ietf-dccp-tfrc-v)103.55 98 Q(oip-01.txt)-.22
E 2.75(*A)106.3 111 S
(dded modi\214ed algorithm for calculating the loss e)-2.75 E -.165(ve)
-.275 G(nt rate,).165 E(for short interv)117.3 124 Q
(als with multiple pack)-.275 E(et drops.)-.11 E 2.75(*M)106.3 137 S
-.165(ove)-2.75 G 2.75(dF).165 G(aster Restart to a separate document.)
-2.915 E 2.75(*A)106.3 150 S
(dded simulations with a con\214gured byte drop rate.)-2.75 E 2.75(*A)
106.3 163 S(dded man)-2.75 E 2.75(ym)-.165 G
(ore simulations, including DropT)-2.75 E(ail with a queue)-.88 E
(in bytes.)111.8 176 Q 2.75(*A)106.3 189 S(dded a discussion of unf)
-2.75 E(airness for DropT)-.11 E(ail with a queue)-.88 E(in bytes.)111.8
202 Q(Changes from draft-ietf-dccp-tfrc-v)103.55 215 Q(oip-00.txt)-.22 E
2.75(*A)106.3 228 S(dded more simulations.)-2.75 E 2.75(*A)106.3 241 S
(dded a Related W)-2.75 E(ork section.)-.88 E(Flo)72 769 Q(yd/K)-.11 E
333.671(ohler [P)-.385 F(age 3])-.165 E 0 Cg EP
%%Page: 4 4
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E/F1 13/Times-Bold@0 SF -1.196
(Ta)253.526 85 S(ble of Contents)1.196 E/F2 10/Times-Roman@0 SF(1. Con)
100.8 110 Q -.15(ve)-.4 G(ntions).15 E F0 11(.......................)
10.44 G F2(5)11.5 E(2. V)100.8 122 Q(oIP V)-1.29 E(ariant Introduction)
-1.11 E F0 11(...................)12.31 G F2(5)11.5 E(3. V)100.8 134 Q
(oIP V)-1.29 E(ariant Congestion Control)-1.11 E F0 11
(.................)10.63 G F2(6)11.5 E(4. V)100.8 146 Q(oIP V)-1.29 E
(ariant Discussion)-1.11 E F0 11(....................)4.11 G F2(7)11.5 E
(4.1. The TCP Throughput Equation)115.8 158 Q F0 11(................)
9.48 G F2(7)11.5 E(4.2. Accounting for Header Size)115.8 170 Q F0 11
(.................)8.54 G F2(7)11.5 E(4.3. The V)115.8 182 Q
(oIP Min Interv)-1.29 E(al)-.25 E F0 11(...................)2.56 G F2(8)
11.5 E(4.4. Counting P)115.8 194 Q(ack)-.15 E(et Losses)-.1 E F0 11
(..................)10.84 G F2(9)11.5 E(5. A Comparison with RFC 3714)
100.8 206 Q F0 11(..................)6.41 G F2(9)11.5 E(6. The V)100.8
218 Q(oIP V)-1.29 E(ariant with Applications that Modify the P)-1.11 E
(ack)-.15 E(et Size)-.1 E F0 11(........)10.21 G F2(9)11.5 E
(7. Simulation Results)100.8 230 Q F0 11(.....................)12.1 G F2
(9)11.5 E(7.1. Simulations with Con\214gured Drop Rates)115.8 242 Q F0
11(.............)12.11 G F2(9)11.5 E(7.2. P)115.8 254 Q(ack)-.15 E
(et Dropping Beha)-.1 E(vior at Routers with DropT)-.2 E(ail Queues)-.8
E F0 11(........)3.54 G F2(14)6.5 E(7.3. P)115.8 266 Q(ack)-.15 E
(et Dropping Beha)-.1 E(vior at Routers with A)-.2 E(QM)-.55 E F0 11
(...........)7.86 G F2(15)6.5 E(8. Discussion)100.8 278 Q F0 11
(........................)2.81 G F2(18)6.5 E(9. Implementation Issues)
100.8 290 Q F0 11(....................)11.43 G F2(18)6.5 E
(10. Security Considerations)100.8 302 Q F0 11(....................).87
G F2(18)6.5 E(11. IAN)100.8 314 Q 2.5(AC)-.35 G(onsiderations)-2.5 E F0
11(....................)9.56 G F2(18)6.5 E(12. Thanks)100.8 326 Q F0 11
(........................)12.26 G F2(19)6.5 E(Normati)100.8 338 Q .3
-.15(ve R)-.25 H(eferences).15 E F0 11(.....................)8.66 G F2
(19)6.5 E(Informati)100.8 350 Q .3 -.15(ve R)-.25 H(eferences).15 E F0
11(.....................)4.22 G F2(19)6.5 E(Authors' Addresses)100.8 362
Q F0 11(......................)5.05 G F2(19)6.5 E(13. Related W)100.8
374 Q(ork on V)-.8 E(oIP V)-1.29 E(ariants of TFRC)-1.11 E F0 11
(...............)2 G F2(20)6.5 E
(14. Simulation Results with RED Queue Management)100.8 386 Q F0 11
(............)4.76 G F2(21)6.5 E(15. A Discussion of P)100.8 398 Q(ack)
-.15 E(et Size and P)-.1 E(ack)-.15 E(et Dropping)-.1 E F0 11
(............)6.95 G F2(22)6.5 E(Full Cop)100.8 410 Q(yright Statement)
-.1 E F0 11(....................)10.12 G F2(22)6.5 E
(Intellectual Property)100.8 422 Q F0 11(......................)2.27 G
F2(23)6.5 E F0(Flo)72 769 Q(yd/K)-.11 E 333.671(ohler [P)-.385 F(age 4])
-.165 E 0 Cg EP
%%Page: 5 5
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E/F1 11/Times-Bold@0 SF(1.)72 85
Q/F2 14/Times-Bold@0 SF(Con)5.5 E -.14(ve)-.56 G(ntions).14 E F0(The k)
100.8 101.6 Q .33 -.165(ey w)-.11 H(ords "MUST", "MUST NO).055 E
(T", "REQ)-.44 E(UIRED", "SHALL", "SHALL NO)-.11 E(T",)-.44 E
("SHOULD", "SHOULD NO)100.8 114.6 Q(T", "RECOMMENDED", "MA)-.44 E
(Y", and "OPTION)-1.155 E(AL" in this)-.385 E
(document are to be interpreted as described in [RFC 2119].)100.8 127.6
Q F1(2.)72 153.6 Q F2 -1.4(Vo)5.5 G(IP V)1.4 E(ariant Intr)-1.288 E
(oduction)-.252 E F0(This document speci\214es a V)100.8 170.2 Q(oIP v)
-1.419 E(ariant for TCP-friendly rate control \(TFRC\) [RFC 3448].)-.275
E(TFRC w)100.8 183.2 Q(as designed to be reasonably f)-.11 E
(air when competing for bandwidth with TCP \215o)-.11 E(ws,)-.275 E -.22
(bu)100.8 196.2 S 2.75(tt).22 G 2.75(oa)-2.75 G -.22(vo)-2.97 G
(id the abrupt changes in the sending rate characteristic of TCP').22 E
2.75(sc)-.605 G(ongestion)-2.75 E(control mechanisms.)100.8 209.2 Q
(TFRC is intended for applications such as streaming media)5.5 E
(applications where a relati)100.8 222.2 Q -.165(ve)-.275 G
(ly smooth sending rate is of importance.).165 E(The V)100.8 238.8 Q
(oIP v)-1.419 E(ariant is intended for \215o)-.275 E
(ws that need to send frequent small pack)-.275 E(ets \(limited by)-.11
E 2.75(am)100.8 251.8 S(inimum interv)-2.75 E(al between pack)-.275 E
(ets of 10 ms\).)-.11 E(Con)5.5 E -.165(ve)-.44 G
(ntional TFRC measures loss rates).165 E(by estimating the loss e)100.8
264.8 Q -.165(ve)-.275 G
(nt ratio as described in [RFC 3448], and uses this loss e).165 E -.165
(ve)-.275 G(nt rate).165 E(to determine the sending rate in pack)100.8
277.8 Q(ets per round-trip time.)-.11 E(This has consequences for the)
5.5 E(rate a TFRC \215o)100.8 290.8 Q 2.75(wc)-.275 G(an achie)-2.75 E
.33 -.165(ve w)-.275 H(hen sharing a bottleneck with lar).165 E(ge-pack)
-.198 E(et TCP \215o)-.11 E 2.75(ws. In)-.275 F(particular)100.8 303.8 Q
2.75(,al)-.44 G -.275(ow)-2.75 G(-bandwidth, small-pack).275 E
(et TFRC \215o)-.11 E 2.75(ws)-.275 G(haring a bottleneck with high-)
-2.75 E(bandwidth, lar)100.8 316.8 Q(ge-pack)-.198 E(et TCP \215o)-.11 E
(ws may be forced to slo)-.275 E 2.75(wd)-.275 G -.275(ow)-2.75 G(n, e)
.275 E -.165(ve)-.275 G 2.75(nt).165 G(hough the)-2.75 E(application')
100.8 329.8 Q 2.75(sn)-.605 G
(ominal rate in bytes per second is less than the rate achie)-2.75 E
-.165(ve)-.275 G 2.75(db).165 G 2.75(yt)-2.75 G(he TCP)-2.75 E<8d6f>
100.8 342.8 Q 2.75(ws. Intuiti)-.275 F -.165(ve)-.275 G(ly).165 E 2.75
(,t)-.715 G(his w)-2.75 E(ould be "f)-.11 E(air" only if the netw)-.11 E
(ork limitation w)-.11 E(as in pack)-.11 E(ets per)-.11 E(second \(such\
 as a routing lookup\), rather than bytes per second \(such as link ban\
dwidth\).)100.8 355.8 Q(Con)100.8 368.8 Q -.165(ve)-.44 G
(ntional wisdom is that man).165 E 2.75(yo)-.165 G 2.75(ft)-2.75 G
(he netw)-2.75 E(ork limitations in today')-.11 E 2.75(sI)-.605 G
(nternet are in)-2.75 E(bytes per second, e)100.8 381.8 Q -.165(ve)-.275
G 2.75(nt).165 G(hough the netw)-2.75 E
(ork limitations of the future might mo)-.11 E .33 -.165(ve b)-.165 H
(ack).165 E(to)100.8 394.8 Q -.11(wa)-.275 G(rds limitations in pack).11
E(ets per second.)-.11 E(The V)100.8 411.4 Q(oIP v)-1.419 E(ariant of T\
FRC described here will better support applications that do not w)-.275
E(ant)-.11 E(their sending rates in bytes per second to suf)100.8 424.4
Q(fer from their use of small pack)-.275 E 2.75(ets. This)-.11 F -.275
(va)100.8 437.4 S(riant is restricted to applications that send pack)
.275 E(ets no more than once e)-.11 E -.165(ve)-.275 G(ry 10 ms \(the)
.165 E(Min Interv)100.8 450.4 Q 2.75(al\). Gi)-.275 F -.165(ve)-.275 G
2.75(nt).165 G(his restriction, the V)-2.75 E(oIP v)-1.419 E(ariant ef)
-.275 E(fecti)-.275 E -.165(ve)-.275 G(ly calculates the TFRC f).165 E
(air)-.11 E(rate as if the bottleneck restriction w)100.8 463.4 Q
(as in bytes per second.)-.11 E(Applications using the V)5.5 E(oIP)
-1.419 E -.275(va)100.8 476.4 S(riant of TFRC could ha).275 E .33 -.165
(ve a \214)-.22 H -.165(xe).165 G 2.75(dp).165 G(ack)-2.75 E
(et size, or could v)-.11 E(ary their pack)-.275 E(et size in response)
-.11 E(to congestion.)100.8 489.4 Q(The V)100.8 506 Q(oIP v)-1.419 E
(ariant of TFRC is moti)-.275 E -.275(va)-.275 G
(ted in part by the approach in RFC 3714, which ar).275 E(gues)-.198 E
(that it is acceptable for V)100.8 519 Q(oIP \215o)-1.419 E
(ws to assume that the netw)-.275 E(ork limitation is in bytes per)-.11
E(second \(Bps\) rather in pack)100.8 532 Q
(ets per second \(pps\), and to ha)-.11 E .33 -.165(ve t)-.22 H
(he same sending rate in bytes).165 E(per second as a TCP \215o)100.8
545 Q 2.75(ww)-.275 G(ith 1500-byte pack)-2.75 E(ets and the same pack)
-.11 E(et drop rate.)-.11 E(RFC)5.5 E(3714 states the follo)100.8 558 Q
(wing:)-.275 E("While the ideal w)129.6 574.6 Q(ould be to ha)-.11 E .33
-.165(ve a t)-.22 H(ransport protocol that is able to detect whether)
.165 E(the bottleneck links along the path are limited in Bps or in pps\
, and to respond)129.6 587.6 Q(appropriately when the limitation is in \
pps, such an ideal is hard to achie)129.6 600.6 Q -.165(ve)-.275 G 2.75
(.W).165 G(e)-3.63 E -.11(wo)129.6 613.6 S(uld not w).11 E
(ant to delay the deplo)-.11 E(yment of congestion control for telephon)
-.11 E 2.75(yt)-.165 G(raf)-2.75 E<8c63>-.275 E
(until such an ideal could be accomplished.)129.6 626.6 Q
(In addition, we note that the current TCP)5.5 E
(congestion control mechanisms are themselv)129.6 639.6 Q(es not v)-.165
E(ery ef)-.165 E(fecti)-.275 E .33 -.165(ve i)-.275 H 2.75(na).165 G
2.75(ne)-2.75 G -.44(nv)-2.75 G(ironment).44 E
(where there is a limitation along the re)129.6 652.6 Q -.165(ve)-.275 G
(rse path in pps.).165 E(While the TCP)5.5 E(mechanisms do pro)129.6
665.6 Q(vide an incenti)-.165 E .33 -.165(ve t)-.275 H 2.75(ou).165 G
(se lar)-2.75 E(ge data pack)-.198 E(ets, TCP does not include)-.11 E
(an)129.6 678.6 Q 2.75(ye)-.165 G -.275(ff)-2.75 G(ecti).275 E .33 -.165
(ve c)-.275 H(ongestion control mechanisms for the stream of small).165
E(ackno)129.6 691.6 Q(wledgement pack)-.275 E(ets on the re)-.11 E -.165
(ve)-.275 G(rse path.).165 E(Gi)5.5 E -.165(ve)-.275 G 2.75(nt).165 G
(he ar)-2.75 E(guments abo)-.198 E -.165(ve)-.165 G 2.75(,i).165 G 2.75
(ts)-2.75 G(eems)-2.75 E(acceptable to us to assume a netw)129.6 704.6 Q
(ork limitation in Bps rather than in pps in)-.11 E
(considering the minimum sending rate of telephon)129.6 717.6 Q 2.75(yt)
-.165 G(raf)-2.75 E(\214c.")-.275 E(Flo)72 769 Q(yd/K)-.11 E 284.171
(ohler Section)-.385 F 2.75(2. [P)2.75 F(age 5])-.165 E 0 Cg EP
%%Page: 6 6
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E -.385(Tr)100.8 85 S(anslating \
the discussion in [RFC 3714] to the congestion control mechanisms of TF\
RC,).385 E(it seems acceptable to standardize a v)100.8 98 Q
(ariant of TFRC that allo)-.275 E(ws V)-.275 E(oIP \215o)-1.419 E
(ws sending small)-.275 E(pack)100.8 111 Q(ets to achie)-.11 E .33 -.165
(ve a r)-.275 H(ough f).165 E(airness with TCP \215o)-.11 E
(ws in terms of the sending rate in Bps,)-.275 E
(rather than in terms of the sending rate in pps.)100.8 124 Q
(This is accomplished by a small)5.5 E
(modi\214cation to TFRC, as described belo)100.8 137 Q -.715(w.)-.275 G
(Ho)100.8 153.6 Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, b).165 H
(ecause the bottlenecks in the netw).44 E(ork in f)-.11 E
(act can include limitations in pps as)-.11 E(well as in Bps, we pay sp\
ecial attention to the potential dangers of encouraging a lar)100.8
166.6 Q(ge)-.198 E(deplo)100.8 179.6 Q(yment of best-ef)-.11 E
(fort traf)-.275 E
(\214c in the Internet consisting entirely of small pack)-.275 E 2.75
(ets. This)-.11 F(is)2.75 E
(discussed in more detail in Section 4.3. In addition, as ag)100.8 192.6
Q(ain discussed in Section 4.3, the)-.055 E -1.419(Vo)100.8 205.6 S
(IP v)1.419 E(ariant of TFRC includes the limitation of the Min Interv)
-.275 E(al between pack)-.275 E(ets of 10 ms.)-.11 E(The V)100.8 222.2 Q
(oIP v)-1.419 E(ariant of TFRC essentially assumes that the small-pack)
-.275 E(et V)-.11 E(oIP TFRC \215o)-1.419 E(w)-.275 E(recei)100.8 235.2
Q -.165(ve)-.275 G 2.75(sr).165 G(oughly the same pack)-2.75 E
(et drop rate as a lar)-.11 E(ge-pack)-.198 E(et TFRC or TCP \215o)-.11
E 4.18 -.715(w. A)-.275 H 2.75(sw).715 G(e)-2.75 E(sho)100.8 248.2 Q
1.43 -.715(w, t)-.275 H(his assumption is not correct for all of the si\
mulations in this document.).715 E(The V)100.8 264.8 Q(oIP v)-1.419 E
(ariant of TFRC requires a modi\214cation in TFRC')-.275 E 2.75(sc)-.605
G(alculation of the loss e)-2.75 E -.165(ve)-.275 G(nt).165 E
(rate, because a V)100.8 277.8 Q(oIP TFRC connection can send man)-1.419
E 2.75(ys)-.165 G(mall pack)-2.75 E(ets when a standard)-.11 E
(TFRC or TCP connection w)100.8 290.8 Q(ould send a single lar)-.11 E
(ge pack)-.198 E 2.75(et. It)-.11 F(is not possible for a standard)2.75
E(TFRC or TCP connection to repeatedly send multiple pack)100.8 303.8 Q
(ets per round-trip time in the)-.11 E -.11(fa)100.8 316.8 S
(ce of a high pack).11 E(et drop rate.)-.11 E
(As a result, TCP and standard TFRC only respond to a)5.5 E
(single loss e)100.8 329.8 Q -.165(ve)-.275 G(nt per round-trip time, a\
nd are still able to detect and respond to increasingly).165 E(hea)100.8
342.8 Q(vy pack)-.22 E(et loss rates.)-.11 E(Ho)5.5 E(we)-.275 E -.165
(ve)-.275 G .88 -.44(r, i).165 H 2.75(nah).44 G(ighly-congested en)-2.75
E(vironment, when a TCP)-.44 E(connection might be sending, on a)100.8
355.8 Q -.165(ve)-.22 G(rage, one lar).165 E(ge pack)-.198 E
(et per round-trip time, a)-.11 E(corresponding V)100.8 368.8 Q
(oIP TFRC connection might be sending man)-1.419 E 2.75(ys)-.165 G
(mall pack)-2.75 E(ets per round-trip)-.11 E 2.75(time. As)100.8 381.8 R
2.75(ar)2.75 G(esult, in order to maintain f)-2.75 E(airness with TCP)
-.11 E 2.75(,a)-1.221 G(nd to be able to detect changes)-2.75 E
(in the de)100.8 394.8 Q(gree of congestion, V)-.165 E
(oIP TFRC needs to be sensiti)-1.419 E .33 -.165(ve t)-.275 H 2.75(ot)
.165 G(he actual pack)-2.75 E(et drop rate)-.11 E
(during periods of high congestion.)100.8 407.8 Q
(This is discussed in more detail in the section belo)5.5 E -.715(w.)
