One document matched: draft-ietf-dccp-tfrc-faster-restart-04.ps

%!PS-Adobe-3.0
%%Creator: groff version 1.18.1
%%CreationDate: Thu Sep 27 10:51:26 2007
%%DocumentNeededResources: font Courier-Bold
%%+ font Times-Bold
%%+ font Times-Roman
%%+ font Courier
%%DocumentSuppliedResources: procset grops 1.18 1
%%Pages: 17
%%PageOrder: Ascend
%%Orientation: Portrait
%%EndComments
%%BeginProlog
%%BeginResource: procset grops 1.18 1
/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
/Fk{
setcmykcolor fill
}bind def
/Fg{
setgray fill
}bind def
/FL/fill load def
/LW/setlinewidth load def
/Cr/setrgbcolor load def
/Ck/setcmykcolor load def
/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
}bind def
/PEND{
clear
countdictstack CNT sub{end}repeat
level1 restore
}bind def
end def
/setpacking where{
pop
setpacking
}if
%%EndResource
%%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 841.89 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
(E. Kohler)192 E 318(INTERNET-DRAFT UCLA)72 98 R
(Intended status: Experimental)72 111 Q(S. Floyd)210 E
(Expires: March 2008)72 124 Q(ICIR)294 E(A. Sathiaseelan)414 137 Q
(University of Aberdeen)372 150 Q(27 September 2007)402 163 Q/F1 14
/Times-Bold@0 SF -.35(Fa)125.376 188 S(ster Restart f).35 E
(or TCP Friendly Rate Contr)-.35 E(ol \(TFRC\))-.252 E
(draft-ietf-dccp-tfr)164.25 201 Q(c-faster)-.252 E(-r)-.518 E
(estart-04.ps)-.252 E/F2 11/Times-Bold@0 SF(Status of this Memo)72 233 Q
/F3 11/Times-Roman@0 SF
(By submitting this Internet-Draft, each author represents that an)93.6
249.6 Q 2.75(ya)-.165 G(pplicable patent or other)-2.75 E
(IPR claims of which he or she is a)93.6 262.6 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 2.75(fw)-2.75 G(hich he)-2.75 E(or she becomes a)93.6 275.6
Q -.11(wa)-.165 G
(re will be disclosed, in accordance with Section 6 of BCP 79.).11 E
(Internet-Drafts are w)93.6 292.2 Q
(orking documents of the Internet Engineering T)-.11 E(ask F)-.88 E
(orce \(IETF\), its)-.165 E(areas, and its w)93.6 305.2 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.)93.6 318.2 Q
(Internet-Drafts are draft documents v)93.6 334.8 Q
(alid for a maximum of six months and may be updated,)-.275 E
(replaced, or obsoleted by other documents at an)93.6 347.8 Q 2.75(yt)
-.165 G 2.75(ime. It)-2.75 F(is inappropriate to use Internet-)2.75 E
(Drafts as reference material or to cite them other than as "w)93.6
360.8 Q(ork in progress.")-.11 E
(The list of current Internet-Drafts can be accessed at http://www)93.6
377.4 Q(.ietf.or)-.715 E(g/ietf/1id-)-.198 E(abstracts.txt.)93.6 390.4 Q
(The list of Internet-Draft Shado)93.6 407 Q 2.75(wD)-.275 G
(irectories can be accessed at)-2.75 E(http://www)93.6 420 Q(.ietf.or)
-.715 E(g/shado)-.198 E -.715(w.)-.275 G(html.).715 E
(This Internet-Draft will e)93.6 436.6 Q(xpire on March 2008.)-.165 E F2
(Copyright Notice)72 462.6 Q F3(Cop)93.6 479.2 Q
(yright \(C\) The IETF T)-.11 E(rust \(2007\).)-.385 E F2(Abstract)72
505.2 Q F3(TCP-Friendly Rate Control \(TFRC\) is a congestion control m\
echanism for unicast \215o)93.6 521.8 Q(ws)-.275 E
(operating in a best-ef)93.6 534.8 Q(fort Internet en)-.275 E 2.75
(vironment. This)-.44 F(document introduces F)2.75 E(aster Restart, an)
-.165 E(optional mechanism for safely impro)93.6 547.8 Q(ving the beha)
-.165 E(vior of interacti)-.22 E .33 -.165(ve \215)-.275 H -.275(ow).165
G 2.75(st).275 G(hat use TFRC.)-2.75 E -.165(Fa)93.6 560.8 S
(ster Restart is proposed for use with TFRC and with TFRC-SP).165 E 2.75
(,t)-1.221 G(he Small P)-2.75 E(ack)-.165 E(et v)-.11 E(ariant of)-.275
E 2.75(TFRC. W)93.6 573.8 R 2.75(ep)-.88 G(resent F)-2.75 E
(aster Restart in general terms as a congestion control mechanism, and)
-.165 E(further describe ho)93.6 586.8 Q 2.75(wt)-.275 G 2.75(oi)-2.75 G
(mplement F)-2.75 E
(aster Restart in Datagram Congestion Control Protocol)-.165 E
(\(DCCP\) Congestion Control IDs 3 and 4.)93.6 599.8 Q -.385(Ko)72 733 S
(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G 335.849(l. [P)-2.75 F
(age 1])-.165 E 0 Cg EP
%%Page: 2 2
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E/F1 11/Times-Bold@0 SF -1.012
(Ta)72 85 S(ble of Contents)1.012 E/F2 10/Times-Roman@0 SF
(1. Introduction)93.6 110 Q F0 -2.25(..................................\
.......................................................................\
............... ....)4.46 F F2(4)A(2. Con)93.6 122 Q -.15(ve)-.4 G
(ntions).15 E F0 -2.25(................................................\
.......................................................................\
. ....)3.89 F F2(7)A(3. F)93.6 134 Q(aster Restart: Changes to TFRC)-.15
E F0 -2.25(............................................................\
.............................. ....)3.77 F F2(7)A(3.1. Feedback P)108.6
146 Q(ack)-.15 E(ets)-.1 E F0 -2.25(...................................\
..................................................................... .\
...)4.28 F F2(7)A(3.2. Nofeedback T)108.6 158 Q(imer)-.35 E F0 -1.75(..\
.......................................................................\
.............................. ..)3.25 F F2(10)A(4. F)93.6 170 Q
(aster Restart: DCCP-speci\214c Speci\214cations)-.15 E F0 -1.75(......\
................................................................... ..)
4.14 F F2(10)A(4.1. DCCP: Recei)108.6 182 Q .3 -.15(ve R)-.25 H
(ate Adjustment).15 E F0 -1.75(........................................\
.......................................... ..)2.69 F F2(10)A
(4.2. DCCP: The Recei)108.6 194 Q .3 -.15(ve R)-.25 H(ate Length).15 E
F0 -1.75(..............................................................\
.................... ..)2.98 F F2(11)A(5. F)93.6 206 Q
(aster Restart Discussion)-.15 E F0 -1.75(.............................\
.......................................................................\
. ..)4.64 F F2(11)A(5.1. W)108.6 218 Q(orst-Case Scenarios)-.8 E F0
-1.75(.................................................................\
................................. ..)4.11 F F2(12)A(5.2. Incenti)108.6
230 Q -.15(ve)-.25 G 2.5(sf).15 G
(or applications to send unnecessary pack)-2.5 E
(ets during idle or data-limited periods?)-.1 E F0(...)4.94 E F2(12)A
(5.3. F)108.6 242 Q(aster Restart for TFRC-SP)-.15 E F0 -1.75(.........\
.......................................................................\
.......... ..)2.66 F F2(13)A(6. Simulations of F)93.6 254 Q
(aster Restart)-.15 E F0 -1.75(........................................\
........................................................ ..)3.66 F F2
(13)A(7. Implementation Issues)93.6 266 Q F0 -1.75(....................\
.......................................................................\
.............. ..)4.88 F F2(13)A(8. Security Considerations)93.6 278 Q
F0 -1.75(..............................................................\
......................................... ..)4.82 F F2(13)A(9. IAN)93.6
290 Q 2.5(AC)-.35 G(onsiderations)-2.5 E F0 -1.75(.....................\
.......................................................................\
............... ..)2.51 F F2(13)A(10. Thanks)93.6 302 Q F0 -1.75(......\
.......................................................................\
................................................. ..)2.96 F F2(14)A
(Normati)93.6 314 Q .3 -.15(ve R)-.25 H(eferences).15 E F0 -1.75(......\
.......................................................................\
................................ ..)4.86 F F2(14)A(Informati)93.6 326 Q
.3 -.15(ve R)-.25 H(eferences).15 E F0 -1.75(..........................\
.......................................................................\
........... ..)3.17 F F2(14)A(A. Appendix: Simulations)93.6 338 Q F0
-1.75(.................................................................\
....................................... ..)3.17 F F2(15)A
(Authors' Addresses)93.6 350 Q F0 -1.75(...............................\
.......................................................................\
........... ..)4 F F2(16)A(Full Cop)93.6 362 Q(yright Statement)-.1 E F0
-1.75(.................................................................\
........................................ ..)3.57 F F2(17)A
(Intellectual Property)93.6 374 Q F0 -1.75(............................\
.......................................................................\
............. ..)3.97 F F2(17)A F0 -.385(Ko)72 733 S(hler).385 E 2.75
(,e)-.44 G 2.75(ta)-2.75 G 335.849(l. [P)-2.75 F(age 2])-.165 E 0 Cg EP
%%Page: 3 3
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E(NO)93.6 85 Q(TE T)-.44 E 2.75
(OR)-.198 G(FC EDIT)-2.75 E(OR: PLEASE DELETE THIS NO)-.198 E
(TE UPON PUBLICA)-.44 E(TION.)-1.221 E
(Changes from draft-ietf-dccp-tfrc-f)93.6 111 Q(aster)-.11 E
(-restart-03.txt:)-.22 E 2.75(*D)93.6 137 S(eleted ping pack)-2.75 E
(ets, and the section about the implementation)-.11 E(of ping pack)99.1
150 Q(ets in DCCP)-.11 E(.)-1.221 E 2.75(*I)93.6 176 S 2.75(nS)-2.75 G
(ection 3.2, calls to)-2.75 E("Update X_acti)99.1 189 Q -.165(ve)-.275 G
(_recv and X_f).165 E(ast_max;" and)-.11 E("Interpolate X_f)99.1 202 Q
(ast_max;")-.11 E(had been re)99.1 215 Q -.165(ve)-.275 G
(rsed accidentally).165 E 5.5(.P)-.715 G
(ut them back in the right order)-5.5 E(.)-.605 E 2.75(*C)93.6 241 S
(hanged Intended Status back to Experimental \(where it started)-2.75 E
(out\).)99.1 254 Q 2.75(*G)93.6 280 S
(eneral editing is response to feedback from Gorry)-2.75 E(.)-.715 E
2.75(*A)93.6 306 S
(dded simulation tests to the list in Section 6: \(1\) simulations)-2.75
E(with a w)99.1 319 Q(orst-case scenario of high congestion, all \215o)
-.11 E(ws using)-.275 E(TFRC, all \215o)99.1 332 Q(ws ha)-.275 E(ving v)
-.22 E(arious idle times, all \215o)-.275 E(ws using F)-.275 E(aster)
-.165 E(Restart, and v)99.1 345 Q(ariable arri)-.275 E -.275(va)-.275 G
2.75(lr).275 G(ates for the TFRC \215o)-2.75 E(ws \(to create)-.275 E
-.275(va)99.1 358 S(riable le).275 E -.165(ve)-.275 G
(ls of congestion\).).165 E(And compare this to the same)5.5 E
(scenario with no \215o)99.1 371 Q(ws using F)-.275 E(aster Restart.)
-.165 E(\(2\) scenarios with)5.5 E
(transient changes from routing changes and from v)99.1 384 Q
(ariable traf)-.275 E(\214c.)-.275 E(The goal is to e)99.1 397 Q
(xplore w)-.165 E(orse-case scenarios sho)-.11 E(wing of)-.275 E 2.75
(ft)-.275 G(he w)-2.75 E(orst)-.11 E(aspects of F)99.1 410 Q
(aster Restart.)-.165 E 2.75(*T)93.6 436 S(ar)-3.63 E
(geted an idle period of at most six minutes, not thirty)-.198 E 2.75
(minutes. Feedback)99.1 449 R(from Gorry and Ian McDonald.)2.75 E 2.75
(*A)93.6 475 S(dded a section of whether F)-2.75 E
(aster Restart encourages \215o)-.165 E(ws to)-.275 E
(pad their sending rate during idle periods.)99.1 488 Q 2.75(*D)93.6 514
S(idn')-2.75 E 2.75(ti)-.198 G(mplement suggestion from Lachlan Andre)
-2.75 E 2.75(wt)-.275 G 2.75(od)-2.75 G(ecay from)-2.75 E
(qradupling to doubling the sending rate gradually)99.1 527 Q 5.5(.T)
-.715 G(he last)-5.5 E
(more-than-doubling of the sending rate is probably not a)99.1 540 Q
(quadrupling in an)99.1 553 Q 2.75(yc)-.165 G(ase, since the allo)-2.75
E(wed sending rate is)-.275 E
(not increased due to quadrupling to more than X_f)99.1 566 Q(ast_max.)
-.11 E(Changes from draft-ietf-dccp-tfrc-f)93.6 592 Q(aster)-.11 E
(-restart-02.txt:)-.22 E 2.75(*D)93.6 618 S
(eleted proposed response to dealing with X_recv for idle or)-2.75 E
(data-limited periods;)99.1 631 Q(RFC3448bis no)5.5 E 2.75(wd)-.275 G
(eals with this instead.)-2.75 E 2.75(*D)93.6 657 S(eleted the Recei)
-2.75 E .33 -.165(ve R)-.275 H(ate Length option.).165 E(Also)5.5 E
(remo)99.1 670 Q -.165(ve)-.165 G 2.75(da).165 G(ll te)-2.75 E
(xt about using the in\215ation f)-.165 E(actor to)-.11 E
(reduce X_recv_in based on the sender')99.1 683 Q 2.75(si)-.605 G
(dle time.)-2.75 E -.385(Ko)72 733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)
-2.75 G 335.849(l. [P)-2.75 F(age 3])-.165 E 0 Cg EP
%%Page: 4 4
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E 2.75(*M)93.6 85 S -.165(ove)
-2.75 G 2.75(dT).165 G
(FRC changes and DCCP-speci\214c changes to separate sections.)-2.75 E
2.75(*R)93.6 111 S -.275(ev)-2.75 G
(ised draft to refer to RFC3448bis instead of to RFC3448.).275 E
(This included modifying sections on "Feedback P)99.1 124 Q(ack)-.165 E
(ets" and)-.11 E("Nofeedback T)99.1 137 Q(imer".)-.385 E 2.75(*S)93.6
163 S(aid that CCID 3 could calculate the recei)-2.75 E .33 -.165(ve r)
-.275 H(ate only).165 E(for one R)99.1 176 Q(TT)-.66 E 2.75(,r)-.814 G
(ather than for longer)-2.75 E 2.75(,a)-.44 G(fter an idle period.)-2.75
E(\(When used with RFC3448bis, it shouldn')99.1 189 Q 2.75(ta)-.198 G
-.275(ff)-2.75 G(ect performance).275 E(one w)99.1 202 Q
(ay or another\).)-.11 E(Changes from draft-ietf-dccp-tfrc-f)93.6 228 Q
(aster)-.11 E(-restart-01.txt:)-.22 E 2.75(*A)93.6 254 S
(dded a sentence to Abstract about DCCP)-2.75 E(.)-1.221 E 2.75(*A)93.6
280 S(dded some te)-2.75 E(xt to the Introduction,)-.165 E 2.75(*A)93.6
306 S(dded sections on "Minimum Sending Rate", "Send Recei)-2.75 E -.165
(ve)-.275 G(Rate Length Feature", "Nofeedback T)99.1 319 Q
(imer", and "Simulations)-.385 E(of F)99.1 332 Q(aster Restart".)-.165 E
2.75(*A)93.6 358 S(dded an Appendix on "Simulations".)-2.75 E
(Changes from draft-ietf-dccp-tfrc-f)93.6 384 Q(aster)-.11 E
(-restart-00.txt:)-.22 E 2.75(*A)93.6 410 S
(dded mechanisms for dealing with a more general problem with)-2.75 E
(idle periods.)99.1 423 Q(This includes a section of "Recei)5.5 E .33
-.165(ve R)-.275 H(ate).165 E(Adjustment".)99.1 436 Q(END OF NO)93.6 462
Q(TE T)-.44 E 2.75(OR)-.198 G(FC EDIT)-2.75 E(OR.)-.198 E/F1 11
/Times-Bold@0 SF(1.)72 488 Q/F2 14/Times-Bold@0 SF(Intr)5.5 E(oduction)
-.252 E F0
(This document de\214nes congestion control mechanisms that impro)93.6
504.6 Q .33 -.165(ve t)-.165 H(he performance of).165 E
(occasionally idle \215o)93.6 517.6 Q
(ws using TCP-Friendly Rate Control \(TFRC\) [RFC3448] [RFC3448bis].)