-.275 G/F1 11/Times-Bold@0 SF(3.)72 433.8 Q/F2 14/Times-Bold@0 SF -1.4
(Vo)5.5 G(IP V)1.4 E(ariant Congestion Contr)-1.288 E(ol)-.252 E F0
(TFRC uses the TCP throughput equation gi)100.8 450.4 Q -.165(ve)-.275 G
2.75(ni).165 G 2.75(nS)-2.75 G(ection 3.1 of [RFC 3448], which gi)-2.75
E -.165(ve)-.275 G(s).165 E(the allo)100.8 463.4 Q
(wed sending rate X in bytes per second as a function of the loss e)
-.275 E -.165(ve)-.275 G(nt rate, pack).165 E(et)-.11 E
(size, and round-trip time.)100.8 476.4 Q
([RFC 3448] speci\214es that the pack)5.5 E(et size s used in the)-.11 E
(throughput equation should be the pack)100.8 489.4 Q
(et size used by the application, or the estimated)-.11 E(mean pack)
100.8 502.4 Q(et size if there are v)-.11 E(ariations in the pack)-.275
E(et size depending on the data.)-.11 E(This gi)5.5 E -.165(ve)-.275 G
(s).165 E(rough f)100.8 515.4 Q(airness with TCP \215o)-.11 E
(ws using the same pack)-.275 E(et size.)-.11 E(The V)100.8 532 Q(oIP v)
-1.419 E(ariant changes this beha)-.275 E(vior in the follo)-.22 E
(wing w)-.275 E(ays.)-.11 E 12.65<8354>100.8 548.6 S(he nominal pack)
-12.65 E(et size: The nominal pack)-.11 E
(et size s is set to 1460 bytes.)-.11 E -.165(Fo)5.5 G(llo).165 E(wing)
-.275 E([RFC 3714], this pro)117.3 561.6 Q(vides a goal of f)-.165 E
(airness, in terms of the sending rate in bytes per)-.11 E
(second, with a TCP \215o)117.3 574.6 Q 2.75(ww)-.275 G
(ith 1460 bytes of application data per pack)-2.75 E(et b)-.11 E
(ut with the)-.22 E(same pack)117.3 587.6 Q(et drop rate.)-.11 E 12.65
<8354>100.8 604.2 S(aking pack)-13.53 E
(et headers into account: The allo)-.11 E
(wed transmit rate X in bytes per second is)-.275 E(reduced by a f)117.3
617.2 Q(actor that accounts for pack)-.11 E(et header size.)-.11 E
(This gi)5.5 E -.165(ve)-.275 G 2.75(st).165 G(he application)-2.75 E
(some incenti)117.3 630.2 Q -.165(ve)-.275 G 2.75(,b).165 G -.165(ey)
-2.75 G(ond the Min Interv).165 E
(al, not to use unnecessarily small pack)-.275 E 2.75(ets. In)-.11 F
(particular)117.3 643.2 Q 2.75(,w)-.44 G 2.75(ei)-2.75 G(ntroduce a ne)
-2.75 E 2.75(wp)-.275 G(arameter H, which represents the e)-2.75 E
(xpected size in bytes)-.165 E(of netw)117.3 656.2 Q
(ork and transport headers to be used on the TFRC connection')-.11 E
2.75(sp)-.605 G(ack)-2.75 E 2.75(ets. This)-.11 F(is)2.75 E
(used to reduce the allo)117.3 669.2 Q(wed transmit rate X as follo)
-.275 E(ws:)-.275 E 2.75(X<)117.3 685.8 S 2.75(-X*s)-2.75 G
(_true / \(s_true + H\),)-2.75 E(where s_true is the true a)117.3 702.4
Q -.165(ve)-.22 G(rage data pack).165 E
(et size for the connection in bytes, e)-.11 E(xcluding)-.165 E
(the transport and netw)117.3 715.4 Q(ork headers.)-.11 E(Flo)72 769 Q
(yd/K)-.11 E 284.171(ohler Section)-.385 F 2.75(3. [P)2.75 F(age 6])
-.165 E 0 Cg EP
%%Page: 7 7
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E
(The H parameter is set to the constant 40 bytes.)117.3 85 Q
(Thus, if the V)5.5 E(oIP TFRC application)-1.419 E
(used 40-byte data se)117.3 98 Q(gments, the allo)-.165 E
(wed transmit rate X w)-.275 E(ould be halv)-.11 E(ed to account for)
-.165 E(the f)117.3 111 Q(act that half of the sending rate w)-.11 E
(ould be used by the headers.)-.11 E(Section 4.2 justi\214es)5.5 E
(this de\214nition.)117.3 124 Q(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88
-.44(r, a c).165 H(onnection using the V).44 E(oIP v)-1.419 E(ariant MA)
-.275 E 2.75(Yi)-1.155 G(nstead use a more)-2.75 E
(precise estimate of H, based on the actual netw)117.3 137 Q
(ork and transport headers to be used on)-.11 E(the connection')117.3
150 Q 2.75(sp)-.605 G(ack)-2.75 E 2.75(ets. F)-.11 F(or e)-.165 E
(xample, a DCCP connection [DCCP] o)-.165 E -.165(ve)-.165 G 2.75(rI)
.165 G(Pv4, where)-2.75 E(data pack)117.3 163 Q
(ets use the DCCP-Data pack)-.11 E
(et type, and there are no IP or DCCP options,)-.11 E
(could set H to 20 + 12 = 32 bytes.)117.3 176 Q 12.65<834d>100.8 192.6 S
(easuring the loss e)-12.65 E -.165(ve)-.275 G
(nt rate in times of high loss: During short loss interv).165 E
(als \(those at)-.275 E(most tw)117.3 205.6 Q 2.75(or)-.11 G(ound-trip \
times\), the loss rate is computed by counting the actual number of)
-2.75 E(pack)117.3 218.6 Q(ets lost or mark)-.11 E
(ed, not by counting at most one loss e)-.11 E -.165(ve)-.275 G
(nt per loss interv).165 E(al.)-.275 E -.44(Wi)117.3 231.6 S
(thout this change, V).44 E(oIP TFRC could send multiple pack)-1.419 E
(ets per round-trip time e)-.11 E -.165(ve)-.275 G(n).165 E(in the f)
117.3 244.6 Q(ace of hea)-.11 E(vy congestion, for a steady-state beha)
-.22 E(vior of multiple pack)-.22 E(ets dropped)-.11 E
(each round-trip time.)117.3 257.6 Q(In standard TFRC, the TFRC recei)
117.3 283.6 Q -.165(ve)-.275 G 2.75(re).165 G(stimates the loss e)-2.75
E -.165(ve)-.275 G(nt rate by calculating the).165 E -2.475 -.22(av e)
117.3 296.6 T(rage loss interv).22 E(al in pack)-.275 E(ets, and in)-.11
E -.165(ve)-.44 G(rting to get the loss e).165 E -.165(ve)-.275 G
(nt rate.).165 E(Thus, for a)5.5 E(short loss interv)117.3 309.6 Q
(al with N pack)-.275 E
(ets and K losses, standard TFRC calculates the size of)-.11 E
(that loss interv)117.3 322.6 Q(al as N pack)-.275 E(ets, contrib)-.11 E
(uting to a loss e)-.22 E -.165(ve)-.275 G(nt rate of 1/N.).165 E(Ho)5.5
E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, f).165 H(or).44 E -1.419(Vo)
117.3 335.6 S(IP TFRC, for small loss interv)1.419 E(als of at most tw)
-.275 E 2.75(or)-.11 G(ound-trip times, if the loss interv)-2.75 E(al)
-.275 E(consists of N pack)117.3 348.6 Q
(ets including K losses, the size of the loss interv)-.11 E
(al is calculated as)-.275 E(N/K, contrib)117.3 361.6 Q
(uting to a loss e)-.22 E -.165(ve)-.275 G
(nt rate of K/N instead of 1/N.).165 E 12.65<8341>100.8 378.2 S
(minimum interv)-9.9 E(al between pack)-.275 E(ets: The V)-.11 E(oIP v)
-1.419 E(ariant of TFRC enforces a Min)-.275 E(Interv)117.3 391.2 Q
(al between pack)-.275 E(ets of 10 ms.)-.11 E 2.75<418d>5.5 G .55 -.275
(ow t)-2.75 H(hat wishes its transport protocol to e).275 E(xceed)-.165
E(this Min Interv)117.3 404.2 Q(al MUST use the con)-.275 E -.165(ve)
-.44 G(ntional TFRC equations, rather than the V).165 E(oIP)-1.419 E
-.275(va)117.3 417.2 S 2.75(riant. The).275 F(moti)2.75 E -.275(va)-.275
G(tion for this is discussed belo).275 E -.715(w.)-.275 G/F1 11
/Times-Bold@0 SF(4.)72 443.2 Q/F2 14/Times-Bold@0 SF -1.4(Vo)5.5 G(IP V)
1.4 E(ariant Discussion)-1.288 E F1(4.1.)72 469.2 Q/F3 13/Times-Bold@0
SF(The TCP Thr)5.5 E(oughput Equation)-.234 E F0(The V)100.8 485.8 Q
(oIP v)-1.419 E(ariant of TFRC uses the TCP throughput equation gi)-.275
E -.165(ve)-.275 G 2.75(ni).165 G 2.75(n[)-2.75 G(RFC 3448].)-2.75 E(As)
5.5 E(sho)100.8 498.8 Q(wn in T)-.275 E
(able 1 of [RFC 3714], for high pack)-.88 E
(et drop rates, this throughput equation gi)-.11 E -.165(ve)-.275 G(s)
.165 E(rough f)100.8 511.8 Q(airness with most aggressi)-.11 E .33 -.165
(ve p)-.275 H(ossible current TCP: a SA).165 E(CK TCP \215o)-.44 E 2.75
(wu)-.275 G(sing)-2.75 E(timestamps and ECN.)100.8 524.8 Q
(Because it is not recommended for routers to use ECN-marking in)5.5 E
(highly-congested en)100.8 537.8 Q(vironments \(e.g., with pack)-.44 E
(et drop rates greater than 10%\), we note)-.11 E(that it w)100.8 550.8
Q(ould be useful to ha)-.11 E .33 -.165(ve a t)-.22 H
(hroughput equation with a some).165 E(what more moderate)-.275 E
(sending rate for pack)100.8 563.8 Q(et drop rates of 40% and abo)-.11 E
-.165(ve)-.165 G(.).165 E F1(4.2.)72 589.8 Q F3(Accounting f)5.5 E
(or Header Size)-.325 E F0([RFC 3714] mak)100.8 606.4 Q
(es the optimistic assumption that the limitation of the netw)-.11 E
(ork is in)-.11 E
(bandwidth in bytes per second \(Bps\), and not in CPU c)100.8 619.4 Q
(ycles or in pack)-.165 E(ets per second \(pps\).)-.11 E(Ho)100.8 632.4
Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, s).165 H(ome attention must b\
e paid to the load in pps as well as to the load in Bps.).44 E(Ev)100.8
645.4 Q(en aside from the Min Interv)-.165 E(al, the V)-.275 E(oIP v)
-1.419 E(ariant of TFRC gi)-.275 E -.165(ve)-.275 G 2.75(st).165 G
(he application some)-2.75 E(incenti)100.8 658.4 Q .33 -.165(ve t)-.275
H 2.75(ou).165 G(se fe)-2.75 E(wer b)-.275 E(ut lar)-.22 E(ger pack)
-.198 E(ets, when lar)-.11 E(ger pack)-.198 E(ets w)-.11 E(ould suf)-.11
E(\214ce, by including)-.275 E(the bandwidth used by the pack)100.8
671.4 Q(et header in the allo)-.11 E(wed sending rate.)-.275 E(As an e)
100.8 688 Q(xample, a sender using 120-byte pack)-.165 E
(ets needs a TCP-friendly rate of 128 Kbps to)-.11 E
(send 96 Kbps of application data.)100.8 701 Q
(This is because the TCP-friendly rate is reduced by a)5.5 E -.11(fa)
100.8 714 S
(ctor of s_true/\(s_true + H\) = 120/160, to account for the ef).11 E
(fect of pack)-.275 E(et headers.)-.11 E(If the)5.5 E(Flo)72 769 Q(yd/K)
-.11 E 275.921(ohler Section)-.385 F 2.75(4.2. [P)2.75 F(age 7])-.165 E
0 Cg EP
%%Page: 8 8
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E
(sender suddenly switched to 40-byte data se)100.8 85 Q
(gments, the allo)-.165 E(wed sending rate w)-.275 E(ould reduce)-.11 E
(to 64 Kbps of application data; and one-byte data se)100.8 98 Q
(gments w)-.165 E(ould reduce the allo)-.11 E(wed)-.275 E
(sending rate to 3.12 Kbps of application data.)100.8 111 Q(\(In f)5.5 E
(act, the Min Interv)-.11 E(al w)-.275 E(ould pre)-.11 E -.165(ve)-.275
G(nt).165 E(senders from achie)100.8 124 Q
(ving these rates, since applications using the V)-.275 E(oIP v)-1.419 E
(ariant cannot send)-.275 E(more than 100 pack)100.8 137 Q
(ets per second.\))-.11 E(The V)100.8 153.6 Q(oIP v)-1.419 E(ariant ass\
umes 40 bytes for the header size, although the header could be lar)
-.275 E(ger)-.198 E(\(due to IP options, IPv6, IP tunnels, and the lik)
100.8 166.6 Q(e\) or smaller \(due to header compression\))-.11 E
(on the wire, because using the e)100.8 179.6 Q
(xact header size in bytes w)-.165 E(ould ha)-.11 E .33 -.165(ve l)-.22
H(ittle additional).165 E 2.75(bene\214t. The)100.8 192.6 R -1.419(Vo)
2.75 G(IP v)1.419 E(ariant')-.275 E 2.75(su)-.605 G
(se of an assumed 40-byte header is suf)-2.75 E
(\214cient to get a rough)-.275 E(estimate of the throughput, and to gi)
100.8 205.6 Q .33 -.165(ve t)-.275 H(he application some incenti).165 E
.33 -.165(ve n)-.275 H(ot to use).165 E(unnecessarily-man)100.8 218.6 Q
2.75(ys)-.165 G(mall pack)-2.75 E 2.75(ets. Because)-.11 F
(we are only aiming at rough f)2.75 E(airness, and at a)-.11 E
(rough incenti)100.8 231.6 Q .33 -.165(ve f)-.275 H(or applications, th\
e use of a 40-byte header in the calculations of the).165 E
(header bandwidth seems suf)100.8 244.6 Q(\214cient.)-.275 E/F1 11
/Times-Bold@0 SF(4.3.)72 270.6 Q/F2 13/Times-Bold@0 SF(The V)5.5 E
(oIP Min Inter)-1.3 E -.13(va)-.13 G(l).13 E F0
(The header size calculation pro)100.8 287.2 Q(vides an incenti)-.165 E
.33 -.165(ve f)-.275 H(or applications to use fe).165 E(wer)-.275 E 2.75
(,b)-.44 G(ut lar)-2.97 E(ger)-.198 E(,)-.44 E(pack)100.8 300.2 Q 2.75
(ets. Another)-.11 F(incenti)2.75 E .33 -.165(ve i)-.275 H 2.75(st).165
G(hat when the path limitation is in pps, the application using)-2.75 E
(more small pack)100.8 313.2 Q(ets is lik)-.11 E
(ely to cause higher pack)-.11 E(et drop rates, and to ha)-.11 E .33
-.165(ve t)-.22 H 2.75(or).165 G(educe its)-2.75 E
(sending rate accordingly)100.8 326.2 Q 5.5(.T)-.715 G
(hat is, if the congestion is in terms of pps, then the \215o)-5.5 E
2.75(ws)-.275 G(ending)-2.75 E(more pps will increase the pack)100.8
339.2 Q(et drop rate, and ha)-.11 E .33 -.165(ve t)-.22 H 2.75(oa).165 G
(djust its sending rate accordingly)-2.75 E(.)-.715 E(Ho)100.8 352.2 Q
(we)-.275 E -.165(ve)-.275 G .88 -.44(r, t).165 H
(he increased congestion caused by the use of small pack).44 E
(ets in an en)-.11 E(vironment)-.44 E(limited by pps is e)100.8 365.2 Q
(xperienced not only by the \215o)-.165 E 2.75(wu)-.275 G
(sing the small pack)-2.75 E(ets, b)-.11 E(ut by all of the)-.22 E
(competing traf)100.8 378.2 Q(\214c on that congested link.)-.275 E
(These incenti)5.5 E -.165(ve)-.275 G 2.75(sa).165 G(re therefore insuf)
-2.75 E(\214cient to)-.275 E(pro)100.8 391.2 Q(vide suf)-.165 E
(\214cient protection for pps netw)-.275 E(ork limitations.)-.11 E
(The V)100.8 407.8 Q(oIP v)-1.419 E
(ariant for TFRC, then, includes a Min Interv)-.275 E(al of 10 ms.)-.275
E(This pro)5.5 E(vides)-.165 E
(additional restrictions on the use of unnecessarily man)100.8 420.8 Q
2.75(ys)-.165 G(mall pack)-2.75 E(ets.)-.11 E
(One justi\214cation for the Min Interv)100.8 437.4 Q
(al is the practical one that the applications that currently)-.275 E
-.11(wa)100.8 450.4 S(nt to send small pack).11 E(ets are the V)-.11 E
(oIP applications that send at most one pack)-1.419 E(et e)-.11 E -.165
(ve)-.275 G(ry 10).165 E(ms, so this restriction does not af)100.8 463.4
Q(fect current traf)-.275 E 2.75(\214c. A)-.275 F
(second justi\214cation is that there is)2.75 E
(no pressing need for best-ef)100.8 476.4 Q(fort traf)-.275 E
(\214c in the current Internet to send small pack)-.275 E(ets more)-.11
E(frequently than once e)100.8 489.4 Q -.165(ve)-.275 G
(ry 10 ms \(rather than taking the 10 ms delay at the sender).165 E 2.75
(,a)-.44 G(nd)-2.75 E(mer)100.8 502.4 Q(ging the tw)-.198 E 2.75(os)-.11
G(mall pack)-2.75 E(ets into one lar)-.11 E(ger one\).)-.198 E
(This 10 ms delay for mer)5.5 E(ging small)-.198 E(pack)100.8 515.4 Q
(ets is lik)-.11 E(ely to be dominated by the netw)-.11 E(ork propag)
-.11 E(ation, transmission, and queueing)-.055 E(delays of best-ef)100.8
528.4 Q(fort traf)-.275 E(\214c in the current Internet.)-.275 E
(As a result, our judgement w)5.5 E(ould be that)-.11 E
(the bene\214t to the user of ha)100.8 541.4 Q
(ving less than 10 ms between pack)-.22 E(ets is outweighed by the)-.11
E(bene\214t to the netw)100.8 554.4 Q(ork of a)-.11 E -.22(vo)-.22 G
(iding unnecessarily man).22 E 2.75(ys)-.165 G(mall pack)-2.75 E(ets.)