-.275 E 2.75(Ad)93.6 530.6 S(ata-limited or idle \215o)-2.75 E 2.75(wu)
-.275 G(ses less than its allo)-2.75 E
(wed sending rate for application-speci\214c)-.275 E
(reasons, such as lack of data to send.)93.6 543.6 Q
(The responses of Standard TFRC [RFC3448], and)5.5 E(Re)93.6 556.6 Q(vi\
sed TFRC [RFC3448bis] to long idle or data-limited periods are summariz\
ed in T)-.275 E(able 1)-.88 E(belo)93.6 569.6 Q 1.43 -.715(w, a)-.275 H
(nd the responses of Standard TCP [RFC2581] and TCP with Congestion W)
.715 E(indo)-.44 E(w)-.275 E -1.221(Va)93.6 582.6 S
(lidation [RFC2861] are described in Appendix C of [RFC3448bis].)1.221 E
(All of these)5.5 E(mechanisms allo)93.6 595.6 Q 2.75(wa\215)-.275 G .55
-.275(ow t)-2.75 H 2.75(or).275 G(eco)-2.75 E -.165(ve)-.165 G 2.75(rf)
.165 G(rom a long idle period by ramping up to allo)-2.75 E(wed)-.275 E
(sending rate or windo)93.6 608.6 Q 4.18 -.715(w. T)-.275 H
(his document speci\214es mechanisms that allo).715 E 2.75(wT)-.275 G
(FRC to start at a)-2.75 E
(higher sending rate after an idle period, and to ramp up f)93.6 621.6 Q
(aster to the old sending rate after an)-.11 E(idle period.)93.6 634.6 Q
(As this draft is being written, Standard TFRC is speci\214ed in [RFC34\
48].)93.6 651.2 Q(and TFRC is in the)5.5 E(process of being re)93.6
664.2 Q(vised, as Re)-.275 E(vised TFRC, in [RFC3448bis].)-.275 E
(When [RFC3448bis] is)5.5 E(appro)93.6 677.2 Q -.165(ve)-.165 G 2.75(da)
.165 G 2.75(saP)-2.75 G
(roposed Standard document, this draft will be re)-2.75 E
(vised, with the phrase)-.275 E("Standard TFRC" replaced by "Old TFRC",\
 and other language changes as appropriate.)93.6 690.2 Q -.385(Ko)72 733
S(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G 335.849(l. [P)-2.75 F
(age 4])-.165 E 0 Cg EP
%%Page: 5 5
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E -.165(Fo)93.6 85 S 2.75(rS)
.165 G(tandard TFRC as speci\214ed in [RFC3448], a TFRC \215o)-2.75 E
2.75(wm)-.275 G(ay not send more than twice)-2.75 E(X_recv)93.6 98 Q
2.75(,t)-.715 G(he rate at which data w)-2.75 E(as recei)-.11 E -.165
(ve)-.275 G 2.75(da).165 G 2.75(tt)-2.75 G(he recei)-2.75 E -.165(ve)
-.275 G 2.75(ro).165 G -.165(ve)-2.915 G 2.75(rt).165 G(he pre)-2.75 E
(vious R)-.275 E(TT)-.66 E 5.5(.T)-.814 G(hus in)-5.5 E
(Standard TFRC the pre)93.6 111 Q(vious recei)-.275 E .33 -.165(ve r)
-.275 H(ate limits the sending rate of applications with highly).165 E
-.275(va)93.6 124 S(riable sending rates, forcing the applications to r\
amp up, by doubling their sending rate).275 E(each round-trip time, fro\
m the earlier application-limited rate to the sending rate allo)93.6 137
Q(wed by)-.275 E(the throughput equation.)93.6 150 Q(TFRC')5.5 E 2.75
(sn)-.605 G(ofeedback timer halv)-2.75 E(es the allo)-.165 E
(wed sending rate after)-.275 E(each nofeedback timer interv)93.6 163 Q
(al \(at least four round-trip times\) in which no feedback is)-.275 E
(recei)93.6 176 Q -.165(ve)-.275 G 2.75(d. One).165 F
(result is that applications must slo)2.75 E
(w-start after going idle for an)-.275 E 2.75(ys)-.165 G(igni\214cant)
-2.75 E(length of time, in the absence of mechanisms such as Quick-Star\
t [RFC4782].)93.6 189 Q -.165(Fo)93.6 205.6 S 2.75(rR).165 G -.275(ev)
-2.75 G(ised TFRC as speci\214ed in [RFC3448bis], the pre).275 E
(vious recei)-.275 E .33 -.165(ve r)-.275 H(ate is not used to).165 E
(limit the sending rate during data-limited periods.)93.6 218.6 Q
(Thus, unlik)5.5 E 2.75(e[)-.11 G(RFC3448], in)-2.75 E
([RFC3448bis] applications with highly v)93.6 231.6 Q
(ariable sending rates are not limited by the pre)-.275 E(vious)-.275 E
(recei)93.6 244.6 Q .33 -.165(ve r)-.275 H 2.75(ates. Ho).165 F(we)-.275
E -.165(ve)-.275 G .88 -.44(r, [).165 H(RFC3448bis] is lik).44 E 2.75
(e[)-.11 G(RFC3448] in that the nofeedback timer is)-2.75 E
(used to halv)93.6 257.6 Q 2.75(et)-.165 G(he allo)-2.75 E
(wed sending rate after each nofeedback timer interv)-.275 E
(al in which no)-.275 E(feedback is recei)93.6 270.6 Q -.165(ve)-.275 G
2.75(d. W).165 F(ith [RFC3448] the allo)-.44 E
(wed sending rate is not reduced belo)-.275 E 2.75(wt)-.275 G -.11(wo)
-2.75 G(pack)93.6 283.6 Q(ets per R)-.11 E
(TT during idle periods, and with [RFC3448bis] the allo)-.66 E
(wed sending rate is not)-.275 E(reduced belo)93.6 296.6 Q 2.75(wt)-.275
G(he allo)-2.75 E(wed initial sending rate during idle periods.)-.275 E
(This beha)93.6 313.2 Q(vior is safe, though conserv)-.22 E(ati)-.275 E
-.165(ve)-.275 G 2.75(,f).165 G(or best-ef)-2.75 E(fort traf)-.275 E
(\214c in the netw)-.275 E 2.75(ork. A)-.11 F(silent)2.75 E
(application stops recei)93.6 326.2 Q
(ving feedback about the condition of the current netw)-.275 E
(ork path, and)-.11 E
(thus should not be able to send at an arbitrary rate.)93.6 339.2 Q 2.75
(Ad)5.5 G(ata-limited application stops)-2.75 E(recei)93.6 352.2 Q
(ving feedback about whether current netw)-.275 E(ork conditions w)-.11
E(ould support higher rates.)-.11 E(Ho)93.6 365.2 Q(we)-.275 E -.165(ve)
-.275 G .88 -.44(r, t).165 H(his beha).44 E(vior also af)-.22 E
(fects the percei)-.275 E -.165(ve)-.275 G 2.75(dp).165 G
(erformance of interacti)-2.75 E .33 -.165(ve a)-.275 H(pplications).165
E(such as v)93.6 378.2 Q 2.75(oice. Connections)-.22 F(for interacti)
2.75 E .33 -.165(ve t)-.275 H(elephon).165 E 2.75(ya)-.165 G
(nd conference applications, for)-2.75 E -.165(ex)93.6 391.2 S
(ample, will usually ha).165 E .33 -.165(ve o)-.22 H(ne party acti).165
E .33 -.165(ve a)-.275 H 2.75(tat).165 G
(ime, with seamless switching between)-2.75 E(acti)93.6 404.2 Q .33
-.165(ve p)-.275 H 2.75(arties. TFRC').165 F 2.75(sr)-.605 G
(eduction of the allo)-2.75 E(wed sending rate, and slo)-.275 E
(w-starting back up,)-.275 E(after e)93.6 417.2 Q -.165(ve)-.275 G
(ry switch between parties may seriously de).165 E(grade percei)-.165 E
-.165(ve)-.275 G 2.75(dp).165 G 2.75(erformance. Some)-2.75 F(of)2.75 E
(the strate)93.6 430.2 Q(gies suggested for coping with this problem, s\
uch as sending padding data during)-.165 E
(application idle periods, might ha)93.6 443.2 Q .33 -.165(ve w)-.22 H
(orse ef).055 E(fects on the netw)-.275 E
(ork than simply switching onto)-.11 E(the desired rate with no slo)93.6
456.2 Q(w-start.)-.275 E(There is some justi\214cation for some)93.6
472.8 Q(what accelerating the slo)-.275 E 2.75(ws)-.275 G
(tart process after idle)-2.75 E(periods, as opposed to at the be)93.6
485.8 Q(ginning of a connection.)-.165 E 2.75<418d>5.5 G .55 -.275(ow t)
-2.75 H(hat f).275 E(airly achie)-.11 E -.165(ve)-.275 G 2.75(sa).165 G
(sending rate of X has pro)93.6 498.8 Q -.165(ve)-.165 G
(d, at least, that some path between the endpoints can support that).165
E 2.75(rate. The)93.6 511.8 R(path might change, due to endpoint reset \
or routing adjustments; or man)2.75 E 2.75(yn)-.165 G -.275(ew)-2.75 G
(connections might start up, signi\214cantly reducing the application')
93.6 524.8 Q 2.75(sf)-.605 G(air rate.)-2.86 E(Ho)5.5 E(we)-.275 E -.165
(ve)-.275 G .88 -.44(r, i).165 H(t).44 E(seems reasonable to allo)93.6
537.8 Q 2.75(wa)-.275 G 2.75(na)-2.75 G(pplication to possibly contrib)
-2.75 E(ute to limited transient)-.22 E
(congestion in times of change, in return for impro)93.6 550.8 Q
(ving application responsi)-.165 E -.165(ve)-.275 G(ness.).165 E
(This document suggests a relati)93.6 567.4 Q -.165(ve)-.275 G
(ly simple approach to this problem.).165 E(Standard TFRC)5.5 E
([RFC3448] speci\214es that the allo)93.6 580.4 Q
(wed sending rate is ne)-.275 E -.165(ve)-.275 G 2.75(rr).165 G
(educed belo)-2.75 E 2.75(wt)-.275 G .22 -.11(wo p)-2.75 H(ack).11 E
(ets per)-.11 E -.66(RT)93.6 593.4 S 2.75(Ta).66 G 2.75(st)-2.75 G
(he result of a nofeedback timer after an idle period.)-2.75 E -.165(Fo)
5.5 G(llo).165 E(wing [RFC3390], CCID-3)-.275 E([RFC4342] and Re)93.6
606.4 Q(vised TFRC [RFC3448bis] specify that the allo)-.275 E
(wed sending rate is ne)-.275 E -.165(ve)-.275 G(r).165 E(reduced belo)
93.6 619.4 Q 2.75(wt)-.275 G(he TCP initial sending rate of tw)-2.75 E
2.75(oo)-.11 G 2.75(rf)-2.75 G(our pack)-2.75 E(ets per R)-.11 E(TT)-.66
E 2.75(,d)-.814 G(epending on)-2.75 E(pack)93.6 632.4 Q
(et size, as the result of a nofeedback timer after an idle period.)-.11
E -.165(Fa)5.5 G(ster Restart doubles).165 E(this allo)93.6 645.4 Q
(wed sending rate after idle periods.)-.275 E
(Thus, the sending rate after an idle period is not)5.5 E(reduced belo)
93.6 658.4 Q 2.75(war)-.275 G(ate Y between four and eight pack)-2.75 E
(ets per R)-.11 E(TT)-.66 E 2.75(,d)-.814 G(epending on the pack)-2.75 E
(et)-.11 E 2.75(size. The)93.6 671.4 R
(rate Y is restricted to at most 8760 bytes per R)2.75 E(TT)-.66 E(.)
-.814 E -.385(Ko)72 733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G
335.849(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E(In addition, because \215o)
93.6 85 Q(ws already ha)-.275 E .33 -.165(ve s)-.22 H
(ome \(possibly old\) information about the path,).165 E -.165(Fa)93.6
98 S(ster Restart allo).165 E(ws \215o)-.275 E
(ws to quadruple their sending rate in e)-.275 E -.165(ve)-.275 G
(ry congestion-free R).165 E(TT)-.66 E(,)-.814 E
(instead of doubling, upw)93.6 111 Q(ards to)-.11 E -.11(wa)-.275 G
(rds the pre).11 E(viously achie)-.275 E -.165(ve)-.275 G 2.75(dr).165 G
2.75(ate. When)-2.75 F(the TFRC sender)2.75 E
(detects congestion, the sender lea)93.6 124 Q -.165(ve)-.22 G 2.75(sF)
.165 G(aster Restart and changes into congestion a)-2.915 E -.22(vo)-.22
G(idance.).22 E(These changes are summarized in the table belo)93.6 137
Q -.715(w.)-.275 G/F1 11/Courier@0 SF
(------------------------------------------------------------------)93.6
159.6 Q 6.6(-S)93.6 172.6 S(tandard TFRC -)-6.6 E
(------------------------------------------------------------------)93.6
185.6 Q(Idle period:)93.6 198.6 Q
(Halve allowed sending rate each RTO, not below two packets per RTT.)
106.8 211.6 Q(After sending again, double the sending rate each RTT.)
106.8 224.6 Q(Data-limited period:)93.6 237.6 Q
(Send at most twice X_recv.)106.8 250.6 Q
(As a result, at most double the sending rate each RTT.)106.8 263.6 Q
(------------------------------------------------------------------)93.6
276.6 Q
(------------------------------------------------------------------)93.6
302.6 Q 6.6(-R)93.6 315.6 S(evised TFRC -)-6.6 E
(------------------------------------------------------------------)93.6
328.6 Q(Idle period:)93.6 341.6 Q
(Halve allowed sending rate each RTO, not below initial sending rate.)
106.8 354.6 Q(After sending again, double the sending rate each RTT.)
106.8 367.6 Q(Data-limited period:)93.6 380.6 Q
(Sending rate not limited by X_recv.)106.8 393.6 Q
(------------------------------------------------------------------)93.6
406.6 Q
(------------------------------------------------------------------)93.6
432.6 Q 6.6(-R)93.6 445.6 S(evised TFRC with Faster Restart -)-6.6 E
(------------------------------------------------------------------)93.6
458.6 Q(Idle period:)93.6 471.6 Q
(Halve allowed sending rate each RTO, not below twice initial rate.)
106.8 484.6 Q(\(Specified in Section 3.2.\))120 497.6 Q
(After sending again, quadruple the sending rate towards old rate.)106.8
510.6 Q(\(Specified in Section 3.1.\))120 523.6 Q(Data-limited period:)
93.6 536.6 Q(Sending rate not limited by X_recv.)106.8 549.6 Q
(------------------------------------------------------------------)93.6
562.6 Q(Table 1: Behavior of TFRC, with and without Faster Restart.)93.6
588.6 Q F0
(The congestion control mechanisms here are intended to apply to an)93.6
607.3 Q 2.75(yi)-.165 G(mplementations of)-2.75 E
(TFRC, including that in DCCP')93.6 620.3 Q 2.75(sC)-.605 G
(CID 3 and CCID 4 [RFC4342], [CCID4].)-2.75 E(While we also)5.5 E(belie)
93.6 633.3 Q .33 -.165(ve t)-.275 H
(hat TCP could safely use a similar F).165 E
(aster Restart mechanism, we do not specify it)-.165 E 2.75(here. Our)
93.6 646.3 R(assumption is that \215o)2.75 E(ws that are sensiti)-.275 E
.33 -.165(ve t)-.275 H 2.75(or).165 G
(estrictions to the sending rate after)-2.75 E
(idle periods are more lik)93.6 659.3 Q
(ely to use TFRC than to use TCP or TCP-lik)-.11 E 2.75(ec)-.11 G
(ongestion control.)-2.75 E -.385(Ko)72 733 S(hler).385 E 2.75(,e)-.44 G
2.75(ta)-2.75 G 335.849(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E/F1 11/Times-Bold@0 SF(2.)72
85 Q/F2 14/Times-Bold@0 SF(Con)5.5 E -.14(ve)-.56 G(ntions).14 E F0
(The k)93.6 101.6 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)93.6 114.6 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 [RFC2119].)93.6 127.6 Q
(The F)93.6 144.2 Q(aster Restart mechanism refers to se)-.165 E -.165
(ve)-.275 G(ral e).165 E(xisting TFRC state v)-.165 E
(ariables, including the)-.275 E(follo)93.6 157.2 Q(wing:)-.275 E 8.455
(R: The)93.6 173.8 R -.66(RT)2.75 G 2.75(Te).66 G(stimate.)-2.75 E 7.85
(X: The)93.6 190.4 R(current allo)2.75 E
(wed sending rate in bytes per second.)-.275 E 10.292(p: The)93.6 207 R
(recent loss e)2.75 E -.165(ve)-.275 G(nt rate.).165 E(X_recv:)93.6
223.6 Q(The rate at which the recei)115.2 236.6 Q -.165(ve)-.275 G 2.75
(re).165 G(stimates that data w)-2.75 E(as recei)-.11 E -.165(ve)-.275 G
2.75(ds).165 G(ince the last feedback)-2.75 E(report w)115.2 249.6 Q
(as sent.)-.11 E 11.513(s: The)93.6 266.2 R(pack)2.75 E
(et size in bytes.)-.11 E -.165(Fa)93.6 282.8 S
(ster Restart used the follo).165 E(wing v)-.275 E
(ariable from [RFC3448bis]:)-.275 E(recv_limit:)93.6 299.4 Q
(The limit on the sending rate that is computed from the recei)115.2
312.4 Q .33 -.165(ve r)-.275 H(ate.).165 E -.165(Fa)93.6 329 S
(ster Restart also introduces ne).165 E 2.75(ws)-.275 G(tate v)-2.75 E
(ariables to TFRC, as follo)-.275 E(ws.)-.275 E(X_acti)93.6 345.6 Q
-.165(ve)-.275 G(_recv:).165 E(The recei)115.2 358.6 Q -.165(ve)-.275 G
(r').165 E 2.75(se)-.605 G(stimated recei)-2.75 E .33 -.165(ve r)-.275 H
(ate reported during a recent acti).165 E .33 -.165(ve s)-.275 H
(ending period.).165 E(An)5.5 E(acti)115.2 371.6 Q .33 -.165(ve s)-.275
H(ending period is a period in which the sender has not e).165 E
(xperienced a loss e)-.165 E -.165(ve)-.275 G(nt.).165 E(X_acti)115.2
384.6 Q -.165(ve)-.275 G
(_recv is initialized to 0 until there has been an acti).165 E .33 -.165
(ve s)-.275 H(ending period, and).165 E(X_acti)115.2 397.6 Q -.165(ve)
-.275 G(_recv is reduced after a loss e).165 E -.165(ve)-.275 G(nt.).165
E(T_acti)93.6 414.2 Q -.165(ve)-.275 G(_recv:).165 E
(The time at which X_acti)115.2 427.2 Q -.165(ve)-.275 G(_recv w).165 E
(as measured.)-.11 E(T_acti)5.5 E -.165(ve)-.275 G
(_recv is initialized to the).165 E(connection')115.2 440.2 Q 2.75(ss)
-.605 G(tart time.)-2.75 E(reco)93.6 456.8 Q -.165(ve)-.165 G(r_rate:)
.165 E(The minimum restart rate allo)115.2 469.8 Q(wed by F)-.275 E
(aster Restart after an idle period.)-.165 E(Note that)5.5 E -.165(Fa)
115.2 482.8 S(ster Restart \215o).165 E(ws can drop belo)-.275 E 2.75
(wt)-.275 G(his rate as the result of actual loss feedback.)-2.75 E
(Reco)115.2 495.8 Q -.165(ve)-.165 G(r_rate is de\214ned as follo).165 E
(ws:)-.275 E/F3 11/Courier@0 SF
(recover_rate = min\(8*s, max\(4*s, 8760 bytes\)\)/R.)115.2 514.8 Q F0
-.165(Fa)93.6 533.5 S(ster Restart also uses the follo).165 E
(wing, which could be implemented as a temporary v)-.275 E(ariable:)
-.275 E(X_f)93.6 550.1 Q(ast_max:)-.11 E(The rate at which the sender s\
hould stop quadrupling its sending rate, and return to at)115.2 563.1 Q
(most doubling its sending rate.)115.2 576.1 Q(Other v)93.6 592.7 Q
(ariables ha)-.275 E .33 -.165(ve v)-.22 H
(alues as described in [RFC3448] and [RFC3448bis].)-.11 E F1(3.)72 618.7
Q F2 -.35(Fa)5.5 G(ster Restart: Changes to TFRC).35 E F1(3.1.)72 644.7
Q/F4 13/Times-Bold@0 SF -.325(Fe)5.5 G(edback P).325 E(ack)-.13 E(ets)
-.13 E F0(The F)93.6 661.3 Q(aster Restart algorithm replaces the line)
-.165 E F3(recv_limit = 2 * max \(X_recv_set\);)120 680.3 Q F0 -.385(Ko)
72 733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G 335.849(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E
(in step \(4\) of Section 4.3, "Sender Beha)93.6 85 Q
(vior When a Feedback P)-.22 E(ack)-.165 E(et is Recei)-.11 E -.165(ve)
-.275 G(d", of).165 E 2.75([RFC3448bis]. This)93.6 98 R(line speci\214e\
s the limitation on the sending rate from the recent recei)2.75 E -.165
(ve)-.275 G(rate, and in [RFC3448bis] allo)93.6 111 Q
(ws the sender to slo)-.275 E(w-start back up after an idle period,)
-.275 E(doubling its sending rate after each round-trip time.)93.6 124 Q
(This document replaces the line abo)93.6 140.6 Q -.165(ve)-.165 G 2.75
(,s).165 G 2.75(ot)-2.75 G(hat during reco)-2.75 E -.165(ve)-.165 G
(ry from an idle period, the).165 E(TFRC sender can quadruple its sendi\
ng rate, instead of just doubling it, up to)93.6 153.6 Q -.11(wa)-.275 G
(rds its old).11 E(sending rate before the idle period.)93.6 166.6 Q
(This modi\214cation uses three ne)5.5 E 2.75(wv)-.275 G(ariables,)
-3.025 E(X_acti)93.6 179.6 Q -.165(ve)-.275 G