-.11 E(The Min Interv)100.8 571 Q(al causes the V)-.275 E(oIP v)-1.419 E
(ariant of TFRC not to support applications sending)-.275 E(small pack)
100.8 584 Q(ets v)-.11 E(ery frequently)-.165 E 5.5(.C)-.715 G
(onsider a TFRC \215o)-5.5 E 2.75(ww)-.275 G(ith a \214x)-2.75 E
(ed pack)-.165 E(et size of 100 bytes,)-.11 E -.22(bu)100.8 597 S 2.75
(tw).22 G(ith a v)-2.75 E(ariable sending rate and a f)-.275 E
(airly uncongested path.)-.11 E(When this \215o)5.5 E 2.75(ww)-.275 G
(as sending)-2.86 E(at most 100 pps, it w)100.8 610 Q
(ould be able to use the V)-.11 E(oIP v)-1.419 E(ariant of TFRC.)-.275 E
(If the \215o)5.5 E 2.75(ww)-.275 G(ished to)-2.75 E
(increase its sending rate to more than 100 pps, b)100.8 623 Q(ut to k)
-.22 E(eep the same pack)-.11 E(et size, it w)-.11 E(ould)-.11 E
(no longer be able to achie)100.8 636 Q .33 -.165(ve t)-.275 H
(his with the V).165 E(oIP v)-1.419 E(ariant to TFRC, and w)-.275 E
(ould ha)-.11 E .33 -.165(ve t)-.22 H 2.75(os).165 G(witch)-2.75 E
(to the def)100.8 649 Q(ault TFRC, recei)-.11 E
(ving a dramatic, discontinuous decrease in its allo)-.275 E
(wed sending)-.275 E 2.75(rate. This)100.8 662 R
(seems not only acceptable, b)2.75 E
(ut desirable for the global Internet.)-.22 E(What is to pre)100.8 678.6
Q -.165(ve)-.275 G(nt \215o).165 E
(ws from opening multiple connections, each with a 10 ms Min)-.275 E
(Interv)100.8 691.6 Q
(al, and thereby getting around the limitation of the Min Interv)-.275 E
2.75(al? Ob)-.275 F(viously)-.165 E 2.75(,t)-.715 G(here is)-2.75 E
(nothing to pre)100.8 704.6 Q -.165(ve)-.275 G(nt \215o).165 E
(ws from doing this, just as there is currently nothing to pre)-.275 E
-.165(ve)-.275 G(nt \215o).165 E(ws)-.275 E(from using UDP)100.8 717.6 Q
2.75(,o)-1.221 G 2.75(rf)-2.75 G
(rom opening multiple parallel TCP connections, or from using their)
-2.75 E(Flo)72 769 Q(yd/K)-.11 E 275.921(ohler Section)-.385 F 2.75
(4.3. [P)2.75 F(age 8])-.165 E 0 Cg EP
%%Page: 9 9
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E -.275(ow)100.8 85 S 2.75(nc)
.275 G(ongestion control mechanism.)-2.75 E
(Of course, implementations or middlebox)5.5 E(es are also)-.165 E(free\
 to limit the number of parallel TFRC connections opened to the same de\
stination in)100.8 98 Q(times of congestion, if that seems desirable.)
100.8 111 Q(And \215o)5.5 E(ws that open multiple parallel)-.275 E
(connections are subject to the incon)100.8 124 Q -.165(ve)-.44 G
(niences of reordering and the lik).165 E 2.75(e. But)-.11 F -2.365
-.275(ev e)2.75 H 2.75(nw).275 G(ithout)-2.75 E 2.75(am)100.8 137 S
(echanism to pre)-2.75 E -.165(ve)-.275 G(nt \215o).165 E(ws from sub)
-.275 E -.165(ve)-.165 G(rting the Min Interv).165 E
(al by opening multiple)-.275 E
(parallel connections, it seems useful to include the Min Interv)100.8
150 Q(al in the V)-.275 E(oIP v)-1.419 E(ariant of)-.275 E(TFRC.)100.8
163 Q/F1 11/Times-Bold@0 SF(4.4.)72 189 Q/F2 13/Times-Bold@0 SF
(Counting P)5.5 E(ack)-.13 E(et Losses)-.13 E F0(It is not possible for\
 a TCP connection to persistently send multiple pack)100.8 205.6 Q
(ets per round-trip)-.11 E(time in the f)100.8 218.6 Q
(ace of high congestion, with a steady-state with multiple pack)-.11 E
(ets dropped per)-.11 E(round-trip time.)100.8 231.6 Q -.165(Fo)5.5 G
2.75(rT).165 G(CP)-2.75 E 2.75(,w)-1.221 G(hen one or more pack)-2.75 E
(ets are dropped each round-trip time, the)-.11 E
(sending rate is quickly dropped to less than one pack)100.8 244.6 Q
(et per round-trip time.)-.11 E(In addition, for)5.5 E(TCP with T)100.8
257.6 Q(ahoe, Ne)-.88 E(wReno, or SA)-.275 E
(CK congestion control mechanisms, the response to)-.44 E
(congestion is lar)100.8 270.6 Q(gely independent of the number of pack)
-.198 E(ets dropped per round-trip time.)-.11 E
(As a result, standard TFRC can best achie)100.8 287.2 Q .33 -.165(ve f)
-.275 H(airness with TCP).055 E 2.75(,e)-1.221 G -.165(ve)-3.025 G 2.75
(ni).165 G 2.75(nh)-2.75 G(ighly congested)-2.75 E(en)100.8 300.2 Q
(vironments, by calculating the loss e)-.44 E -.165(ve)-.275 G
(nt rate rather than the pack).165 E(et drop rate, where a)-.11 E
(loss e)100.8 313.2 Q -.165(ve)-.275 G(nt is one or more pack).165 E
(ets dropped or mark)-.11 E(ed from a windo)-.11 E 2.75(wo)-.275 G 2.75
(fd)-2.75 G(ata.)-2.75 E(Ho)100.8 329.8 Q(we)-.275 E -.165(ve)-.275 G
.88 -.44(r, w).165 H(ith the V).44 E(oIP v)-1.419 E
(ariant of TFRC, it is no longer possible to achie)-.275 E .33 -.165
(ve f)-.275 H(airness with).055 E
(TCP or with standard TFRC by counting only the loss e)100.8 342.8 Q
-.165(ve)-.275 G(nt rate [WBL04].).165 E(Instead of)5.5 E
(sending one lar)100.8 355.8 Q(ge pack)-.198 E
(et per round-trip time, the V)-.11 E(oIP v)-1.419 E
(ariant of TFRC could be sending N)-.275 E(small pack)100.8 368.8 Q
(ets \(where N small pack)-.11 E(ets equal one lar)-.11 E
(ge 1500-byte pack)-.198 E 2.75(et\). The)-.11 F(loss)2.75 E
(measurement used with the V)100.8 381.8 Q(oIP v)-1.419 E
(ariant of TFRC has to be able to detect a connection that)-.275 E
(is sending multiple pack)100.8 394.8 Q
(ets per round-trip time in the f)-.11 E(ace of multiple pack)-.11 E
(et losses or)-.11 E(marks per round-trip time.)100.8 407.8 Q(In the V)
100.8 424.4 Q(oIP v)-1.419 E(ariant of TFRC, the loss e)-.275 E -.165
(ve)-.275 G(nt rate is calculated by counting at most one loss).165 E
-2.365 -.275(ev e)100.8 437.4 T(nt in loss interv).275 E
(als longer than tw)-.275 E 2.75(or)-.11 G
(ound-trip times, and by counting each pack)-2.75 E(et lost or)-.11 E
(mark)100.8 450.4 Q(ed in shorter loss interv)-.11 E 2.75(als. In)-.275
F(particular)2.75 E 2.75(,f)-.44 G(or a short loss interv)-2.75 E
(al with N pack)-.275 E(ets,)-.11 E(including K lost or mark)100.8 463.4
Q(ed pack)-.11 E(ets, the loss interv)-.11 E
(al length is calculated as N/K, instead as)-.275 E 2.75(N. The)100.8
476.4 R -2.475 -.22(av e)2.75 H(rage loss interv).22 E
(al I_mean is still a)-.275 E -.165(ve)-.22 G(raged o).165 E -.165(ve)
-.165 G 2.75(rt).165 G(he most recent eight loss)-2.75 E(interv)100.8
489.4 Q(als, as speci\214ed in Section 5.4 of RFC 3448.)-.275 E
(Thus, if eight successi)5.5 E .33 -.165(ve l)-.275 H(oss interv).165 E
(als)-.275 E(are short loss interv)100.8 502.4 Q(als with N pack)-.275 E
(ets and K losses, the loss e)-.11 E -.165(ve)-.275 G
(nt rate is calculated as).165 E(K/N, rather than as 1/N.)100.8 515.4 Q
F1(5.)72 541.4 Q/F3 14/Times-Bold@0 SF 3.5(AC)5.5 G
(omparison with RFC 3714)-3.5 E F1(6.)72 567.4 Q F3(The V)5.5 E(oIP V)
-1.4 E(ariant with A)-1.288 E(pplications that Modify the P)-.35 E(ack)
-.14 E(et Size)-.14 E F0 1.76 -.88(To b)100.8 584 T 2.75(ed).88 G(one.)
-2.75 E F1(7.)72 610 Q F3(Simulation Results)5.5 E F1(7.1.)72 636 Q F2
(Simulations with Con\214gur)5.5 E(ed Dr)-.234 E(op Rates)-.234 E F0(In\
 this section we describe simulation results from simulations comparing\
 the throughput)100.8 652.6 Q(of standard \(SA)100.8 665.6 Q
(CK\) TCP \215o)-.44 E(ws, TCP \215o)-.275 E
(ws with timestamps and ECN, V)-.275 E(oIP TFRC \215o)-1.419 E(ws,)-.275
E(and standard TFRC \(Stnd TFRC\) \215o)100.8 678.6 Q 2.75(ws. In)-.275
F(these simulations we con\214gure the router to)2.75 E
(randomly drop or mark pack)100.8 691.6 Q
(ets at a speci\214ed rate, independently of the pack)-.11 E(et size.)
-.11 E -.165(Fo)5.5 G(r).165 E(each speci\214ed pack)100.8 704.6 Q
(et drop rate, we gi)-.11 E .33 -.165(ve a \215)-.275 H -.275(ow).165 G
1.21 -.605('s a).275 H -.165(ve).385 G(rage sending rate in Kbps o).165
E -.165(ve)-.165 G 2.75(rt).165 G(he)-2.75 E
(second half of a 100-second simulation, a)100.8 717.6 Q -.165(ve)-.22 G
(raged o).165 E -.165(ve)-.165 G 2.75(rt).165 G(en \215o)-2.75 E(ws.)