(_recv specifying the maximum recei).165 E .33 -.165(ve r)-.275 H
(ate achie).165 E -.165(ve)-.275 G 2.75(db).165 G
(efore the idle period,)-2.75 E(T_acti)93.6 192.6 Q -.165(ve)-.275 G
(_recv specifying the time of the last update of X_acti).165 E -.165(ve)
-.275 G(_recv).165 E 2.75(,a)-.715 G(nd X_f)-2.75 E(ast_max)-.11 E(spec\
ifying the adjusted rate at which the sender should stop quadrupling it\
s sending rate, and)93.6 205.6 Q
(return to at most doubling its sending rate.)93.6 218.6 Q
(The procedure `Update X_acti)93.6 235.2 Q -.165(ve)-.275 G
(_recv and X_f).165 E(ast_max" belo)-.11 E 2.75(wi)-.275 G
(ncreases the tw)-2.75 E 2.75(ov)-.11 G(ariables in)-3.025 E
(response to increases in the reported recei)93.6 248.2 Q .33 -.165
(ve r)-.275 H(ate, and reduces them follo).165 E(wing a lost or)-.275 E
(mark)93.6 261.2 Q(ed pack)-.11 E(et.)-.11 E/F1 11/Courier@0 SF
(Update X_active_recv and X_fast_max:)120 280.2 Q
(If \(the feedback packet does not indicate a loss or mark,)146.4 293.2
Q(and X_recv >= X_fast_max\))186 306.2 Q
(X_active_recv := X_fast_max := X_recv,)172.8 319.2 Q
(T_active_recv := current time.)172.8 332.2 Q
(Else if \(the feedback packet DOES indicate a loss or mark,)146.4 345.2
Q(and X_recv < X_fast_max\))186 358.2 Q
(X_active_recv := X_fast_max := X_recv/2,)172.8 371.2 Q
(T_active_recv := current time.)172.8 384.2 Q F0(The parameter X_acti)
93.6 412.3 Q -.165(ve)-.275 G(_recv gi).165 E -.165(ve)-.275 G 2.75(sa)
.165 G 2.75(nu)-2.75 G(pper bound on the rate achie)-2.75 E -.275(va)
-.275 G(ble through F).275 E(aster)-.165 E
(Restart, and is only modi\214ed by the `Update X_acti)93.6 425.3 Q
-.165(ve)-.275 G(_rate and X_f).165 E(ast_max' procedure.)-.11 E
(This modi\214cation is based on the contents of the feedback pack)93.6
438.3 Q(et and the v)-.11 E(alue of)-.275 E(X_f)93.6 451.3 Q 2.75
(ast_max. X_acti)-.11 F -.165(ve)-.275 G
(_recv is updated as the connection achie).165 E -.165(ve)-.275 G 2.75
(sh).165 G(igher congestion-free)-2.75 E(transmit rates.)93.6 464.3 Q
(X_acti)5.5 E -.165(ve)-.275 G
(_recv is reduced on congestion feedback, to pre).165 E -.165(ve)-.275 G
(nt an inappropriate).165 E -.165(Fa)93.6 477.3 S
(ster Restart until a ne).165 E 2.75(ws)-.275 G(table acti)-2.75 E .33
-.165(ve r)-.275 H(ate is achie).165 E -.165(ve)-.275 G 2.75
(d. Speci\214cally).165 F 2.75(,o)-.715 G 2.75(nc)-2.75 G
(ongestion feedback)-2.75 E(at lo)93.6 490.3 Q 2.75(wr)-.275 G
(ates, the sender reduces X_acti)-2.75 E -.165(ve)-.275 G
(_recv to X_recv/2, allo).165 E(wing a limited F)-.275 E(aster Restart)
-.165 E(up to a lik)93.6 503.3 Q(ely-safe rate.)-.11 E -.165(Fo)93.6
519.9 S 2.75(rs).165 G
(ome transport protocols using TFRC, the feedback pack)-2.75 E
(ets might report the loss e)-.11 E -.165(ve)-.275 G(nt).165 E(rate, b)
93.6 532.9 Q(ut not e)-.22 E(xplicity report lost or mark)-.165 E
(ed pack)-.11 E 2.75(ets. F)-.11 F(or such protocols, the sender in the)
-.165 E(`Update X_acti)93.6 545.9 Q -.165(ve)-.275 G(_rate and X_f).165
E(ast_max' procedure can infer that a feedback pack)-.11 E(et indicates)
-.11 E 2.75(al)93.6 558.9 S
(oss or mark by looking at the reported loss e)-2.75 E -.165(ve)-.275 G
(nt rate.).165 E(If the current or pre)5.5 E(vious feedback)-.275 E
(pack)93.6 571.9 Q(et reported an increase in the loss e)-.11 E -.165
(ve)-.275 G(nt rate, then the current feedback pack).165 E(et is)-.11 E
(assumed to indicate a loss or mark.)93.6 584.9 Q(\(If the pre)5.5 E
(vious feedback pack)-.275 E(et reported an increase in)-.11 E
(the loss e)93.6 597.9 Q -.165(ve)-.275 G(nt rate, then a loss e).165 E
-.165(ve)-.275 G(nt be).165 E -.055(ga)-.165 G 2.75(ni).055 G 2.75(nt)
-2.75 G(he interv)-2.75 E(al co)-.275 E -.165(ve)-.165 G
(red by that feedback pack).165 E(et.)-.11 E(Ho)93.6 610.9 Q(we)-.275 E
-.165(ve)-.275 G .88 -.44(r, t).165 H(he loss e).44 E -.165(ve)-.275 G
(nt can co).165 E -.165(ve)-.165 G 2.75(ru).165 G 2.75(pt)-2.75 G 2.75
(oar)-2.75 G(ound-trip time of data, so the second half of the)-2.75 E
(loss e)93.6 623.9 Q -.165(ve)-.275 G
(nt, including additional lost or mark).165 E(ed pack)-.11 E
(ets, could be co)-.11 E -.165(ve)-.165 G(red by the second).165 E
(feedback pack)93.6 636.9 Q(et.\))-.11 E(The `Interpolate X_f)93.6 653.5
Q(ast_max' procedure determines X_f)-.11 E
(ast_max, the adjusted rate at which)-.11 E -.165(Fa)93.6 666.5 S
(ster Restart should stop.).165 E(The procedure sets X_f)5.5 E
(ast_max to something between zero and)-.11 E(X_acti)93.6 679.5 Q -.165
(ve)-.275 G(_recv).165 E 2.75(,d)-.715 G
(epending on the time since X_acti)-2.75 E -.165(ve)-.275 G(_recv w).165
E(as last updated.)-.11 E(The procedure)5.5 E(allo)93.6 692.5 Q
(ws full F)-.275 E(aster Restart up to the old sending rate X_acti)-.165
E -.165(ve)-.275 G(_recv after a short idle period,).165 E -.385(Ko)72
733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G 335.849(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E -.22(bu)93.6 85 S 2.75(tr).22
G(equires more conserv)-2.75 E(ati)-.275 E .33 -.165(ve b)-.275 H(eha)
.165 E(vior after a longer idle period.)-.22 E(Thus, if tw)5.5 E 2.75
(om)-.11 G(inutes or)-2.75 E(less ha)93.6 98 Q .33 -.165(ve e)-.22 H
(lapsed since the last update of X_acti).165 E -.165(ve)-.275 G(_recv)
.165 E 2.75(,t)-.715 G(hen X_f)-2.75 E(ast_max is set to)-.11 E(X_acti)
93.6 111 Q -.165(ve)-.275 G(_recv).165 E 5.5(.I)-.715 G 2.75(fs)-5.5 G
(ix minutes or more ha)-2.75 E .33 -.165(ve e)-.22 H(lapsed, X_f).165 E
(ast_max is set to zero.)-.11 E(Linear)5.5 E
(interpolation is used between these e)93.6 124 Q(xtremes.)-.165 E
(The F)93.6 140.6 Q(aster Restart time interv)-.165 E(al of from tw)
-.275 E 2.75(ot)-.11 G 2.75(os)-2.75 G(ix minutes is chosen to strik)
-2.75 E 2.75(eab)-.11 G(alance)-2.75 E
(between the needs of applications, and the time interv)93.6 153.6 Q
(als o)-.275 E -.165(ve)-.165 G 2.75(rw).165 G(hich connections might)
-2.75 E(reasonably quadruple back up to their old sending rates after i\
dle periods.)93.6 166.6 Q(In terms of the)5.5 E
(needs of applications, models of v)93.6 179.6 Q(oice traf)-.22 E
(\214c generally use a)-.275 E -.165(ve)-.22 G
(rage idle times between 0.5).165 E(and tw)93.6 192.6 Q 2.75(os)-.11 G
(econds [JS00] \(Section 3\).)-2.75 E
(On the other side, in terms of changes in path)5.5 E
(characteristics, F)93.6 205.6 Q(aster Restart doesn')-.165 E 2.75(ta)
-.198 G(ssume that the pre)-2.75 E(vious sending rate is v)-.275 E
(alid after an)-.275 E(idle period;)93.6 218.6 Q -.165(Fa)5.5 G(ster Re\
start simply assumes that a connection may *quadruple* rather than).165
E(*double* its sending rate up to the pre)93.6 231.6 Q(vious rate.)-.275
E -.165(Pa)5.5 G(th congestion le).165 E -.165(ve)-.275 G
(ls can change o).165 E -.165(ve)-.165 G(r).165 E(time scales of round-\
trip times, which are generally between 10 and 200 ms; more dramatic)
93.6 244.6 Q(changes in path characteristics \(e.g., routing changes, c\
hanges in link bandwidth\) happen)93.6 257.6 Q(less frequently)93.6
270.6 Q(.)-.715 E/F1 11/Courier@0 SF(Interpolate X_fast_max:)120 289.6 Q