-.275 E(Flo)72 769 Q(yd/K)-.11 E 275.921(ohler Section)-.385 F 2.75
(7.1. [P)2.75 F(age 9])-.165 E 0 Cg EP
%%Page: 10 10
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E/F1 11/Courier@0 SF 33
(Packet TCP)114 85 R(ECN TCP)52.8 E(TFRC)33 E 19.8(DropRate SendRate)114
98 R 19.8(SendRate SendRate)19.8 F 19.8(-------- --------)114 111 R 19.8
(-------- --------)19.8 F 19.8(0.001 2020.85 1904.61)133.8 124 R(982.09)
33 E 26.4(0.005 811.10 792.11 878.08)133.8 137 R 33(0.01 515.45)133.8
150 R 26.4(533.19 598.90)33 F 33(0.02 362.93)133.8 163 R 26.4
(382.67 431.41)33 F 33(0.04 250.06)133.8 176 R 26.4(252.64 284.82)33 F
33(0.05 204.48)133.8 189 R 26.4(218.16 268.51)33 F 26.4
(0.066 176.40 178.16 211.05)133.8 202 R 39.6(0.1 143.30)133.8 215 R 26.4
(148.41 146.03)33 F 46.2(0.2 78.65)133.8 228 R 26.4(93.23* 55.14)39.6 F
46.2(0.3 26.26)133.8 241 R 26.4(59.65* 32.87)39.6 F 52.8(0.4 9.87)133.8
254 R 26.4(47.79* 25.45)39.6 F 52.8(0.5 3.53)133.8 267 R 26.4
(32.01* 18.52)39.6 F(Table 1:)100.8 293 Q
("Total Sending Rates \(Kbps\) for Configured Packet Drop Rates.)100.8
306 Q(The TFRC flow uses 1460-byte data packets, with a maximum)100.8
319 Q(data sending rate of 1000 Kbps.")100.8 332 Q 6.6(*N)100.8 358 S
(ote: These ECN scenarios are not realistic, as routers are)-6.6 E
(not likely to mark packets when packet drop/mark rates are 20%)100.8
371 Q(or higher.)100.8 384 Q F0 -.88(Ta)100.8 416 S(ble 1 sho).88 E
(ws the sending rate for a TCP and a standard TFRC \215o)-.275 E 2.75
(wf)-.275 G(or a range of)-2.75 E(con\214gured pack)100.8 429 Q
(et drop rates, when both \215o)-.11 E(ws ha)-.275 E .33 -.165(ve 1)-.22
H(460-byte data pack).165 E(ets, in order to)-.11 E
(illustrate the relati)100.8 442 Q .33 -.165(ve f)-.275 H
(airness of TCP and TFRC when both \215o).055 E(ws use the same pack)
-.275 E(et size.)-.11 E -.165(Fo)100.8 455 S 2.75(re).165 G
(xample, a pack)-2.915 E
(et drop rate of 0.1 means that 10% of the TCP and TFRC pack)-.11 E
(ets are)-.11 E 2.75(dropped. There)100.8 468 R(is good relati)2.75 E
.33 -.165(ve f)-.275 H(airness until the pack).055 E
(et drop percentages reach 40 and)-.11 E(50%, when the TFRC \215o)100.8
481 Q 2.75(wr)-.275 G(ecei)-2.75 E -.165(ve)-.275 G 2.75(st).165 G
(hree to \214v)-2.75 E 2.75(et)-.165 G
(imes more bandwidth than the standard)-2.75 E(TCP \215o)100.8 494 Q
4.18 -.715(w. W)-.275 H 2.75(en)-.165 G(ote that an ECN TCP \215o)-2.75
E 2.75(ww)-.275 G(ould recei)-2.86 E .33 -.165(ve a h)-.275 H
(igher throughput than the).165 E(TFRC \215o)100.8 507 Q 1.43 -.715
(w, b)-.275 H(ut this w).495 E(ould not be rele)-.11 E -.275(va)-.275 G
(nt in practice, as routers are advised to drop rather).275 E
(than mark pack)100.8 520 Q(ets during high le)-.11 E -.165(ve)-.275 G
(ls of congestion.).165 E(Flo)72 769 Q(yd/K)-.11 E 270.421
(ohler Section)-.385 F 2.75(7.1. [P)2.75 F(age 10])-.165 E 0 Cg EP
%%Page: 11 11
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E/F1 11/Courier@0 SF 33
(Packet TCP)114 85 R(ECN TCP)52.8 E(VoIP TFRC)39.6 E(Stnd TFRC)13.2 E
19.8(DropRate SendRate)114 98 R 26.4(SendRate SendRate)19.8 F(SendRate)
19.8 E 19.8(-------- --------)114 111 R 26.4(-------- --------)19.8 F
(--------)19.8 E 19.8(0.001 1787.54 1993.03)133.8 124 R 33(17.71 17.69)
39.6 F 26.4(0.005 785.11 823.75)133.8 137 R 33(18.11 17.69)39.6 F 33
(0.01 533.38)133.8 150 R 33(529.01 17.69 17.80)33 F 33(0.02 317.16)133.8
163 R 33(399.62 17.69 13.41)33 F 33(0.04 245.42)133.8 176 R 33
(260.57 17.69)33 F(8.84)46.2 E 33(0.05 216.38)133.8 189 R 33
(223.75 17.69)33 F(7.63)46.2 E 26.4(0.066 174.07 184.37)133.8 202 R 39.6
(17.69 6.46)39.6 F 39.6(0.1 142.75)133.8 215 R 33(138.36 17.69)33 F
(4.29)46.2 E 46.2(0.2 58.61)133.8 228 R 26.4(91.54* 17.80)39.6 F(1.94)
46.2 E 46.2(0.3 21.62)133.8 241 R 26.4(63.96* 10.26)39.6 F(1.00)46.2 E
46.2(0.4 10.51)133.8 254 R 33(41.74* 4.78)39.6 F(0.77)46.2 E 52.8
(0.5 1.92)133.8 267 R 33(19.03* 2.41)39.6 F(0.56)46.2 E(Table 2:)100.8
293 Q("Total Sending Rates \(Kbps\) for Configured Packet Drop Rates.)
100.8 306 Q(The TFRC flows use 14-byte data packets, with a maximum)
100.8 319 Q(data sending rate of 5.6 Kbps.")100.8 332 Q 6.6(*N)100.8 358
S(ote: These ECN scenarios are not realistic, as routers are)-6.6 E
(not likely to mark packets when packet drop/mark rates are 20%)100.8
371 Q(or higher.)100.8 384 Q F0 -.88(Ta)100.8 406.6 S(ble 2 sho).88 E
(ws the results of simulations where each V)-.275 E(oIP TFRC \215o)
-1.419 E 2.75(wh)-.275 G(as a maximum data)-2.75 E
(sending rate of 5.6 Kbps, with 14-byte data pack)100.8 419.6 Q
(ets and a 32-byte pack)-.11 E(et header for DCCP)-.11 E(and IP)100.8
432.6 Q 5.5(.E)-1.221 G(ach TCP \215o)-5.5 E 2.75(wh)-.275 G(as a recei)
-2.75 E .33 -.165(ve w)-.275 H(indo).165 E 2.75(wo)-.275 G 2.75(f1)-2.75
G(00 pack)-2.75 E(ets and a data pack)-.11 E(et size of 1460)-.11 E
(bytes, with a 40-byte pack)100.8 445.6 Q(et header for TCP and IP)-.11
E 5.5(.T)-1.221 G(he TCP \215o)-5.5 E 2.75(wu)-.275 G(ses Sack TCP with)
-2.75 E(Limited T)100.8 458.6 Q(ransmit, b)-.385 E
(ut without timestamps or ECN.)-.22 E(Each \215o)5.5 E 2.75(wh)-.275 G
(as a round-trip time of 240)-2.75 E(ms.)100.8 471.6 Q
(The TFRC sending rate in T)100.8 497.6 Q
(able 2 is the sending rate for the 14-byte data pack)-.88 E
(et with the)-.11 E(32-byte pack)100.8 510.6 Q(et header)-.11 E 5.5(.T)
-.605 G(hus, only 30% of the TFRC sending rate is for data, and with a)
-5.5 E(pack)100.8 523.6 Q
(et drop rate of p, only a fraction 1-p of that data mak)-.11 E
(es it to the recei)-.11 E -.165(ve)-.275 G 3.96 -.605(r. T).165 H
(hus, the).605 E(TFRC data recei)100.8 536.6 Q .33 -.165(ve r)-.275 H
(ate can be considerably less than the TFRC sending rate in the table.)
.165 E(Because TCP uses lar)100.8 549.6 Q(ge pack)-.198 E
(ets, 97% of the TCP sending rate is for data, and the same)-.11 E
(fraction 1-p of that data mak)100.8 562.6 Q(es it to the recei)-.11 E
-.165(ve)-.275 G -.605(r.).165 G -.88(Ta)100.8 588.6 S(ble 2 sho).88 E(\
ws that for the 5.6 Kbps data stream with TFRC, Standard TFRC \(Stnd TF\
RC\))-.275 E(gi)100.8 601.6 Q -.165(ve)-.275 G 2.75(sav).165 G
(ery poor sending rate in bps, relati)-2.915 E .33 -.165(ve t)-.275 H
2.75(ot).165 G(he sending rate for the lar)-2.75 E(ge-pack)-.198 E
(et TCP)-.11 E 2.75(connection. In)100.8 614.6 R
(contrast, the sending rate for the V)2.75 E(oIP TFRC \215o)-1.419 E
2.75(wi)-.275 G 2.75(sr)-2.75 G(elati)-2.75 E -.165(ve)-.275 G
(ly close to the).165 E(desired goal of f)100.8 627.6 Q
(airness in bps with TCP)-.11 E(.)-1.221 E -.88(Ta)100.8 653.6 S
(ble 2 sho).88 E(ws that with V)-.275 E
(oIP TFRC, the 5.6 Kbps data stream doesn')-1.419 E 2.75(tr)-.198 G
(educe its sending)-2.75 E(rate until pack)100.8 666.6 Q
(et drop rates greater than 20%, as desired.)-.11 E -.44(Wi)5.5 G
(th pack).44 E(et drop rates of)-.11 E
(30-40%, the sending rate for the V)100.8 679.6 Q(oIP TFRC \215o)-1.419
E 2.75(wi)-.275 G 2.75(ss)-2.75 G(ome)-2.75 E
(what less than that of the a)-.275 E -.165(ve)-.22 G(rage).165 E(lar)
100.8 692.6 Q(ge-pack)-.198 E(et TCP \215o)-.11 E 1.43 -.715(w, w)-.275
H(hile for pack).715 E(et drop rates of 50% the sending rate for the V)
-.11 E(oIP)-1.419 E(TFRC \215o)100.8 705.6 Q 2.75(wi)-.275 G 2.75(ss)
-2.75 G(ome)-2.75 E(what greater than that of the a)-.275 E -.165(ve)
-.22 G(rage lar).165 E(ge-pack)-.198 E(et TCP \215o)-.11 E -.715(w.)
-.275 G(Flo)72 769 Q(yd/K)-.11 E 270.421(ohler Section)-.385 F 2.75
(7.1. [P)2.75 F(age 11])-.165 E 0 Cg EP
%%Page: 12 12
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E 1.76 -.88(We n)100.8 85 T
(ote that the high sending rate for ECN TCP in en).88 E
(vironments with high marking rates)-.44 E(is lar)100.8 98 Q
(gely irrele)-.198 E -.275(va)-.275 G
(nt, as routers are advised to drop rather than mark ECN-capable pack)
.275 E(ets)-.11 E(in times of high congestion.)100.8 111 Q
(The sending rate of a TCP connection using timestamps is)5.5 E
(similar to the sending rate sho)100.8 124 Q
(wn for a standard TCP connection without timestamps.)-.275 E/F1 11
/Courier@0 SF 46.2(Byte TCP ECN)114 143 R 33(TCP VoIP)6.6 F 6.6
(TFRC Stnd)6.6 F(TFRC)6.6 E 19.8(DropRate SendRate)114 156 R 26.4
(SendRate SendRate)19.8 F(SendRate)19.8 E 19.8(-------- --------)114 169
R 26.4(-------- --------)19.8 F(--------)19.8 E 13.2(0.00001 423.02)
133.8 182 R 33(406.44 17.69 17.69)33 F 19.8(0.0001 117.41)133.8 195 R 33
(114.34 17.69 17.69)33 F 39.6(0.001 0.41 3.38*)133.8 208 R 39.6
(17.69 8.37)33 F 39.6(0.005 0.26 0.26*)133.8 221 R 39.6(18.39 1.91)33 F
46.2(0.01 0.31)133.8 234 R 33(0.26* 7.07)46.2 F(0.84)46.2 E 46.2
(0.02 0.29)133.8 247 R 33(0.26* 1.61)46.2 F(0.43)46.2 E 46.2(0.04 0.12)
133.8 260 R 33(0.26* 0.17)46.2 F(0.12)46.2 E 46.2(0.05 0.15)133.8 273 R
33(0.26* 0.08)46.2 F(0.06)46.2 E(Table 3:)100.8 299 Q
("Total Sending Rates \(Kbps\) for Configured Byte Drop Rates.)100.8 312
Q(The TFRC flows use 14-byte data packets, with a maximum)100.8 325 Q
(data sending rate of 5.6 Kbps.")100.8 338 Q 6.6(*N)100.8 364 S
(ote: These ECN scenarios are not realistic, as routers are)-6.6 E
(not likely to mark packets when packet drop/mark rates are 20%)100.8
377 Q(or higher.)100.8 390 Q 46.2(Byte TCP)114 415 R 19.8(Pkt TFRC)6.6 F
13.2(Pkt TCP/TFRC)6.6 F 19.8(DropRate DropRate)114 428 R 13.2
(DropRate Pkt)19.8 F(Drop Ratio)6.6 E 19.8(-------- --------)114 441 R
13.2(-------- --------------)19.8 F 33(0.00001 0.015)120.6 454 R 26.4
(0.0006 26.59)33 F 39.6(0.0001 0.13)120.6 467 R 26.4(0.0056 24.94)39.6 F
46.2(0.001 0.77)120.6 480 R 33(0.054 14.26)39.6 F 46.2(0.005 0.99)120.6
493 R 46.2(0.24 4.08)39.6 F 52.8(0.01 1.0)120.6 506 R 46.2(0.43 2.32)
46.2 F 52.8(0.05 1.0)120.6 519 R 46.2(0.94 1.05)46.2 F(Table 4:)100.8
545 Q("Converting Byte Drop Rates to Packet Drop Rates,)100.8 558 Q
(for 1500-byte TCP packets and 56-byte TFRC packets.")100.8 571 Q F0
(In contrast, T)100.8 593.6 Q(able 3 sho)-.88 E
(ws the TCP and TFRC send rates for v)-.275 E(arious byte drop rates.)
-.275 E -.165(Fo)5.5 G(r).165 E
(these simulations, for each *byte*, the pack)100.8 606.6 Q
(et containing that byte is dropped with)-.11 E(probability p.)100.8
619.6 Q -.88(Ta)5.5 G(ble 4 con).88 E -.165(ve)-.44 G
(rts the byte drop rate p to pack).165 E(et drop rates for the TCP and)
-.11 E(TFRC pack)100.8 632.6 Q(ets, where the pack)-.11 E
(et drop rate for an N-byte pack)-.11 E(et is 1-\(1-p\)^N.)-.11 E
(Thus, a byte)5.5 E(drop rate of 0.001, with each byte being dropped wi\
th probability 0.001, con)100.8 645.6 Q -.165(ve)-.44 G(rts to a).165 E
(pack)100.8 658.6 Q
(et drop rate of 0.77, or 77%, for the 1500-byte TCP pack)-.11 E
(ets, and a pack)-.11 E(et drop rate of)-.11 E
(0.054, or 5.4%, for the 56-byte TFRC pack)100.8 671.6 Q(ets.)-.11 E
(The last column of T)100.8 688.2 Q(able 3 sho)-.88 E
(ws the ratio between the TCP pack)-.275 E(et drop rate and the)-.11 E
(TFRC pack)100.8 701.2 Q(et drop rate.)-.11 E -.165(Fo)5.5 G 2.75(rl)
.165 G .55 -.275(ow b)-2.75 H
(yte drop rates, this ratio is close to 26.8, the ratio).275 E
(between the TCP and TFRC pack)100.8 714.2 Q(et sizes.)-.11 E -.165(Fo)
5.5 G 2.75(rh).165 G(igh byte drop rates, where e)-2.75 E -.165(ve)-.275
G 2.75(nm).165 G(ost small)-2.75 E(Flo)72 769 Q(yd/K)-.11 E 270.421
(ohler Section)-.385 F 2.75(7.1. [P)2.75 F(age 12])-.165 E 0 Cg EP
%%Page: 13 13
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E(TFRC pack)100.8 85 Q
(ets are lik)-.11 E(ely to be dropped, this drop ratio approaches 1.)
-.11 E(As T)100.8 101.6 Q(able 3 sho)-.88 E
(ws. with moderate byte drop rates \(between 0.1% and 4%\), e)-.275 E
-.165(ve)-.275 G 2.75(nt).165 G(he)-2.75 E(standard TFRC \215o)100.8
114.6 Q 2.75(ww)-.275 G(ith 14-byte data pack)-2.75 E
(ets has higher throughput than the lar)-.11 E(ge-pack)-.198 E(et)-.11 E
(TCP \215o)100.8 127.6 Q 4.18 -.715(w. I)-.275 H 2.75(nt).715 G(hese re)
-2.75 E(gimes, the pack)-.165 E
(et drop rate for TCP is greater than 50%, and the TCP)-.11 E
(sending rate no longer v)100.8 140.6 Q
(aries as 1/sqrt\(p\), as it does with smaller pack)-.275 E
(et drop rates.)-.11 E -.44(Wi)5.5 G(th).44 E
(these equal byte drop rates for the TCP and TFRC \215o)100.8 153.6 Q
(ws, the V)-.275 E(oIP v)-1.419 E(ariant of TFRC mak)-.275 E(es)-.11 E
(the unf)100.8 166.6 Q(airness in f)-.11 E -.22(avo)-.11 G 2.75(ro).22 G
2.75(fT)-2.75 G(FRC e)-2.75 E -.165(ve)-.275 G 2.75(nw).165 G
(orse, with the V)-2.86 E(oIP TFRC \215o)-1.419 E 2.75(wr)-.275 G(ecei)
-2.75 E(ving)-.275 E
(signi\214cantly higher throughput than the TCP \215o)100.8 179.6 Q 2.75
(wf)-.275 G(or byte drop rates from 0.1% to 2%.)-2.75 E/F1 11/Courier@0
SF 33(Packet TCP)114 202.2 R(ECN TCP)52.8 E(VoIP TFRC)33 E(Stnd TFRC)
13.2 E 19.8(DropRate SendRate)114 215.2 R 19.8(SendRate SendRate)19.8 F
(SendRate)19.8 E 19.8(-------- --------)114 228.2 R 19.8
(-------- --------)19.8 F(--------)19.8 E 19.8(0.001 1908.98 1869.24)
133.8 241.2 R 26.4(183.45 178.35)33 F 26.4
(0.005 854.69 835.10 185.06 138.06)133.8 254.2 R 33(0.01 564.10)133.8
267.2 R 26.4(531.10 185.33)33 F(92.43)39.6 E 33(0.02 365.38)133.8 280.2
R 26.4(369.10 185.57)33 F(62.18)39.6 E 33(0.04 220.80)133.8 293.2 R 26.4
(257.81 185.14)33 F(45.43)39.6 E 33(0.05 208.97)133.8 306.2 R 26.4
(219.41 180.08)33 F(39.44)39.6 E 26.4(0.066 179.04 184.68 168.51)133.8
319.2 R(31.16)39.6 E 39.6(0.1 141.67)133.8 332.2 R 26.4(143.88 127.33)33
F(21.96)39.6 E 46.2(0.2 62.66)133.8 345.2 R 26.4(91.87* 54.66)39.6 F
(9.40)46.2 E 46.2(0.3 16.63)133.8 358.2 R 26.4(65.52* 24.50)39.6 F(4.73)
46.2 E 52.8(0.4 6.62)133.8 371.2 R 26.4(42.00* 13.47)39.6 F(3.35)46.2 E
52.8(0.5 1.01)133.8 384.2 R 26.4(21.34* 10.51)39.6 F(2.92)46.2 E
(Table 5:)100.8 410.2 Q
("Total Sending Rates \(in Kbps\) for Configured Packet Drop Rates.)
100.8 423.2 Q(The TFRC flows use 200-byte data packets, with a maximum)
100.8 436.2 Q(data sending rate of 160 Kbps.")100.8 449.2 Q 6.6(*N)100.8
475.2 S(ote: These ECN scenarios are not realistic, as routers are)-6.6
E(not likely to mark packets when packet drop/mark rates are 20%)100.8
488.2 Q(or higher.)100.8 501.2 Q F0(Using con\214gured pack)100.8 533.2
Q(et drop rates, T)-.11 E(able 5 compares the a)-.88 E -.165(ve)-.22 G
(rage per).165 E(-\215o)-.22 E 2.75(ws)-.275 G(ending rates)-2.75 E
(when the TFRC \215o)100.8 546.2 Q 2.75(wh)-.275 G
(as a maximum data sending rate of 160 Kbps, with the application)-2.75
E(generating 200-byte data pack)100.8 559.2 Q(ets at 100 pack)-.11 E
(ets per second.)-.11 E(As e)5.5 E(xpected with equal pack)-.165 E(et)
-.11 E(drop rates, the performance of Standard TFRC is quite poor)100.8
572.2 Q 2.75(,w)-.44 G(hile the performance of)-2.75 E -1.419(Vo)100.8
585.2 S(IP TFRC is essentially as desired for pack)1.419 E
(et drop rates up to 30%.)-.11 E(Ag)5.5 E(ain as e)-.055 E(xpected,)
-.165 E(with pack)100.8 598.2 Q(et drop rates of 40-50% the V)-.11 E
(oIP TFRC sending rate is some)-1.419 E(what higher than)-.275 E
(desired.)100.8 611.2 Q(In general, T)100.8 637.2 Q(ables 2 and 5 sho)
-.88 E 2.75(wa)-.275 G(cceptable performance for V)-2.75 E
(oIP TFRC in en)-1.419 E(vironments)-.44 E(with stable pack)100.8 650.2
Q(et drop rates, where the decision to drop a pack)-.11 E
(et is independent of the)-.11 E(pack)100.8 663.2 Q(et size.)-.11 E(Ho)
5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, i).165 H 2.75(nr).44 G
(ealistic en)-2.75 E(vironments, the pack)-.44 E(et size might af)-.11 E
(fect the lik)-.275 E(elihood)-.11 E(that a pack)100.8 676.2 Q
(et is dropped.)-.11 E -.165(Fo)5.5 G 2.75(re).165 G(xample, with hea)
-2.915 E(vy congestion and a Drop T)-.22 E(ail queue with a)-.88 E<8c78>
100.8 689.2 Q(ed number of bytes rather than a \214x)-.165 E
(ed number of pack)-.165 E(et-sized b)-.11 E(uf)-.22 E(fers, small pack)
-.275 E(ets)-.11 E(might be more lik)100.8 702.2 Q(ely than lar)-.11 E
(ge pack)-.198 E(ets to \214nd room at the end of an almost-full queue.)
-.11 E(As a further complication, in a scenario with Acti)100.8 715.2 Q
.33 -.165(ve Q)-.275 H(ueue Management, the A).165 E(QM)-.605 E(Flo)72
769 Q(yd/K)-.11 E 270.421(ohler Section)-.385 F 2.75(7.1. [P)2.75 F
(age 13])-.165 E 0 Cg EP
%%Page: 14 14
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E
(mechanism could either be in pack)100.8 85 Q
(et mode, dropping each pack)-.11 E(et with equal probability)-.11 E
2.75(,o)-.715 G(r)-2.75 E
(in byte mode, dropping each byte with equal probability)100.8 98 Q 5.5
(.S)-.715 G(ections 7.2 and 7.3 sho)-5.5 E(w)-.275 E
(simulations with pack)100.8 111 Q(ets dropped at Drop T)-.11 E
(ail or A)-.88 E(QM queues, rather that from a)-.605 E
(probabilistic process.)100.8 124 Q -1.419(Vo)100.8 150 S
(IP mode for TFRC has been added to the NS simulator)1.419 E 2.75(,a)
-.44 G(nd is illustrated in the)-2.75 E -.275(va)100.8 163 S
(lidation test "./test-all-friendly" in the directory tcl/tests.).275 E
(The simulation scripts for the)5.5 E
(simulations in this document will be made a)100.8 176 Q -.275(va)-.22 G
(ilable in).275 E("http://www)100.8 189 Q(.icir)-.715 E(.or)-.605 E
(g/tfrc/v)-.198 E(oipsims.html".)-.22 E/F1 11/Times-Bold@0 SF(7.2.)72
215 Q/F2 13/Times-Bold@0 SF -.13(Pa)5.5 G(ck).13 E(et Dr)-.13 E
(opping Beha)-.234 E(vior at Routers with Dr)-.325 E(opT)-.234 E
(ail Queues)-1.196 E F0
(One of the problems with comparing the throughput of tw)100.8 231.6 Q
2.75<6f8d>-.11 G -.275(ow)-2.75 G 2.75(su).275 G(sing dif)-2.75 E
(ferent pack)-.275 E(et)-.11 E(sizes is that the pack)100.8 244.6 Q
(et size itself can in\215uence the pack)-.11 E
(et drop rate [V00, WBL04].)-.11 E(The def)100.8 261.2 Q
(ault TFRC, without the V)-.11 E(oIP v)-1.419 E(ariant, w)-.275 E
(as designed for rough f)-.11 E(airness with TCP)-.11 E 2.75(,f)-1.221 G
(or)-2.75 E(TFRC and TCP \215o)100.8 274.2 Q(ws with the same pack)-.275
E(et size and e)-.11 E(xperiencing the same pack)-.165 E(et drop)-.11 E
2.75(rate. When)100.8 287.2 R(the issue of f)2.75 E
(airness between \215o)-.11 E(ws with dif)-.275 E(ferent pack)-.275 E
(ets sizes is addressed, it)-.11 E(matters whether the pack)100.8 300.2
Q(et drop rates e)-.11 E(xperienced by the \215o)-.165 E
(ws is related to the pack)-.275 E(et size.)-.11 E(That is, are small V)
100.8 313.2 Q(oIP pack)-1.419 E(ets just as lik)-.11 E
(ely to be dropped as lar)-.11 E(ge TCP pack)-.198 E(ets, or are the)
-.11 E(smaller pack)100.8 326.2 Q(ets less lik)-.11 E
(ely to be dropped [WBL04]? And what is the relationship between)-.11 E
(the pack)100.8 339.2 Q(et-dropping beha)-.11 E
(vior of the path, and the loss e)-.22 E -.165(ve)-.275 G
(nt measurements of TFRC?).165 E/F3 11/Courier@0 SF 46.2(Web TCP)140.4
358.2 R 39.6(TCP VoIP_TFRC)46.2 F(VoIP_TFRC)6.6 E 13.2
(Sessions DropRate)127.2 371.2 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)127.2 384.2 R 19.8(-------- --------)13.2
F(--------)13.2 E 39.6(10 0.04)153.6 397.2 R 39.6(316.18 0.05)33 F
(183.05)33 E 39.6(25 0.07)153.6 410.2 R 39.6(227.47 0.07)33 F(181.23)33
E 39.6(50 0.08)153.6 423.2 R 39.6(181.10 0.08)33 F(178.32)33 E 39.6
(100 0.14)147 436.2 R 39.6(85.97 0.12)39.6 F(151.42)33 E 39.6(200 0.17)
147 449.2 R 39.6(61.20 0.14)39.6 F(73.88)39.6 E 39.6(400 0.20)147 462.2
R 39.6(27.79 0.18)39.6 F(36.81)39.6 E 39.6(800 0.29 3.50 0.27)147 475.2
R(16.33)39.6 E 39.6(1600 0.37 0.63 0.33 6.29)140.4 488.2 R(Table 6:)
100.8 514.2 Q("Simulation Results with Drop-Tail Queues in Packets.")
100.8 527.2 Q F0 -.88(Ta)100.8 549.8 S(ble 6 sho).88 E(ws the results o\
f the second half of 100-second simulations, with \214v)-.275 E 2.75(eT)
-.165 G(CP)-2.75 E(connections and \214v)100.8 562.8 Q 2.75(eV)-.165 G
(oIP TFRC connections competing with web traf)-4.169 E
(\214c in a topology)-.275 E(with a 3 Mbps shared link.)100.8 575.8 Q
(The V)5.5 E(oIP TFRC application generates 200-byte data pack)-1.419 E
(ets)-.11 E -2.365 -.275(ev e)100.8 588.8 T
(ry 10 ms, for a maximum data rate of 160 Kbps.).275 E(The \214v)5.5 E
2.75(eT)-.165 G(CP connections ha)-2.75 E -.165(ve)-.22 G
(roundtrip times from 40 to 240 ms, and the \214v)100.8 601.8 Q 2.75(eT)
-.165 G(FRC connections ha)-2.75 E .33 -.165(ve t)-.22 H(he same set of)
.165 E(round-trip times.)100.8 614.8 Q(The SA)5.5 E
(CK TCP connections in these simulations use the def)-.44 E(ault)-.11 E
(parameters in the NS simulator)100.8 627.8 Q 2.75(,w)-.44 G
(ith Limited T)-2.75 E(ransmit, and a minimum R)-.385 E .396 -.198(TO o)
-.66 H 2.75(f2).198 G(00 ms.)-2.75 E
(Adding timestamps to the TCP connection didn')100.8 640.8 Q 2.75(tc)
-.198 G(hange the results appreciably)-2.75 E 5.5(.T)-.715 G(he)-5.5 E
(simulations include re)100.8 653.8 Q -.165(ve)-.275 G(rse-path traf)
.165 E(\214c, to add some small control pack)-.275 E(ets to the forw)
-.11 E(ard)-.11 E(path, and some queueing delay to the re)100.8 666.8 Q
-.165(ve)-.275 G(rse path.).165 E(The number of web sessions is v)5.5 E
(aried)-.275 E(to create dif)100.8 679.8 Q(ferent le)-.275 E -.165(ve)
-.275 G(ls of congestion.).165 E(The DropT)5.5 E
(ail queue is in units of pack)-.88 E(ets, which)-.11 E(each pack)100.8
692.8 Q(et in the queue requires a single b)-.11 E(uf)-.22 E(fer)-.275 E
2.75(,r)-.44 G -2.475 -.165(eg a)-2.75 H(rdless of the pack).165 E
(et size.)-.11 E -.88(Ta)100.8 709.4 S(ble 6 sho).88 E(ws the a)-.275 E
-.165(ve)-.22 G(rage TCP and TFRC sending rates, each a).165 E -.165(ve)
-.22 G(raged o).165 E -.165(ve)-.165 G 2.75(rt).165 G(he \214v)-2.75 E
2.75<658d>-.165 G -.275(ow)-2.75 G(s.).275 E(As e)100.8 722.4 Q
(xpected, the V)-.165 E(oIP TCP \215o)-1.419 E(ws see similar pack)-.275
E(et drop rates as the TCP \215o)-.11 E(ws, though)-.275 E(Flo)72 769 Q
(yd/K)-.11 E 270.421(ohler Section)-.385 F 2.75(7.2. [P)2.75 F(age 14])
-.165 E 0 Cg EP
%%Page: 15 15
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E(the V)100.8 85 Q
(oIP TFRC \215o)-1.419 E(ws recei)-.275 E -.165(ve)-.275 G 2.75(sh).165
G(igher throughput than the TCP \215o)-2.75 E(ws with pack)-.275 E
(et drop rates)-.11 E(of 25% or higher)100.8 98 Q(.)-.605 E/F1 11
/Courier@0 SF 46.2(Web TCP)140.4 117 R 39.6(TCP VoIP_TFRC)46.2 F
(VoIP_TFRC)6.6 E 13.2(Sessions DropRate)127.2 130 R 19.8
(SendRate DropRate)13.2 F(SendRate)13.2 E 13.2(-------- --------)127.2
143 R 19.8(-------- --------)13.2 F(--------)13.2 E 39.6(10 0.06)153.6
156 R 39.6(239.81 0.00)33 F(185.19)33 E 39.6(25 0.09)153.6 169 R 39.6
(189.02 0.01)33 F(184.95)33 E 39.6(50 0.14)153.6 182 R 39.6(99.46 0.01)
39.6 F(185.07)33 E 39.6(100 0.20)147 195 R 39.6(16.42 0.02)39.6 F
(183.77)33 E 39.6(200 0.26 4.46 0.03)147 208 R(181.98)33 E 39.6
(400 0.29 4.61 0.05)147 221 R(151.88)33 E 39.6(800 0.49 1.01 0.08)147
234 R(113.10)33 E 39.6(1600 0.65 0.67 0.12)140.4 247 R(65.17)39.6 E
(Table 7:)100.8 273 Q
("Simulation Results with Drop-Tail Queues in Bytes.")100.8 286 Q F0(Ho)
100.8 308.6 Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, t).165 H(he f).44
E(airness results can change signi\214cantly if the DropT)-.11 E
(ail queue at the)-.88 E
(congested output link is in units of bytes rather than pack)100.8 321.6
Q 2.75(ets. F)-.11 F(or a queue in pack)-.165 E(ets, the)-.11 E
(queue has a \214x)100.8 334.6 Q(ed number of b)-.165 E(uf)-.22 E
(fers, and each b)-.275 E(uf)-.22 E(fer can hold e)-.275 E
(xactly one pack)-.165 E(et,)-.11 E(re)100.8 347.6 Q -.055(ga)-.165 G
(rdless of its size in bytes.).055 E -.165(Fo)5.5 G 2.75(raq).165 G
(ueue in bytes, the queue has a \214x)-2.75 E(ed number of)-.165 E
(*bytes*, and an almost-full queue might ha)100.8 360.6 Q .33 -.165
(ve r)-.22 H(oom for a small pack).165 E(et b)-.11 E(ut not for a lar)
-.22 E(ge)-.198 E 2.75(one. This,)100.8 373.6 R
(for a simulation with a Drop-T)2.75 E(ail queue in bytes, lar)-.88 E
(ge pack)-.198 E(ets are more lik)-.11 E(ely)-.11 E
(to be dropped than are small ones.)100.8 386.6 Q
(The NS simulator doesn')5.5 E 2.75(ty)-.198 G(et ha)-2.75 E .33 -.165
(ve a m)-.22 H(ore realistic).165 E
(intermediate model, where the queue has a \214x)100.8 399.6 Q
(ed number of b)-.165 E(uf)-.22 E(fers, each b)-.275 E(uf)-.22 E
(fer has a \214x)-.275 E(ed)-.165 E(number of bytes, and each pack)100.8
412.6 Q(et w)-.11 E(ould require one or more free b)-.11 E(uf)-.22 E
2.75(fers. In)-.275 F(this model, a)2.75 E(small pack)100.8 425.6 Q
(et w)-.11 E(ould use one b)-.11 E(uf)-.22 E(fer)-.275 E 2.75(,w)-.44 G
(hile a lar)-2.75 E(ger pack)-.198 E(et w)-.11 E(ould require se)-.11 E
-.165(ve)-.275 G(ral b).165 E(uf)-.22 E(fers.)-.275 E(As T)100.8 442.2 Q
(able 7 sho)-.88 E(ws, with a DropT)-.275 E(ail queue in bytes, the V)
-.88 E(oIP TFRC \215o)-1.419 E 2.75(ws)-.275 G(ees a much)-2.75 E
(smaller drop rate than the TCP \215o)100.8 455.2 Q 1.43 -.715(w, a)
-.275 H(nd as a consequence recei).715 E -.165(ve)-.275 G 2.75(sam).165
G(uch lar)-2.75 E(ger sending)-.198 E 2.75(rate. F)100.8 468.2 R(or e)
-.165 E(xample, when the \214v)-.165 E 2.75(eT)-.165 G(CP \215o)-2.75 E