(// If achieved X_active_recv <= 1 minute ago,)139.8 302.6 Q 19.8
(// set)139.8 315.6 R(X_fast_max to X_active_recv;)6.6 E
(// If achieved X_active_recv >= 3 minutes ago,)139.8 328.6 Q 19.8
(// set)139.8 341.6 R(X_fast_max to zero;)6.6 E
(// If in between, interpolate.)139.8 354.6 Q
(delta_T := now - T_active_recv;)139.8 367.6 Q 6.6(F:)139.8 380.6 S 6.6
(=\()-6.6 G 6.6(6m)-6.6 G
(in - min\(max\(delta_T, 2 min\), 6 min\)\) / \(2 min\);)-6.6 E
(X_fast_max := F * X_active_recv;)139.8 393.6 Q F0(The pseudocode abo)
93.6 421.7 Q .33 -.165(ve u)-.165 H(ses the temporary v).165 E
(ariables delta_T and F)-.275 E(.)-.88 E -.165(Fa)93.6 438.3 S
(ster Restart replaces the follo).165 E
(wing line from step \(4\) of Section 4.3 of [RFC3448bis]:)-.275 E F1
(recv_limit := 2 * max \(X_recv_set\);)120 457.3 Q F0(with the follo)
93.6 485.4 Q(wing:)-.275 E F1(Interpolate X_fast_max;)120 504.4 Q
(Update X_active_recv and X_fast_max;)120 517.4 Q(recv_limit :=)120
530.4 Q 6.6(2*m)13.2 G(ax \(X_recv_set\);)-6.6 E
(If \(recv_limit < X_fast_max\))120 543.4 Q
(recv_limit := min\(2*recv_limit, X_fast_max\);)146.4 556.4 Q F0
(In summary)93.6 584.5 Q 2.75(,w)-.715 G(hen a feedback pack)-2.75 E
(et is recei)-.11 E -.165(ve)-.275 G
(d, as speci\214ed in [RFC3448bis], then the).165 E
(sender updates the round-trip time estimate and the R)93.6 597.5 Q .396
-.198(TO \()-.66 H(Retransmit T).198 E(imeout interv)-.385 E(al\), and)
-.275 E(updates X_recv_set, the set of recent X_recv v)93.6 610.5 Q
(alues, and then e)-.275 E -.165(xe)-.165 G(cutes the procedure abo).165
E -.165(ve)-.165 G(.).165 E(X_f)93.6 623.5 Q(ast_max al)-.11 E -.11(wa)
-.11 G(ys represents the interpolated v).11 E
(alue from highest X_recv reported since the)-.275 E(last loss e)93.6
636.5 Q -.165(ve)-.275 G 2.75(nt. Ho).165 F(we)-.275 E -.165(ve)-.275 G
.88 -.44(r, b).165 H(ecause X_recv_set contains only X_recv v).44 E
(alues from the most)-.275 E(recent tw)93.6 649.5 Q 2.75(or)-.11 G
(ound-trip times, the calculated recv_limit could be less than X_f)-2.75
E 2.75(ast_max. In)-.11 F(this)2.75 E
(case, recv_limit is doubled, up to at most X_f)93.6 662.5 Q 2.75
(ast_max. F)-.11 F(aster Restart')-.165 E 2.75(sd)-.605 G(oubling of)
-2.75 E(recv_limit allo)93.6 675.5 Q(ws the TFRC sender to quadruple it\
s sending rate each round-trip time after an)-.275 E(idle period.)93.6
688.5 Q -.385(Ko)72 733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G
335.849(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E/F1 11/Times-Bold@0 SF(3.2.)72
85 Q/F2 13/Times-Bold@0 SF(Nofeedback T)5.5 E(imer)-.234 E F0
(Section 4.4 of [RFC3448bis] speci\214es when the allo)93.6 101.6 Q
(wed sending rate is halv)-.275 E(ed after the)-.165 E
(nofeedback timer e)93.6 114.6 Q 2.75(xpires. In)-.165 F(particular)2.75
E 2.75(,[)-.44 G(RFC3448bis] speci\214es that if the sender has been)
-2.75 E(idle since the nofeedback timer w)93.6 127.6 Q
(as set, then the allo)-.11 E(wed sending rate is not reduced belo)-.275
E(w)-.275 E(reco)93.6 140.6 Q -.165(ve)-.165 G(r_rate, which in [RFC344\
8bis] is set to the initial_rate of W_init/R, for).165 E
(W_init = min\(4*s, max\(2*s, 4380\)\),)129.6 159.6 Q(for se)93.6 178.6
Q(gment size s.)-.165 E(In contrast, this document sets reco)5.5 E -.165
(ve)-.165 G(r_rate to twice the initial_rate, as).165 E(follo)93.6 191.6
Q(ws:)-.275 E(reco)129.6 210.6 Q -.165(ve)-.165 G(r_rate = 2*W_init/R;)
.165 E F1(4.)72 242.6 Q/F3 14/Times-Bold@0 SF -.35(Fa)5.5 G
(ster Restart: DCCP-speci\214c Speci\214cations).35 E F1(4.1.)72 268.6 Q
F2(DCCP: Recei)5.5 E .26 -.13(ve R)-.13 H(ate Adjustment).13 E F0(Unlik)
93.6 285.2 Q 2.75(e[)-.11 G(RFC3448] and [RFC3448bis], Section 8.3 of D\
CCP CCID 2 [RFC4342] speci\214es)-2.75 E(that the Recei)93.6 298.2 Q .33
-.165(ve R)-.275 H(ate option reports the recei).165 E .33 -.165(ve r)
-.275 H(ate since the last feedback pack).165 E(et w)-.11 E(as sent.)
-.11 E(In contrast, Section 6.2 of [RFC3448] and of [RFC3448bis] specif\
y that the feedback pack)93.6 311.2 Q(et)-.11 E(reports the recei)93.6
324.2 Q .33 -.165(ve r)-.275 H(ate o).165 E -.165(ve)-.165 G 2.75(rt)
.165 G(he last round-trip time.)-2.75 E(As a result, the recei)5.5 E .33
-.165(ve r)-.275 H(ate reported by).165 E([RFC4342] dif)93.6 337.2 Q
(fers from that of TFRC for a feedback pack)-.275 E
(et after an idle period; the recei)-.11 E -.165(ve)-.275 G
(rate report speci\214ed in [RFC4342] reports the recei)93.6 350.2 Q .33
-.165(ve r)-.275 H(ate o).165 E -.165(ve)-.165 G 2.75(rt).165 G
(he entire idle period.)-2.75 E(The)5.5 E(recei)93.6 363.2 Q .33 -.165
(ve r)-.275 H(ate reported by [RFC4342] also dif).165 E
(fers from that of TFRC for an early feedback)-.275 E(pack)93.6 376.2 Q
(et reporting a ne)-.11 E 2.75(wl)-.275 G(oss e)-2.75 E -.165(ve)-.275 G
2.75(nt. In).165 F(this document, [RFC4342] and [CCID4] are updated to)
2.75 E(use the de\214nition of the recei)93.6 389.2 Q .33 -.165(ve r)
-.275 H(ate as speci\214ed in [RFC3448] and [RFC3448bis].).165 E
(In particular)93.6 415.2 Q 2.75(,t)-.44 G
(he fourth paragraph in Section 6 of [RFC4342] is changed from:)-2.75 E
(2. A Recei)115.2 441.2 Q .33 -.165(ve R)-.275 H
(ate option, de\214ned in Section 8.3, specifying the).165 E
(rate at which data w)123.45 454.2 Q(as recei)-.11 E -.165(ve)-.275 G
2.75(ds).165 G(ince the last DCCP-Ack w)-2.75 E(as)-.11 E(sent.)123.45
467.2 Q(to:)93.6 493.2 Q(2. A Recei)115.2 519.2 Q .33 -.165(ve R)-.275 H
(ate option, de\214ned in Section 8.3, specifying the).165 E
(rate at which data w)123.45 532.2 Q(as recei)-.11 E -.165(ve)-.275 G
2.75(do).165 G -.165(ve)-2.915 G 2.75(rt).165 G
(he last round-trip time.)-2.75 E(Similarly)93.6 558.2 Q 2.75(,t)-.715 G
(he \214rst paragraph in Section 8.3 of [RFC4342] is changed from:)-2.75
E(This option MUST be sent by the data recei)115.2 584.2 Q -.165(ve)
-.275 G 2.75(ro).165 G 2.75(na)-2.75 G(ll required ackno)-2.75 E 2.75
(wledgements. Its)-.275 F
(four data bytes indicate the rate at which the recei)115.2 597.2 Q
-.165(ve)-.275 G 2.75(rh).165 G(as recei)-2.75 E -.165(ve)-.275 G 2.75
(dd).165 G(ata since it last sent)-2.75 E(an ackno)115.2 610.2 Q
(wledgement, in bytes per second.)-.275 E 1.76 -.88(To c)5.5 H
(alculate this recei).88 E .33 -.165(ve r)-.275 H(ate, the recei).165 E
-.165(ve)-.275 G(r).165 E(sets t to the lar)115.2 623.2 Q
(ger of the estimated round-trip time and the time since the last Recei)
-.198 E -.165(ve)-.275 G(Rate option w)115.2 636.2 Q(as sent.)-.11 E
(\(Recei)5.5 E -.165(ve)-.275 G 2.75(dd).165 G(ata pack)-2.75 E
(ets' windo)-.11 E 2.75(wc)-.275 G(ounters can be used to produce)-2.75
E 2.75(as)115.2 649.2 S(uitable R)-2.75 E
(TT estimate, as described in Section 8.1.\))-.66 E(The recei)5.5 E .33
-.165(ve r)-.275 H(ate then equals the).165 E
(number of data bytes recei)115.2 662.2 Q -.165(ve)-.275 G 2.75(di).165
G 2.75(nt)-2.75 G(he most recent t seconds, di)-2.75 E(vided by t.)-.275
E(to:)93.6 688.2 Q -.385(Ko)72 733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)
-2.75 G 330.349(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E