(ws and \214v)-.275 E 2.75(eV)-.165 G(oIP TFRC \215o)-4.169 E
(ws share the link with)-.275 E(800 web sessions, the \214v)100.8 481.2
Q 2.75(eT)-.165 G(CP \215o)-2.75 E(ws see an a)-.275 E -.165(ve)-.22 G
(rage drop rate of 49% in the second half of).165 E
(the simulation, while the \214v)100.8 494.2 Q 2.75(eV)-.165 G
(oIP TFRC \215o)-4.169 E(ws recei)-.275 E .33 -.165(ve a)-.275 H 2.75
(na).165 G -.165(ve)-2.97 G(rage drop rate of 8%, and as).165 E 2.75(ac)
100.8 507.2 S(onsequence recei)-2.75 E .33 -.165(ve m)-.275 H
(ore than 100 times the throughput of the TCP \215o).165 E 2.75
(ws. This)-.275 F(raises)2.75 E
(serious questions about making the assumption that \215o)100.8 520.2 Q
(ws with small pack)-.275 E(ets see the same)-.11 E(pack)100.8 533.2 Q
(et drop rate as \215o)-.11 E(ws with lar)-.275 E(ger pack)-.198 E 2.75
(ets. Further)-.11 F -.11(wo)2.75 G(rk will ha).11 E .33 -.165(ve t)-.22
H 2.75(oi).165 G(nclude an)-2.75 E(in)100.8 546.2 Q -.165(ve)-.44 G
(stig).165 E(ation into the range of realistic Internet scenarios, in t\
erms of whether lar)-.055 E(ge pack)-.198 E(ets)-.11 E
(are considerably more lik)100.8 559.2 Q
(ely to be dropped than are small ones.)-.11 E/F2 11/Times-Bold@0 SF
(7.3.)72 585.2 Q/F3 13/Times-Bold@0 SF -.13(Pa)5.5 G(ck).13 E(et Dr)-.13
E(opping Beha)-.234 E(vior at Routers with A)-.325 E(QM)-.585 E F0(As e)
100.8 601.8 Q(xpected, the pack)-.165 E(et dropping beha)-.11 E
(vior also can be v)-.22 E(aried by v)-.275 E(arying the Acti)-.275 E
.33 -.165(ve Q)-.275 H(ueue).165 E(Management \(A)100.8 614.8 Q
(QM\) mechanism in the router)-.605 E 5.5(.W)-.605 G
(hen the routers use RED \(Random Early)-5.5 E
(Detection\), there are se)100.8 627.8 Q -.165(ve)-.275 G
(ral parameters than can af).165 E(fect the pack)-.275 E
(et dropping or marking)-.11 E(beha)100.8 640.8 Q
(vior as a function of the pack)-.22 E(et size.)-.11 E
(First, as with DropT)100.8 657.4 Q
(ail, the RED queue can be either in units of pack)-.88 E
(ets or of bytes.)-.11 E(This)5.5 E(can af)100.8 670.4 Q(fect the pack)
-.275 E(et dropping beha)-.11 E
(vior when the RED mechanism is unable to control the)-.22 E -2.475 -.22
(av e)100.8 683.4 T(rage queue size, and the queue o).22 E -.165(ve)
-.165 G(r\215o).165 E(ws.)-.275 E(Second, and orthogonally)100.8 709.4 Q
2.75(,R)-.715 G(ED can be either in pack)-2.75 E
(et mode or in byte mode.)-.11 E(In pack)5.5 E(et)-.11 E
(mode, each *pack)100.8 722.4 Q
(et* has the same probability of being dropped by RED, while in byte)
-.11 E(Flo)72 769 Q(yd/K)-.11 E 270.421(ohler Section)-.385 F 2.75
(7.3. [P)2.75 F(age 15])-.165 E 0 Cg EP
%%Page: 16 16
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E
(mode, each *byte* has the same probability of being dropped.)100.8 85 Q
(In pack)5.5 E(et mode, lar)-.11 E(ge-)-.198 E(pack)100.8 98 Q
(et and small-pack)-.11 E(et \215o)-.11 E(ws recei)-.275 E .33 -.165
(ve r)-.275 H(oughly the same pack).165 E(et drop rate, while in byte)
-.11 E(mode, with small to moderate le)100.8 111 Q -.165(ve)-.275 G
(ls of congestion, lar).165 E(ge-pack)-.198 E(et and small-pack)-.11 E
(et \215o)-.11 E(ws)-.275 E(with the same throughput in bps recei)100.8
124 Q .33 -.165(ve r)-.275 H(oughly the same *number* of pack).165 E
(et drops.)-.11 E(The)5.5 E(simulations reported in the appendix sho)
100.8 137 Q 2.75(wt)-.275 G(hat for RED in pack)-2.75 E
(et mode, the pack)-.11 E(et drop)-.11 E(rates for the V)100.8 150 Q
(oIP TFRC \215o)-1.419 E(ws are similar to those for the TCP \215o)-.275
E(ws, with a resulting)-.275 E(acceptable throughput for the V)100.8 163
Q(oIP TFRC \215o)-1.419 E 5.5(ws. This)-.275 F
(is true with the queue in pack)2.75 E(ets or)-.11 E
(in bytes, and with or without Adapti)100.8 176 Q .33 -.165(ve R)-.275 H
(ED \(discussed belo).165 E 2.75(w\). As)-.275 F(we sho)2.75 E 2.75(wb)
-.275 G(elo)-2.75 E 1.43 -.715(w, t)-.275 H(his).715 E -.11(fa)100.8 189
S(irness between TCP and V).11 E(oIP TFRC \215o)-1.419 E
(ws does not hold for RED in byte mode.)-.275 E
(The third RED parameter that af)100.8 215 Q(fects the pack)-.275 E
(et dropping or marking beha)-.11 E(vior as a)-.22 E(function of pack)
100.8 228 Q
(et size is whether the RED mechanism is using Standard RED or Adapti)
-.11 E -.165(ve)-.275 G(RED, where the dropping function is v)100.8 241
Q(aried as the pack)-.275 E(et drop rate changes.)-.11 E(The use of)5.5
E(Adapti)100.8 254 Q .33 -.165(ve R)-.275 H(ED allo).165 E
(ws the RED mechanism to function more ef)-.275 E(fecti)-.275 E -.165
(ve)-.275 G(ly in the presence of).165 E(high pack)100.8 267 Q
(et drop rates \(e.g., greater than 10%\).)-.11 E -.44(Wi)5.5 G
(thout Adapti).44 E .33 -.165(ve R)-.275 H(ED, there is a \214x).165 E
(ed)-.165 E(dropping threshold, and all arri)100.8 280 Q(ving pack)-.275
E(ets are dropped when the dropping or marking rate)-.11 E -.165(ex)
100.8 293 S(ceeds this threshold.).165 E(In contrast, with Adapti)5.5 E
.33 -.165(ve R)-.275 H(ED, the dropping function is adapted).165 E
(to accommodate these high-drop-rate re)100.8 306 Q 2.75(gimes. One)
-.165 F(consequence is that when byte mode is)2.75 E
(combined with Adapti)100.8 319 Q .33 -.165(ve R)-.275 H
(ED, the byte mode e).165 E(xtends e)-.165 E -.165(ve)-.275 G 2.75(nt)
.165 G 2.75(oh)-2.75 G(igh-drop-rate re)-2.75 E(gimes.)-.165 E
(When byte mode is used with standard RED, ho)100.8 332 Q(we)-.275 E
-.165(ve)-.275 G .88 -.44(r, t).165 H(he byte mode is no longer in use)
.44 E(when the drop rate e)100.8 345 Q(xceeds the \214x)-.165 E
(ed dropping threshold \(set by def)-.165 E(ault to 10% in the NS)-.11 E
(simulator\).)100.8 358 Q(In the simulations in this section, we e)100.8
384 Q(xplore the V)-.165 E(oIP TFRC beha)-1.419 E(vior o)-.22 E -.165
(ve)-.165 G 2.75(rs).165 G(ome of this)-2.75 E(range of scenarios.)100.8
397 Q(In this simulations, as in Section 7.2 abo)5.5 E -.165(ve)-.165 G
2.75(,t).165 G(he application for the)-2.75 E -1.419(Vo)100.8 410 S
(IP TFRC \215o)1.419 E 2.75(wu)-.275 G(ses 200-byte data pack)-2.75 E
(ets, generating 100 pack)-.11 E(ets per second.)-.11 E/F1 11/Courier@0
SF 46.2(Web TCP)140.4 442 R 39.6(TCP VoIP_TFRC)46.2 F(VoIP_TFRC)6.6 E
13.2(Sessions DropRate)127.2 455 R 19.8(SendRate DropRate)13.2 F
(SendRate)13.2 E 13.2(-------- --------)127.2 468 R 19.8
(-------- --------)13.2 F(--------)13.2 E 39.6(10 0.05)153.6 481 R 39.6
(305.76 0.04)33 F(182.82)33 E 39.6(25 0.06)153.6 494 R 39.6(224.16 0.06)
33 F(175.91)33 E 39.6(50 0.09)153.6 507 R 39.6(159.12 0.08)33 F(152.51)
33 E 39.6(100 0.13)147 520 R 39.6(90.77 0.11)39.6 F(106.13)33 E 39.6
(200 0.14)147 533 R 39.6(48.53 0.14)39.6 F(70.25)39.6 E 39.6(400 0.20)
147 546 R 39.6(22.08 0.19)39.6 F(41.50)39.6 E 39.6(800 0.27 3.55 0.25)
147 559 R(17.50)39.6 E 39.6(1600 0.42 1.87 0.34 8.81)140.4 572 R
(Table 8:)100.8 598 Q("Simulation Results with RED Queues, Packet Mode.)
100.8 611 Q(RED in packet mode, queue in packets, standard RED.")100.8
624 Q F0 -.165(Fo)100.8 646.6 S 2.75(rt).165 G(he simulations in T)-2.75
E(able 8, with a congested router with a RED queue in pack)-.88 E
(et mode,)-.11 E(the results are similar to those in T)100.8 659.6 Q
(able 6 abo)-.88 E -.165(ve)-.165 G 5.5(.T).165 G(he V)-5.5 E
(oIP TCP \215o)-1.419 E 2.75(wr)-.275 G(ecei)-2.75 E -.165(ve)-.275 G
2.75(ss).165 G(imilar)-2.75 E(pack)100.8 672.6 Q
(et drop rates as the TCP \215o)-.11 E 1.43 -.715(w, t)-.275 H
(hough it recei).715 E -.165(ve)-.275 G 2.75(sh).165 G
(igher throughput in the more)-2.75 E(congested en)100.8 685.6 Q 2.75
(vironments. The)-.44 F
(simulations are similar with the queue in bytes, and with or)2.75 E
(without Adapti)100.8 698.6 Q .33 -.165(ve R)-.275 H 2.75(ED. These).165
F(results seems generally acceptable.)2.75 E(Flo)72 769 Q(yd/K)-.11 E
270.421(ohler Section)-.385 F 2.75(7.3. [P)2.75 F(age 16])-.165 E 0 Cg
EP
%%Page: 17 17
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E/F1 11/Courier@0 SF 46.2
(Web TCP)140.4 85 R 39.6(TCP VoIP_TFRC)46.2 F(VoIP_TFRC)6.6 E 13.2
(Sessions DropRate)127.2 98 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)127.2 111 R 19.8(-------- --------)13.2 F
(--------)13.2 E 39.6(10 0.04)153.6 124 R 39.6(286.90 0.01)33 F(184.92)
33 E 39.6(25 0.07)153.6 137 R 39.6(192.67 0.02)33 F(184.06)33 E 39.6
(50 0.11)153.6 150 R 39.6(92.98 0.04)39.6 F(181.69)33 E 39.6(100 0.17)
147 163 R 39.6(57.02 0.07)39.6 F(154.88)33 E 39.6(200 0.17 9.55 0.11)147
176 R(120.83)33 E 39.6(400 0.29 7.68 0.16)147 189 R(77.38)39.6 E 39.6
(800 0.36 1.97 0.23)147 202 R(23.79)39.6 E 39.6(1600 0.51 0.87 0.30)
140.4 215 R(11.00)39.6 E(Table 9:)100.8 241 Q
("Simulation Results with RED Queues, Byte Mode.)100.8 254 Q
(RED in byte mode, queue in bytes, standard RED.")100.8 267 Q F0 -.88
(Ta)100.8 289.6 S(ble 9 sho).88 E
(ws that with a standard RED queue in byte mode, there is a some)-.275 E
(what greater)-.275 E(dif)100.8 302.6 Q(ferent between the pack)-.275 E
(et drop rates between the TCP and V)-.11 E(oIP TFRC \215o)-1.419 E(ws,)
-.275 E(particularly for lo)100.8 315.6 Q(wer pack)-.275 E
(et drop rates.)-.11 E -.165(Fo)5.5 G 2.75(rt).165 G(he simulation in T)
-2.75 E(able 9, the pack)-.88 E(et drop)-.11 E(rates for the TCP \215o)
100.8 328.6 Q
(ws can range from 1 1/2 to four times greater than the pack)-.275 E
(et drop)-.11 E(rates for the V)100.8 341.6 Q(oIP TFRC \215o)-1.419 E
2.75(ws. Ho)-.275 F(we)-.275 E -.165(ve)-.275 G .88 -.44(r, b).165 H
(ecause the V).44 E(oIP TFRC \215o)-1.419 E 2.75(wh)-.275 G
(as an upper bound)-2.75 E
(on the sending rate, its sending rate is not af)100.8 354.6 Q
(fected in the lo)-.275 E(wer pack)-.275 E(et-drop-rate re)-.11 E
(gimes;)-.165 E(its sending rate is only af)100.8 367.6 Q
(fected in the re)-.275 E(gimes with pack)-.165 E
(et drop rates of 10% or more.)-.11 E(While the sending rate for V)100.8
380.6 Q(oIP TFRC in the scenarios in T)-1.419 E(able 9 with higher pack)
-.88 E(et drop)-.11 E(rates are greater than desired, the results are s\
igni\214cantly better than that of V)100.8 393.6 Q(oIP \215o)-1.419 E
(ws)-.275 E(with no congestion control at all.)100.8 406.6 Q F1 46.2
(Web TCP)140.4 438.6 R 39.6(TCP VoIP_TFRC)46.2 F(VoIP_TFRC)6.6 E 13.2
(Sessions DropRate)127.2 451.6 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)127.2 464.6 R 19.8(-------- --------)13.2
F(--------)13.2 E 39.6(10 0.04)153.6 477.6 R 39.6(297.74 0.02)33 F
(185.06)33 E 39.6(25 0.07)153.6 490.6 R 39.6(209.42 0.03)33 F(184.06)33
E 39.6(50 0.10)153.6 503.6 R 39.6(85.34 0.04)39.6 F(182.30)33 E 39.6
(100 0.16)147 516.6 R 39.6(28.18 0.05)39.6 F(181.17)33 E 39.6
(200 0.19 4.18 0.06)147 529.6 R(177.70)33 E 39.6(400 0.31 1.87 0.08)147
542.6 R(154.40)33 E 39.6(800 0.29 0.77 0.06)147 555.6 R(170.01)33 E 39.6
(1600 0.59 0.48 0.02)140.4 568.6 R(173.91)33 E(Table 10:)100.8 594.6 Q
("Simulation Results with RED Queues, Byte Mode.)100.8 607.6 Q
(RED in byte mode, queue in bytes, Adaptive RED.")100.8 620.6 Q F0
(In contrast, for the simulations in T)100.8 643.2 Q
(able 10, the congested router uses an Adapti)-.88 E .33 -.165(ve R)
-.275 H(ED).165 E(queue in byte mode.)100.8 656.2 Q -.165(Fo)8.25 G 2.75
(rt).165 G(his router)-2.75 E 2.75(,t)-.44 G
(he output queue is in units of bytes rather than of)-2.75 E(pack)100.8
669.2 Q(ets, each *byte* is dropped with the same probability)-.11 E
2.75(,a)-.715 G(nd because of the use of)-2.75 E(Adapti)100.8 682.2 Q
.33 -.165(ve R)-.275 H(ED, this byte-dropping mode e).165 E(xtends e)
-.165 E -.165(ve)-.275 G 2.75(nf).165 G(or the high-pack)-2.75 E
(et-drop-rate)-.11 E(re)100.8 695.2 Q(gime.)-.165 E(As T)100.8 711.8 Q
(able 10 sho)-.88 E(ws, for a scenario with Adapti)-.275 E .33 -.165
(ve R)-.275 H(ED in byte mode, the pack).165 E(et drop rate)-.11 E
(for the V)100.8 724.8 Q(oIP TFRC traf)-1.419 E(\214c is *much* lo)-.275
E(wer than that for the TCP traf)-.275 E(\214c, and as a)-.275 E(Flo)72
769 Q(yd/K)-.11 E 270.421(ohler Section)-.385 F 2.75(7.3. [P)2.75 F
(age 17])-.165 E 0 Cg EP
%%Page: 18 18
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E
(consequence, the sending rate for the V)100.8 85 Q(oIP TFRC traf)-1.419
E(\214c in a highly congested en)-.275 E(vironment)-.44 E
(is *much* higher than that of the TCP traf)100.8 98 Q 2.75(\214c. In)
-.275 F -.11(fa)2.75 G(ct, in these scenarios the TFRC V).11 E(oIP)
-1.419 E(congestion control mechanisms are lar)100.8 111 Q(gely inef)
-.198 E(fecti)-.275 E .33 -.165(ve f)-.275 H(or the V).165 E(oIP traf)
-1.419 E(\214c.)-.275 E 1.76 -.88(We n)100.8 127.6 T(ote that the unf)
.88 E(airness in these simulations, in f)-.11 E -.22(avo)-.11 G 2.75(ro)
.22 G 2.75(fV)-2.75 G(oIP TFRC, is e)-4.169 E -.165(ve)-.275 G 2.75(ng)
.165 G(reater)-2.75 E(than the unf)100.8 140.6 Q(airness sho)-.11 E
(wn in T)-.275 E(able 7 for a DropT)-.88 E(ail queue in bytes.)-.88 E
(At the same time, it is)5.5 E(not kno)100.8 153.6 Q(wn if there is an)
-.275 E 2.75(yd)-.165 G(eplo)-2.75 E(yment in the Internet of an)-.11 E
2.75(yr)-.165 G(outers with Adapti)-2.75 E .33 -.165(ve R)-.275 H(ED in)
.165 E(byte mode, or of an)100.8 166.6 Q 2.75(yA)-.165 G
(QM mechanisms with similar beha)-3.355 E 2.75(vior; we)-.22 F(don')2.75
E 2.75(te)-.198 G -.165(ve)-3.025 G 2.75(nk).165 G(no)-2.75 E 2.75(wt)
-.275 G(he)-2.75 E -.165(ex)100.8 179.6 S(tent of the deplo).165 E
(yment of standard RED, or or an)-.11 E 2.75(yo)-.165 G 2.75(ft)-2.75 G
(he proposed A)-2.75 E(QM mechanisms.)-.605 E/F1 11/Times-Bold@0 SF(8.)
72 218.6 Q/F2 14/Times-Bold@0 SF(Discussion)5.5 E F0(The goal of the V)
100.8 235.2 Q(oIP v)-1.419 E(ariant of TFRC has been for the TCP \215o)
-.275 E(ws and the V)-.275 E(oIP TFRC)-1.419 E<8d6f>100.8 248.2 Q
(ws to ha)-.275 E .33 -.165(ve r)-.22 H(ough f).165 E
(airness in the sending rate in bps, in a scenario where each pack)-.11
E(et)-.11 E(recei)100.8 261.2 Q -.165(ve)-.275 G 2.75(sr).165 G
(oughly the same probability of being dropped.)-2.75 E
(In a scenario where lar)5.5 E(ge pack)-.198 E(ets)-.11 E(are more lik)
100.8 274.2 Q(ely to be dropped than small pack)-.11 E
(ets, or where \215o)-.11 E(ws with a b)-.275 E(ursty sending rate)-.22
E(are more lik)100.8 287.2 Q(ely to ha)-.11 E .33 -.165(ve p)-.22 H(ack)
.165 E(ets dropped than are \215o)-.11 E
(ws with a smooth sending rate, \215o)-.275 E(ws)-.275 E(using the V)
100.8 300.2 Q(oIP v)-1.419 E(ariant of TFRC could recei)-.275 E .33
-.165(ve m)-.275 H(ore bandwidth than competing TCP \215o).165 E(ws.)
-.275 E(The V)100.8 316.8 Q(oIP v)-1.419 E
(ariant of TFRC limits the sending rate in pack)-.275 E(et per second.)
-.11 E(The simulations by)5.5 E -.88(To)100.8 329.8 S 2.75(mP).88 G
(helan e)-2.75 E(xplore ho)-.165 E 2.75(wal)-.275 G
(imitation in sending rate complicates the issue of e)-2.75 E
(xploring the)-.165 E -.11(fa)100.8 342.8 S(irness between TCP and V).11
E(oIP TFRC \215o)-1.419 E(ws.)-.275 E -.165(Fo)100.8 359.4 S 2.75(ra)
.165 G(pplications with a maximum sending rate of 96 Kbps or less, V)
-2.75 E(oIP TFRC only)-1.419 E(restricts the sending rate when the pack)
100.8 372.4 Q(et drop rate is f)-.11 E(airly high.)-.11 E(In this re)5.5
E(gime, the)-.165 E(performance of TFRC is v)100.8 385.4 Q
(ery much determined by the accurac)-.165 E 2.75(yo)-.165 G 2.75(ft)
-2.75 G(he TCP response)-2.75 E
(function in representing the actual sending rate of a TCP connection.)
100.8 398.4 Q(In this re)5.5 E(gime of)-.165 E(high pack)100.8 411.4 Q
(et drop rates, the performance of the TCP connection is v)-.11 E
(ery much af)-.165 E(fected by)-.275 E(the TCP algorithm \(e.g., SA)
100.8 424.4 Q(CK or not\), by the minimum R)-.44 E -.198(TO)-.66 G 2.75
(,b).198 G 2.75(yt)-2.75 G(he use or not of Limited)-2.75 E -.385(Tr)
100.8 437.4 S
(ansmit, by the use of timestamps and/or of ECN, and the lik).385 E 2.75
(e. It)-.11 F(is good to insure that)2.75 E(simulations or e)100.8 450.4
Q(xperiments e)-.165 E(xploring f)-.165 E(airness include the e)-.11 E
(xploration of f)-.165 E(airness with the)-.11 E(most aggressi)100.8
463.4 Q .33 -.165(ve T)-.275 H
(CP mechanisms conformance with the current standards.).165 E(Our)5.5 E
(simulations use SA)100.8 476.4 Q(CK TCP with Limited T)-.44 E
(ransmit and with a minimum R)-.385 E .396 -.198(TO o)-.66 H 2.75(f2)
.198 G(00 ms.)-2.75 E
(Adding the use of timestamps has not made a big dif)100.8 489.4 Q 2.75
(ference. W)-.275 F 2.75(eh)-.88 G -2.475 -.22(av e)-2.75 H(n').22 E
2.75(tu)-.198 G(sed ECN,)-2.75 E(because our judgement is that the use \
of ECN is not advisable in high-pack)100.8 502.4 Q(et-dropping)-.11 E
(re)100.8 515.4 Q(gimes.)-.165 E F1(9.)72 541.4 Q F2
(Implementation Issues)5.5 E F0(TB)100.8 558 Q(A)-.385 E F1(10.)72 584 Q
F2(Security Considerations)5.5 E F0(TB)100.8 600.6 Q(A)-.385 E F1(11.)72
626.6 Q F2(IAN)5.5 E 3.5(AC)-.28 G(onsiderations)-3.5 E F0
(There are no IAN)100.8 643.2 Q 2.75(Ac)-.385 G
(onsiderations in this document.)-2.75 E F1(12.)72 669.2 Q F2(Thanks)5.5
E F0 1.76 -.88(We t)100.8 685.8 T(hank T).88 E
(om Phelan for discussions of the V)-.88 E(oIP v)-1.419 E
(ariant of TFRC and for his paper)-.275 E -.165(ex)100.8 698.8 S
(ploring the f).165 E(airness between TCP and V)-.11 E(oIP TFRC \215o)
-1.419 E 2.75(ws. W)-.275 F 2.75(et)-.88 G(hank Joer)-2.75 E 2.75(gW)
-.198 G(idmer for)-3.19 E(feedback on earlier v)100.8 711.8 Q
(ersions of this draft.)-.165 E 1.76 -.88(We a)5.5 H
(lso thank the DCCP W).88 E(orking Group for)-.88 E
(feedback and discussions.)100.8 724.8 Q(Flo)72 769 Q(yd/K)-.11 E
273.171(ohler Section)-.385 F 2.75(12. [P)2.75 F(age 18])-.165 E 0 Cg EP
%%Page: 19 19
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E/F1 14/Times-Bold@0 SF(Normati)
72 85 Q .28 -.14(ve R)-.14 H(efer).14 E(ences)-.252 E F0
([RFC 2119] S. Bradner)100.8 101.6 Q 2.75(.K)-.605 G .33 -.165(ey W)
-3.025 H(ords F)-.715 E(or Use in RFCs to Indicate Requirement Le)-.165
E -.165(ve)-.275 G(ls. RFC).165 E(2119.)129.6 114.6 Q([RFC 2434] T)100.8
131.2 Q 2.75(.N)-.814 G(arten and H. Alv)-2.75 E 2.75
(estrand. Guidelines)-.165 F(for Writing an IAN)2.75 E 2.75(AC)-.385 G
(onsiderations)-2.75 E(Section in RFCs.)129.6 144.2 Q(RFC 2434.)5.5 E
([RFC 2581] M. Allman, V)100.8 160.8 Q 2.75(.P)-1.419 G(axson, and W)
-2.915 E 2.75(.S)-1.012 G(te)-2.75 E -.165(ve)-.275 G 2.75(ns. TCP).165
F(Congestion Control.)2.75 E(RFC 2581.)5.5 E([RFC 3448] M. Handle)100.8
177.4 Q 1.43 -.715(y, S)-.165 H 2.75(.F).715 G(lo)-2.75 E(yd, J. P)-.11
E(adh)-.165 E(ye, and J. W)-.055 E(idmer)-.44 E 2.75(,T)-.44 G
(CP Friendly Rate Control)-2.75 E(\(TFRC\): Protocol Speci\214cation, R\
FC 3448, Proposed Standard, January 2003.)129.6 190.4 Q F1(Inf)72 216.4
Q(ormati)-.35 E .28 -.14(ve R)-.14 H(efer).14 E(ences)-.252 E F0
([CCID 3 PR)100.8 233 Q(OFILE] S. Flo)-.44 E(yd, E. K)-.11 E(ohler)-.385
E 2.75(,a)-.44 G(nd J. P)-2.75 E(adh)-.165 E 2.75(ye. Pro\214le)-.055 F
(for DCCP Congestion)2.75 E(Control ID 3: TFRC Congestion Control.)129.6
246 Q(draft-ietf-dccp-ccid3-06.txt, w)5.5 E(ork in)-.11 E
(progress, October 2004.)129.6 259 Q([DCCP] E. K)100.8 275.6 Q(ohler)
-.385 E 2.75(,M)-.44 G 2.75(.H)-2.75 G(andle)-2.75 E 1.43 -.715(y, a)
-.165 H(nd S. Flo).715 E 2.75(yd. Datagram)-.11 F
(Congestion Control Protocol,)2.75 E(draft-ietf-dccp-spec-08.txt, w)
129.6 288.6 Q(ork in progress, October 2004.)-.11 E([JF)100.8 305.2 Q
(AS05] A. Jain, S. Flo)-.814 E(yd, M. Allman, and P)-.11 E 2.75(.S)
-1.221 G 2.75(arolahti. Quick-Start)-2.75 F(for TCP and IP)2.75 E(.)
-1.221 E(Internet-draft draft-amit-quick-start-04.txt, w)129.6 318.2 Q
(ork in progress, February 2004.)-.11 E
([MAF04] A. Medina, M. Allman, and A. Flo)100.8 334.8 Q
(yd, Measuring the Ev)-.11 E(olution of T)-.22 E(ransport)-.385 E
(Protocols in the Internet, May 2004, URL "http://www)129.6 347.8 Q
(.icir)-.715 E(.or)-.605 E(g/tbit/".)-.198 E([P04] T)100.8 364.4 Q 2.75
(.P)-.814 G(helan, TFRC with Self-Limiting Sources, October 2004.)-2.75
E(URL)5.5 E("http://www)129.6 377.4 Q(.phelan-4.com/dccp/".)-.715 E
([RFC 2861] M. Handle)100.8 394 Q 1.43 -.715(y, J)-.165 H 2.75(.P).715 G
(adh)-2.915 E(ye, and S. Flo)-.055 E 2.75(yd. TCP)-.11 F(Congestion W)
2.75 E(indo)-.44 E 2.75(wV)-.275 G(alidation.)-3.971 E
(RFC 2861, June 2000.)129.6 407 Q([RFC 3714] S. Flo)100.8 423.6 Q
(yd and J. K)-.11 E(empf, Editors.)-.275 E(IAB Concerns Re)5.5 E -.055
(ga)-.165 G(rding Congestion Control).055 E(for V)129.6 436.6 Q(oice T)
-1.419 E(raf)-.385 E(\214c in the Internet.)-.275 E(RFC 3714.)5.5 E
([V00] P)100.8 453.2 Q 2.75(.V)-1.221 G 2.75(asallo. V)-3.971 F
(ariable P)-1.221 E(ack)-.165 E
(et Size Equation-Based Congestion Control.)-.11 E(ICSI)5.5 E -.77(Te)
129.6 466.2 S(chnical Report TR-00-008, April 2000.).77 E(URL)5.5 E
("http://www)129.6 479.2 Q(.icsi.berk)-.715 E(ele)-.11 E -.715(y.)-.165
G(edu/techreports/2000.abstracts/ tr).715 E(-00-008.html".)-.22 E
([WBL04] J. W)100.8 508.8 Q(idmer)-.44 E 2.75(,C)-.44 G 2.75(.B)-2.75 G
(outremans, and Jean-Yv)-2.75 E(es Le Boudec.)-.165 E
(Congestion Control for)5.5 E(Flo)129.6 521.8 Q(ws with V)-.275 E
(ariable P)-1.221 E(ack)-.165 E(et Size, A)-.11 E
(CM CCR, 34\(2\), 2004.)-.44 E F1 -.7(Au)72 547.8 S(thors' Addr).7 E
(esses)-.252 E F0(Sally Flo)100.8 564.4 Q(yd <\215o)-.11 E(yd@icir)-.11
E(.or)-.605 E(g>)-.198 E(ICSI Center for Internet Research)100.8 577.4 Q
(1947 Center Street, Suite 600)100.8 590.4 Q(Berk)100.8 603.4 Q(ele)-.11
E 1.43 -.715(y, C)-.165 H 2.75(A9).715 G(4704)-2.75 E(USA)100.8 616.4 Q
(Eddie K)100.8 633 Q(ohler <k)-.385 E(ohler@cs.ucla.edu>)-.11 E
(4531C Boelter Hall)100.8 646 Q(UCLA Computer Science Department)100.8
659 Q(Los Angeles, CA 90095)100.8 672 Q(USA)100.8 685 Q(Flo)72 769 Q
(yd/K)-.11 E 328.171(ohler [P)-.385 F(age 19])-.165 E 0 Cg EP
%%Page: 20 20
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E/F1 11/Times-Bold@0 SF(13.)72 85
Q/F2 14/Times-Bold@0 SF(Related W)5.5 E(ork on V)-1.05 E(oIP V)-1.4 E
(ariants of TFRC)-1.288 E F0(Other proposals for v)100.8 111 Q
(ariants of TFRC for applications with v)-.275 E(ariable pack)-.275 E
(et sizes include)-.11 E([WBL04] and [V00]. [V00] proposed that TFRC sh\
ould be modi\214ed so that \215o)100.8 124 Q(ws are not)-.275 E
(penalized by sending smaller pack)100.8 137 Q 2.75(ets. In)-.11 F
(particular)2.75 E 2.75(,[)-.44 G(V00] proposes using the path MTU in)
-2.75 E(the TCP-friendly equation, instead of the actual pack)100.8 150
Q(et size used by TFRC, and counting)-.11 E(the pack)100.8 163 Q
(et drop rate by using an estimation algorithm that aggre)-.11 E -.055
(ga)-.165 G(tes both pack).055 E(et drops and)-.11 E(recei)100.8 176 Q
-.165(ve)-.275 G 2.75(dp).165 G(ack)-2.75 E(ets into MTU-sized units.)