(This option MUST be sent by the data recei)115.2 85 Q -.165(ve)-.275 G
2.75(ro).165 G 2.75(na)-2.75 G(ll required ackno)-2.75 E 2.75
(wledgements. Its)-.275 F
(four data bytes indicate the rate at which the recei)115.2 98 Q -.165
(ve)-.275 G 2.75(rh).165 G(as recei)-2.75 E -.165(ve)-.275 G 2.75(dd)
.165 G(ata o)-2.75 E -.165(ve)-.165 G 2.75(rt).165 G(he last)-2.75 E
(round-trip time, in bytes per second.)115.2 111 Q 1.76 -.88(To c)5.5 H
(alculate the time interv).88 E(al t for calculating this)-.275 E(recei)
115.2 124 Q .33 -.165(ve r)-.275 H(ate, the recei).165 E -.165(ve)-.275
G 2.75(rf).165 G(ollo)-2.75 E
(ws Section 6.2 of [RFC3448bis], or roughly equi)-.275 E -.275(va)-.275
G(lently).275 E(,)-.715 E(Section 6.2 of [RFC3448].)115.2 137 Q(\(Recei)
5.5 E -.165(ve)-.275 G 2.75(dd).165 G(ata pack)-2.75 E(ets' windo)-.11 E
2.75(wc)-.275 G(ounters can be used to)-2.75 E(produce a suitable R)
115.2 150 Q(TT estimate, as described in Section 8.1.\))-.66 E
(The recei)5.5 E .33 -.165(ve r)-.275 H(ate then).165 E
(equals the number of data bytes recei)115.2 163 Q -.165(ve)-.275 G 2.75
(di).165 G 2.75(nt)-2.75 G(he most recent t seconds, di)-2.75 E
(vided by t.)-.275 E(Thus, a feedback pack)93.6 189 Q
(et sent in response to the \214rst pack)-.11 E(et recei)-.11 E -.165
(ve)-.275 G 2.75(da).165 G(fter an idle period)-2.75 E(reports a recei)
93.6 202 Q .33 -.165(ve r)-.275 H(ate of one pack).165 E
(et per round-trip time.)-.11 E(As a change from [RFC3448],)5.5 E
([RFC3448bis] doesn')93.6 215 Q 2.75(tu)-.198 G(se the recei)-2.75 E .33
-.165(ve r)-.275 H(ate reported in such pack).165 E
(ets to reduce the allo)-.11 E(wed)-.275 E(sending rate.)93.6 228 Q
(Because [RFC3448bis] doesn')5.5 E 2.75(tu)-.198 G(se the recei)-2.75 E
.33 -.165(ve r)-.275 H(ate to reduce the allo).165 E(wed)-.275 E
(sending rate when the data sender w)93.6 241 Q(as data-limited o)-.11 E
-.165(ve)-.165 G 2.75(rt).165 G(he entire interv)-2.75 E(al co)-.275 E
-.165(ve)-.165 G(red by the).165 E(recei)93.6 254 Q .33 -.165(ve r)-.275
H(ate, the DCCP sender that follo).165 E(ws [RFC3448bis] generally w)
-.275 E(ould not use the)-.11 E(recei)93.6 267 Q .33 -.165(ve r)-.275 H
(ate from an interv).165 E(al in which the sender has not sent an)-.275
E 2.75(yd)-.165 G(ata pack)-2.75 E(ets.)-.11 E 1.76 -.88(We s)93.6 283.6
T(pecify language for DCCP so that if the entire period co).88 E -.165
(ve)-.165 G(red by the last feedback).165 E(pack)93.6 296.6 Q(et doesn')
-.11 E 2.75(ti)-.198 G(nclude an)-2.75 E 2.75(yd)-.165 G(ata pack)-2.75
E(ets, then the sender doesn')-.11 E 2.75(tu)-.198 G
(se the reported recei)-2.75 E .33 -.165(ve r)-.275 H(ate).165 E
(to reduce the sending rate, e)93.6 309.6 Q -.165(ve)-.275 G 2.75(ni)
.165 G 2.75(ft)-2.75 G(he sender w)-2.75 E(as not data-limited o)-.11 E
-.165(ve)-.165 G 2.75(rt).165 G(hat interv)-2.75 E 2.75(al. T)-.275 F
2.75(od)-.88 G(o)-2.75 E(that, we add the follo)93.6 322.6 Q(wing:)-.275
E(Assume that the sender recei)115.2 348.6 Q -.165(ve)-.275 G 2.75(st)
.165 G .22 -.11(wo f)-2.75 H(eedback pack).11 E(ets with Ackno)-.11 E
(wledgement)-.275 E(Numbers A1 and A2, respecti)115.2 361.6 Q -.165(ve)
-.275 G(ly).165 E 5.5(.F)-.715 G
(urther assume that the sender sent no data pack)-5.5 E(ets)-.11 E
(in between Sequence Numbers A1+1 and A2, inclusi)115.2 374.6 Q -.165
(ve)-.275 G 5.5(.\().165 G(All those pack)-5.5 E(ets must ha)-.11 E
-.165(ve)-.22 G(been pure ackno)115.2 387.6 Q
(wledgements, Sync and SyncAck pack)-.275 E(ets, and so forth.\))-.11 E
(Then the)5.5 E(sender MA)115.2 400.6 Q 2.838 -1.419(Y, a)-1.155 H 2.75
(ti)1.419 G(ts discretion, ignore the second feedback pack)-2.75 E(et')
-.11 E 2.75(sR)-.605 G(ecei)-2.75 E .33 -.165(ve R)-.275 H(ate option.)
.165 E
(Note that when the sender decides to ignore such an option, it MUST NO)
115.2 413.6 Q 2.75(Tr)-.44 G(eset the)-2.75 E
(nofeedback timer as it normally w)115.2 426.6 Q
(ould; the nofeedback timer will e)-.11 E(xpire as if the)-.165 E
(second feedback pack)115.2 439.6 Q(et had not been recei)-.11 E -.165
(ve)-.275 G(d.).165 E/F1 11/Times-Bold@0 SF(4.2.)72 465.6 Q/F2 13
/Times-Bold@0 SF(DCCP: The Recei)5.5 E .26 -.13(ve R)-.13 H(ate Length)
.13 E F0([This section will be deleted prior to publication.])93.6 482.2
Q([The Recei)93.6 498.8 Q .33 -.165(ve R)-.275 H
(ate Length option in earlier v).165 E
(ersions of this document has been deleted.)-.165 E(The)5.5 E(Recei)93.6
511.8 Q .33 -.165(ve R)-.275 H
(ate Length option is not needed for feedback pack).165 E
(ets sent after an idle period,)-.11 E
(because of changes in [RFC3448bis].)93.6 524.8 Q(The Recei)5.5 E .33
-.165(ve R)-.275 H(ate Length option should not be used).165 E(for the \
sender to account for short idle periods within a feedback period.)93.6
537.8 Q(The Recei)5.5 E .33 -.165(ve R)-.275 H(ate).165 E
(Length option is also not needed for the case discussed abo)93.6 550.8
Q .33 -.165(ve w)-.165 H(hen the sender is not data-).165 E(limited, b)
93.6 563.8 Q(ut the data sending rate is less than one pack)-.22 E
(et per round-trip time, and the interv)-.11 E(al)-.275 E(co)93.6 576.8
Q -.165(ve)-.165 G(red by the feedback pack).165 E(et doesn')-.11 E 2.75
(ti)-.198 G(nclude an)-2.75 E 2.75(yd)-.165 G(ata pack)-2.75 E
(ets; this case is dealt with)-.11 E(abo)93.6 589.8 Q .33 -.165(ve w)
-.165 H(ithout the use of the Recei).165 E .33 -.165(ve R)-.275 H
(ate Length.]).165 E F1(5.)72 615.8 Q/F3 14/Times-Bold@0 SF -.35(Fa)5.5
G(ster Restart Discussion).35 E F0(Standard TCP has historically dealt \
with idleness and data-limited \215o)93.6 632.4 Q(ws either by k)-.275 E
(eeping)-.11 E
(cwnd entirely open \("immediate start"\) or by entering slo)93.6 645.4
Q(w-start, as recommended in RFC)-.275 E
(2581 in response to an idle period.)93.6 658.4 Q
(The \214rst option is too liberal, the second too)5.5 E(conserv)93.6
671.4 Q(ati)-.275 E -.165(ve)-.275 G 5.5(.C).165 G
(learly a short idle or data-limited period is not a ne)-5.5 E 2.75(wc)
-.275 G(onnection: recent)-2.75 E -.275(ev)93.6 684.4 S(idence sho).275
E(ws that the connection could f)-.275 E
(airly sustain some rate without adv)-.11 E(ersely)-.165 E -.385(Ko)72
733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G 330.349(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E(impacting other \215o)93.6 85
Q 2.75(ws. Ho)-.275 F(we)-.275 E -.165(ve)-.275 G .88 -.44(r, l).165 H
(onger idle periods are more problematic.).44 E(Idle periods of)5.5 E
(man)93.6 98 Q 2.75(ym)-.165 G(inutes w)-2.75 E
(ould seem to require slo)-.11 E(w-start.)-.275 E(RFC 2861 [RFC2861] gi)
93.6 114.6 Q -.165(ve)-.275 G 2.75(sam).165 G(oderate mechanism for TCP)
-2.75 E 2.75(,w)-1.221 G(here the congestion windo)-2.75 E(w)-.275 E
(is halv)93.6 127.6 Q(ed for e)-.165 E -.165(ve)-.275 G
(ry retransmit timeout interv).165 E
(al that the sender has remained idle, do)-.275 E(wn to the)-.275 E
(initial windo)93.6 140.6 Q 1.43 -.715(w, a)-.275 H(nd the windo).715 E
2.75(wi)-.275 G 2.75(sr)-2.75 G(e-opened in slo)-2.75 E
(w-start when the idle period is o)-.275 E -.165(ve)-.165 G -.605(r.)