-.11 E([WBL04] also ar)100.8 202 Q(gues that adapting TFRC for v)-.198 E
(ariable pack)-.275 E(et sizes by just using the pack)-.11 E(et)-.11 E
(size of a reference TCP \215o)100.8 215 Q 2.75(wi)-.275 G 2.75(nt)-2.75
G(he TFRC equation w)-2.75 E(ould not suf)-.11 E
(\214ce in the high-pack)-.275 E(et-)-.11 E(loss re)100.8 228 Q
(gime; such a modi\214ed TFRC w)-.165 E(ould ha)-.11 E .33 -.165(ve a s)
-.22 H(trong bias in f).165 E -.22(avo)-.11 G 2.75(ro).22 G 2.75(fs)
-2.75 G(maller pack)-2.75 E(ets,)-.11 E(because multiple lost pack)100.8
241 Q(ets in a single round-trip time w)-.11 E(ould be aggre)-.11 E
-.055(ga)-.165 G(ted into a single).055 E(pack)100.8 254 Q(et loss.)-.11
E([WBL04] proposes modifying the loss measurement process to account fo\
r the)5.5 E(bias in f)100.8 267 Q -.22(avo)-.11 G 2.75(ro).22 G 2.75(fs)
-2.75 G(maller pack)-2.75 E(ets.)-.11 E(The V)100.8 293 Q(oIP V)-1.419 E
(ariant proposed in our document dif)-1.221 E
(fers from [WBL04] in restricting its)-.275 E(attention to \215o)100.8
306 Q(ws that send at most 100 pack)-.275 E(ets per second.)-.11 E
(The V)5.5 E(oIP V)-1.419 E(ariant proposed in)-1.221 E
(our document also dif)100.8 319 Q(fers from the stra)-.275 E 2.75(wp)
-.165 G(roposal discussed in [WBL04] in that the)-2.75 E(allo)100.8 332
Q(wed bandwidth includes the bandwidth used by pack)-.275 E(et headers.)
-.11 E([WBL04] discusses four methods for modifying the loss measuremen\
t process, "unbiasing,)100.8 358 Q("virtual pack)100.8 371 Q
(ets", "random sampling", and "Loss Insensiti)-.11 E .33 -.165(ve P)
-.275 H(eriod \(LIP\) scaling".).165 E
([WBL04] \214nds only the second and third methods suf)100.8 384 Q
(\214ciently rob)-.275 E(ust when the netw)-.22 E(ork)-.11 E(drops pack)
100.8 397 Q(ets independently of pack)-.11 E(et size.)-.11 E(The)5.5 E
2.75<798c>-.165 G(nd only the second method suf)-2.75 E(\214ciently)
-.275 E(rob)100.8 410 Q(ust when the netw)-.22 E(ork is more lik)-.11 E
(ely to drop lar)-.11 E(ge pack)-.198 E(ets than small pack)-.11 E 2.75
(ets. Our)-.11 F(method for calculating the loss e)100.8 423 Q -.165(ve)
-.275 G(nt rate is some).165 E(what similar to the random sampling)-.275
E(method proposed in [WBL04], e)100.8 436 Q
(xcept that randomization is not used; instead of)-.165 E
(randomization, the e)100.8 449 Q(xact pack)-.165 E
(et loss rate is computed for short loss interv)-.11 E(als, and the)
-.275 E(standard loss e)100.8 462 Q -.165(ve)-.275 G
(nt rate calculation is used for longer loss interv).165 E 2.75
(als. [WBL04])-.275 F(includes)2.75 E(simulations with a Bernoulli loss\
 model, a Bernoulli loss model with a drop rate v)100.8 475 Q(arying)
-.275 E -.165(ove)100.8 488 S 2.75(rt).165 G(ime, and a Gilbert loss mo\
del, as well as more realistic simulations with a range of)-2.75 E
(TCP and TFRC \215o)100.8 501 Q(ws.)-.275 E
([WBL04] produces both a byte-mode and a pack)100.8 527 Q(et-mode v)-.11
E(ariant of the TFRC transport)-.275 E(protocol, for connections o)100.8
540 Q -.165(ve)-.165 G 2.75(rp).165 G(aths with per)-2.75 E
(-byte and per)-.22 E(-pack)-.22 E(et dropping respecti)-.11 E -.165(ve)
-.275 G(ly).165 E(.)-.715 E 1.76 -.88(We w)100.8 553 T(ould ar).77 E
(gue that in the absence of transport-le)-.198 E -.165(ve)-.275 G 2.75
(lm).165 G(echanisms for determining whether)-2.75 E(the pack)100.8 566
Q(et dropping in the netw)-.11 E(ork is per)-.11 E(-pack)-.22 E(et, per)
-.11 E(-byte, or some)-.22 E(where in between, a)-.275 E
(single TFRC implementation is needed, independently of the pack)100.8
579 Q(et-dropping beha)-.11 E(viors of)-.22 E
(the routers along the path.)100.8 592 Q(It w)5.5 E
(ould of course be preferable to ha)-.11 E .33 -.165(ve a m)-.22 H
(echanism that gi).165 E -.165(ve)-.275 G(s).165 E
(roughly acceptable beha)100.8 605 Q(vior)-.22 E 2.75(,t)-.44 G 2.75(ot)
-2.75 G(he connection and to the netw)-2.75 E(ork as a whole, on paths)
-.11 E(with both per)100.8 618 Q(-byte and per)-.22 E(-pack)-.22 E
(et dropping \(and on paths with multiple congested routers,)-.11 E
(some with per)100.8 631 Q(-byte dropping mechanisms, some with per)-.22
E(-pack)-.22 E(et dropping mechanisms,)-.11 E
(and some with dropping mechanisms that lie some)100.8 644 Q
(where between per)-.275 E(-byte and per)-.22 E(-pack)-.22 E(et\).)-.11
E 2.75<418c>100.8 657 S(rst step will be to in)-2.75 E -.165(ve)-.44 G
(stig).165 E(ate the range of beha)-.055 E
(viors actually present in today')-.22 E(s)-.605 E(netw)100.8 670 Q
(orks.)-.11 E(Flo)72 769 Q(yd/K)-.11 E 273.171(ohler Section)-.385 F
2.75(13. [P)2.75 F(age 20])-.165 E 0 Cg EP
%%Page: 21 21
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E/F1 11/Times-Bold@0 SF(14.)72 85
Q/F2 14/Times-Bold@0 SF(Simulation Results with RED Queue Management)5.5
E/F3 11/Courier@0 SF 46.2(Web TCP)140.4 117 R 39.6(TCP VoIP_TFRC)46.2 F
(VoIP_TFRC)6.6 E 13.2(Sessions DropRate)127.2 130 R 19.8
(SendRate DropRate)13.2 F(SendRate)13.2 E 13.2(-------- --------)127.2
143 R 19.8(-------- --------)13.2 F(--------)13.2 E 39.6(10 0.04)153.6
156 R 39.6(299.95 0.04)33 F(184.15)33 E 39.6(25 0.06)153.6 169 R 39.6
(234.00 0.06)33 F(176.43)33 E 39.6(50 0.08)153.6 182 R 39.6(172.42 0.08)
33 F(147.82)33 E 39.6(100 0.11)147 195 R 39.6(110.83 0.12)33 F(89.78)
39.6 E 39.6(200 0.13)147 208 R 39.6(41.90 0.15)39.6 F(62.37)39.6 E 39.6
(400 0.23)147 221 R 39.6(11.38 0.20)39.6 F(25.05)39.6 E 39.6
(800 0.27 2.88 0.29)147 234 R(12.26)39.6 E 39.6
(1600 0.34 0.62 0.35 6.78)140.4 247 R(Table 11:)100.8 273 Q
("Simulation Results with RED Queues, Packet Mode.)100.8 286 Q
(RED in packet mode, queue in packets, Adaptive RED.")100.8 299 Q 46.2
(Web TCP)140.4 350 R 39.6(TCP VoIP_TFRC)46.2 F(VoIP_TFRC)6.6 E 13.2
(Sessions DropRate)127.2 363 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)127.2 376 R 19.8(-------- --------)13.2 F
(--------)13.2 E 39.6(10 0.04)153.6 389 R 39.6(319.87 0.03)33 F(183.32)
33 E 39.6(25 0.05)153.6 402 R 39.6(246.38 0.05)33 F(175.27)33 E 39.6
(50 0.08)153.6 415 R 39.6(148.18 0.07)33 F(152.02)33 E 39.6(100 0.12)147
428 R 39.6(96.38 0.11)39.6 F(92.87)39.6 E 39.6(200 0.18)147 441 R 39.6
(43.30 0.17)39.6 F(55.87)39.6 E 39.6(400 0.24)147 454 R 39.6(12.58 0.21)
39.6 F(20.25)39.6 E 39.6(800 0.30 4.22 0.27)147 467 R(20.08)39.6 E 39.6
(1600 0.35 1.34 0.35 7.21)140.4 480 R(Table 12:)100.8 506 Q
("Simulation Results with RED Queues, Packet Mode.)100.8 519 Q
(RED in packet mode, queue in bytes, Adaptive RED.")100.8 532 Q F0(Flo)
72 769 Q(yd/K)-.11 E 273.171(ohler Section)-.385 F 2.75(14. [P)2.75 F
(age 21])-.165 E 0 Cg EP
%%Page: 22 22
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E/F1 11/Courier@0 SF 46.2
(Web TCP)140.4 85 R 39.6(TCP VoIP_TFRC)46.2 F(VoIP_TFRC)6.6 E 13.2
(Sessions DropRate)127.2 98 R 19.8(SendRate DropRate)13.2 F(SendRate)
13.2 E 13.2(-------- --------)127.2 111 R 19.8(-------- --------)13.2 F
(--------)13.2 E 39.6(10 0.04)153.6 124 R 39.6(326.74 0.03)33 F(184.79)
33 E 39.6(25 0.05)153.6 137 R 39.6(260.50 0.04)33 F(180.28)33 E 39.6
(50 0.08)153.6 150 R 39.6(173.62 0.07)33 F(154.04)33 E 39.6(100 0.11)147
163 R 39.6(81.94 0.10)39.6 F(103.87)33 E 39.6(200 0.16)147 176 R 39.6
(48.10 0.14)39.6 F(60.50)39.6 E 39.6(400 0.25 8.64 0.19)147 189 R(32.55)
39.6 E 39.6(800 0.35 7.68 0.26)147 202 R(14.62)39.6 E 39.6
(1600 0.38 0.77 0.36 7.13)140.4 215 R(Table 13:)100.8 241 Q
("Simulation Results with RED Queues, Packet Mode.)100.8 254 Q
(RED in packet mode, queue in bytes, standard RED.")100.8 267 Q/F2 11
/Times-Bold@0 SF(15.)72 312 Q/F3 14/Times-Bold@0 SF 3.5(AD)5.5 G
(iscussion of P)-3.5 E(ack)-.14 E(et Size and P)-.14 E(ack)-.14 E(et Dr)
-.14 E(opping)-.252 E F0(This section gi)100.8 338 Q -.165(ve)-.275 G
2.75(sag).165 G(eneral summary of the relati)-2.75 E .33 -.165(ve a)
-.275 H(dv).165 E(antages of pack)-.275 E(et-dropping)-.11 E(beha)100.8
351 Q(vior independent of pack)-.22 E(et size, v)-.11 E(ersus pack)-.165
E(et dropping beha)-.11 E(vior where lar)-.22 E(ge pack)-.198 E(ets)-.11
E(are more lik)100.8 364 Q(ely to be dropped than small ones.)-.11 E F1
(Advantages of Packet Dropping Independent of Packet Size:)100.8 396 Q
(---------------------------------------------------------)100.8 409 Q
6.6(1. Adds)100.8 422 R
(another incentive for end nodes to use large packets.)6.6 E 6.6
(2. Matches)100.8 435 R
(an environment with a limitation in pps rather than)6.6 E(bps.)127.2
448 Q(---------------------------------------------------------)100.8
461 Q(Advantages of Packet Dropping as a Function of Packet Size:)100.8
487 Q(---------------------------------------------------------)100.8
500 Q 6.6(1. Small)100.8 513 R
(control packets are less likely to be dropped than are)6.6 E
(large data packets, improving TCP performance.)127.2 526 Q 6.6
(2. Matches)100.8 539 R
(an environment with a limitation in bps rather than)6.6 E(pps.)127.2
552 Q 6.6(3. Reduces)100.8 565 R
(the penalty of TCP and other transport protocols)6.6 E
(against flows with small packets \(where the allowed sending)127.2 578
Q(rate is roughly a linear function of packet size\).)127.2 591 Q 6.6
(4. A)100.8 604 R
(queue limited in bytes is better than a queue limited in)6.6 E
(packets for matching the worst-case queue size to the)127.2 617 Q
(worst-case queueing delay in seconds.)127.2 630 Q
(---------------------------------------------------------)100.8 643 Q
F3(Full Copyright Statement)72 688 Q F0(Cop)100.8 704.6 Q
(yright \(C\) The Internet Society 2005.)-.11 E
(This document is subject to the rights, licenses)5.5 E
(and restrictions contained in BCP 78, and e)100.8 717.6 Q
(xcept as set forth therein, the authors retain all)-.165 E(Flo)72 769 Q
(yd/K)-.11 E 328.171(ohler [P)-.385 F(age 22])-.165 E 0 Cg EP
%%Page: 23 23
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 71.587(-DRAFT Expires:)-1.012 F
(January 2006)2.75 E(July 2005)123.726 E(their rights.)100.8 85 Q
(This document and the information contained herein are pro)100.8 101.6
Q(vided on an "AS IS" basis and)-.165 E(THE CONTRIB)100.8 114.6 Q(UT)
-.11 E(OR, THE ORGANIZA)-.198 E(TION HE/SHE REPRESENTS OR IS)-1.221 E
(SPONSORED BY \(IF ANY\), THE INTERNET SOCIETY AND THE INTERNET)100.8
127.6 Q(ENGINEERING T)100.8 140.6 Q(ASK FORCE DISCLAIM ALL W)-1.023 E
(ARRANTIES, EXPRESS OR)-1.32 E(IMPLIED, INCLUDING B)100.8 153.6 Q(UT NO)
-.11 E 2.75(TL)-.44 G(IMITED T)-2.75 E 2.75(OA)-.198 G(NY W)-2.75 E
(ARRANTY THA)-1.32 E 2.75(TT)-1.221 G(HE USE)-2.75 E(OF THE INFORMA)
100.8 166.6 Q(TION HEREIN WILL NO)-1.221 E 2.75(TI)-.44 G
(NFRINGE ANY RIGHTS OR ANY)-2.75 E(IMPLIED W)100.8 179.6 Q
(ARRANTIES OF MERCHANT)-1.32 E(ABILITY OR FITNESS FOR A)-1.023 E -1.012
(PA)100.8 192.6 S -.66(RT)1.012 G(ICULAR PURPOSE.).66 E/F1 14
/Times-Bold@0 SF(Intellectual Pr)72 218.6 Q(operty)-.252 E F0
(The IETF tak)100.8 235.2 Q(es no position re)-.11 E -.055(ga)-.165 G
(rding the v).055 E(alidity or scope of an)-.275 E 2.75(yI)-.165 G
(ntellectual Property)-2.75 E(Rights or other rights that might be clai\
med to pertain to the implementation or use of the)100.8 248.2 Q
(technology described in this document or the e)100.8 261.2 Q
(xtent to which an)-.165 E 2.75(yl)-.165 G(icense under such rights)
-2.75 E(might or might not be a)100.8 274.2 Q -.275(va)-.22 G
(ilable; nor does it represent that it has made an).275 E 2.75(yi)-.165
G(ndependent)-2.75 E(ef)100.8 287.2 Q(fort to identify an)-.275 E 2.75
(ys)-.165 G(uch rights.)-2.75 E
(Information on the procedures with respect to rights in)5.5 E
(RFC documents can be found in BCP 78 and BCP 79.)100.8 300.2 Q
(Copies of IPR disclosures made to the IETF Secretariat and an)100.8
316.8 Q 2.75(ya)-.165 G(ssurances of licenses to)-2.75 E(be made a)100.8
329.8 Q -.275(va)-.22 G(ilable, or the result of an attempt made to obt\
ain a general license or).275 E(permission for the use of such propriet\
ary rights by implementers or users of this)100.8 342.8 Q(speci\214cati\
on can be obtained from the IETF on-line IPR repository at)100.8 355.8 Q
(http://www)100.8 368.8 Q(.ietf.or)-.715 E(g/ipr)-.198 E(.)-.605 E
(The IETF in)100.8 385.4 Q(vites an)-.44 E 2.75(yi)-.165 G
(nterested party to bring to its attention an)-2.75 E 2.75(yc)-.165 G
(op)-2.75 E(yrights, patents or)-.11 E
(patent applications, or other proprietary rights that may co)100.8
398.4 Q -.165(ve)-.165 G 2.75(rt).165 G(echnology that may be)-2.75 E
(required to implement this standard.)100.8 411.4 Q
(Please address the information to the IETF at ietf-)5.5 E(ipr@ietf.or)
100.8 424.4 Q(g.)-.198 E(Flo)72 769 Q(yd/K)-.11 E 328.171(ohler [P)-.385
F(age 23])-.165 E 0 Cg EP
%%Trailer
end
%%EOF