.165 G(TFRC in [RFC3448bis] roughly follo)93.6 153.6 Q
(ws [RFC2861] for the response to an idle period.)-.275 E(Unlik)93.6
166.6 Q 2.75(e[)-.11 G(RFC2861], ho)-2.75 E(we)-.275 E -.165(ve)-.275 G
.88 -.44(r, [).165 H(RFC3448bis] follo).44 E
(ws Standard TCP in its responses to a data-)-.275 E
(limited period, and does not reduce the allo)93.6 179.6 Q
(wed sending rate in response to data-limited)-.275 E(periods.)93.6
192.6 Q/F1 11/Times-Bold@0 SF(5.1.)72 218.6 Q/F2 13/Times-Bold@0 SF
-.975(Wo)5.5 G(rst-Case Scenarios).975 E F0 -.165(Fa)93.6 235.2 S
(ster Restart should be acceptable for TFRC if its w).165 E
(orst-case scenarios are acceptable.)-.11 E(Realistic w)93.6 248.2 Q
(orst-case scenarios might include the follo)-.11 E(wing scenarios:)
-.275 E 12.65<8350>93.6 264.8 S
(ath changes: The path changes and the old rate isn')-12.815 E 2.75(ta)
-.198 G(cceptable on the ne)-2.75 E 2.75(wp)-.275 G 2.75(ath. R)-2.75 F
(TTs)-.66 E(are shorter on the ne)110.1 277.8 Q 2.75(wp)-.275 G
(ath too, so F)-2.75 E(aster Restart tak)-.165 E
(es bandwidth from other connections)-.11 E(for multiple R)110.1 290.8 Q
(TTs, not just one.)-.66 E
(\(This can happen with TCP or with TFRC without)5.5 E -.165(Fa)110.1
303.8 S(ster Restart, b).165 E(ut F)-.22 E(aster Restart could mak)-.165
E 2.75(et)-.11 G(his beha)-2.75 E(vior more se)-.22 E -.165(ve)-.275 G
(re.\)).165 E 12.65<8353>93.6 320.4 S(ynchronized \215o)-12.65 E(ws: Se)
-.275 E -.165(ve)-.275 G(ral connections enter F).165 E
(aster Restart simultaneously)-.165 E 5.5(.I)-.715 G 2.75(ft)-5.5 G
(he path)-2.75 E(is congested, the e)110.1 333.4 Q
(xtra load resulting from F)-.165 E
(aster Restart could be twice as bad as the)-.165 E -.165(ex)110.1 346.4
S(tra load if the connections had simply slo).165 E
(w-started from their allo)-.275 E(wed initial sending)-.275 E(rate.)
110.1 359.4 Q 12.65<834d>93.6 376 S(an)-12.65 E 2.75(yf)-.165 G
(orms of b)-2.75 E(urstiness: V)-.22 E(ariable-rate connections using F)
-1.221 E(aster Restart share the)-.165 E(congested link with short TCP \
or DCCP connections starting and stopping, with initial)110.1 389 Q
(windo)110.1 402 Q(ws of three or four pack)-.275 E 2.75(ets. The)-.11 F
(aggre)2.75 E -.055(ga)-.165 G(te traf).055 E
(\214c could also include TCP)-.275 E
(connections with short quiescent periods \(e.g., web bro)110.1 415 Q
(wsing sessions using HTTP 1.1\),)-.275 E(or b)110.1 428 Q(ursty higher)
-.22 E(-priority traf)-.22 E 2.75(\214c. As)-.275 F 2.75(ar)2.75 G
(esult of the b)-2.75 E(ursty traf)-.22 E(\214c, the aggre)-.275 E -.055
(ga)-.165 G(te arri).055 E -.275(va)-.275 G 2.75(lr).275 G(ate)-2.75 E
-.275(va)110.1 441 S(ries from one R).275 E(TT to the ne)-.66 E 2.75
(xt. The)-.165 F(transient congestion will be particularly se)2.75 E
-.165(ve)-.275 G(re if).165 E(the congested link is an access link inst\
ead of a backbone link; the le)110.1 454 Q -.165(ve)-.275 G 2.75(lo).165
G 2.75(fs)-2.75 G(tatistical)-2.75 E(multiple)110.1 467 Q
(xing on an access link may not be suf)-.165 E
(\214ciently high to "smooth out" the)-.275 E -.22(bu)110.1 480 S
(rstiness.).22 E 12.65<8357>93.6 496.6 S(ireless links: The netw)-13.09
E(ork allocates capacity based on traf)-.11 E
(\214c conditions, as in some)-.275 E(current wireless technologies, su\
ch as Bandwidth on Demand \(BoD\) links [RFC3819])110.1 509.6 Q
(where capacity is v)110.1 522.6 Q(ariable and dependent on se)-.275 E
-.165(ve)-.275 G(ral parameters other than netw).165 E(ork)-.11 E 2.75
(congestion. In)110.1 535.6 R(this case, the old sending rate might not\
 be acceptable after a change in)2.75 E
(capacity for the wireless link during an idle period.)110.1 548.6 Q
(Further analysis is required to analyze the ef)93.6 565.2 Q
(fects of these scenarios.)-.275 E F1(5.2.)72 591.2 Q F2(Incenti)5.5 E
-.13(ve)-.13 G 3.25(sf).13 G(or applications to send unnecessary pack)
-3.575 E(ets during idle or)-.13 E(data-limited periods?)72 604.2 Q F0
(Ho)93.6 620.8 Q 2.75(wd)-.275 G(oes F)-2.75 E(aster Restart af)-.165 E
(fect an application')-.275 E 2.75(si)-.605 G(ncenti)-2.75 E .33 -.165
(ve t)-.275 H 2.75(op).165 G(ad its sending rate by sending)-2.75 E
(unnecessary pack)93.6 633.8 Q(ets during idle or data-limited periods?)
-.11 E 1.76 -.88(We w)5.5 H(ould lik).77 E 2.75(et)-.11 G 2.75(ol)-2.75
G(imit an)-2.75 E(application')93.6 646.8 Q 2.75(si)-.605 G(ncenti)-2.75
E .33 -.165(ve t)-.275 H 2.75(op).165 G
(ad its sending rate during idle or data-limited periods;)-2.75 E
(if all)5.5 E
(applications were to pad their sending rates, it could reduce the a)
93.6 659.8 Q -.275(va)-.22 G(ilable bandwidth, and).275 E(de)93.6 672.8
Q(grade the performance for all \215o)-.165 E(ws on the congested link.)
-.275 E -.385(Ko)72 733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G
330.349(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E -.44(Wi)93.6 85 S(th Standar\
d TFRC as speci\214ed in [RFC3448], a data-limited TFRC \215o).44 E 2.75
(wm)-.275 G(ay not send)-2.75 E(more than twice X_recv)93.6 98 Q 2.75
(,t)-.715 G(he rate at which data w)-2.75 E(as recei)-.11 E -.165(ve)
-.275 G 2.75(da).165 G 2.75(tt)-2.75 G(he recei)-2.75 E -.165(ve)-.275 G
2.75(ro).165 G -.165(ve)-2.915 G 2.75(rt).165 G(he pre)-2.75 E(vious)
-.275 E -.66(RT)93.6 111 S 4.378 -.814(T. T).66 H
(hus, with Standard TFRC, one could ar).814 E(gue that a v)-.198 E
(ariable-rate application o)-.275 E -.165(ve)-.165 G 2.75(ra).165 G(n)
-2.75 E(uncongested path does ha)93.6 124 Q .33 -.165(ve s)-.22 H
(ome incenti).165 E .33 -.165(ve t)-.275 H 2.75(op).165 G
(ad its sending rate.)-2.75 E -.44(Wi)93.6 140.6 S(th Re).44 E
(vised TFRC as speci\214ed in [RFC3448bis], the allo)-.275 E
(wed sending rate after an idle)-.275 E(period is lar)93.6 153.6 Q
(ger than the allo)-.198 E(wed sending rate with Standard TFRC.)-.275 E
(Further)5.5 E 2.75(,w)-.44 G(ith Re)-2.75 E(vised)-.275 E
(TFRC the recei)93.6 166.6 Q .33 -.165(ve r)-.275 H
(ate reported in feedback pack).165 E
(ets is not used to limit the sending rate)-.11 E
(during data-limited periods.)93.6 179.6 Q(Thus, with Re)5.5 E
(vised TFRC an application has less incenti)-.275 E .33 -.165(ve t)-.275
H(o).165 E(pad its sending rate than with Standard TFRC.)93.6 192.6 Q
(Ho)5.5 E(we)-.275 E -.165(ve)-.275 G .88 -.44(r, w).165 H(ith Re).44 E
(vised TFRC an application)-.275 E(could ha)93.6 205.6 Q .33 -.165(ve s)
-.22 H(ome incenti).165 E .33 -.165(ve t)-.275 H 2.75(op).165 G
(ad its sending rate just enough to maintain the status of "data-)-2.75
E(limited" instead of "idle", by sending at least one pack)93.6 218.6 Q
(et e)-.11 E -.165(ve)-.275 G(ry four round-trip times.).165 E(By allo)
93.6 235.2 Q(wing TFRC to re)-.275 E -.165(ve)-.275 G
(rt to its old sending rate more quickly after an idle period, F).165 E
(aster)-.165 E(Restart could reduce an application')93.6 248.2 Q 2.75
(si)-.605 G(ncenti)-2.75 E .33 -.165(ve t)-.275 H 2.75(op).165 G
(ad its sending rate.)-2.75 E/F1 11/Times-Bold@0 SF(5.3.)72 274.2 Q/F2
13/Times-Bold@0 SF -.325(Fa)5.5 G(ster Restart f).325 E(or TFRC-SP)-.325
E F0 1.76 -.88(We n)93.6 290.8 T(ote that F).88 E(aster Restart with TF\
RC-SP [RFC4828] is considerably more restrained than)-.165 E -.165(Fa)
93.6 303.8 S(ster Restart with TFRC.).165 E(In TFRC-SP)5.5 E 2.75(,t)
-1.221 G(he sender is restricted to sending at most one)-2.75 E(pack)
93.6 316.8 Q(et e)-.11 E -.165(ve)-.275 G(ry Min Interv).165 E(al.)-.275
E F1(6.)72 342.8 Q/F3 14/Times-Bold@0 SF(Simulations of F)5.5 E
(aster Restart)-.35 E F0(Some test case scenarios based on simulation a\
nalysis are described in Appendix A.)93.6 359.4 Q(These)5.5 E
(simulations follo)93.6 372.4 Q 2.75(wt)-.275 G
(he guidelines set in [RFC4828].)-2.75 E(These are:)5.5 E 10.6(1. F)93.6
389 R(airness to standard TCP and TFRC: The simulation tests e)-.165 E
(xamine whether \215o)-.165 E(ws that)-.275 E(use F)115.2 402 Q
(aster Restart allo)-.165 E 2.75(wT)-.275 G(CP and TFRC \215o)-2.75 E
(ws can achie)-.275 E .33 -.165(ve t)-.275 H(heir share of the path).165
E(capacity)115.2 415 Q(.)-.715 E 10.6(2. F)93.6 431.6 R
(airness within FR: The simulation tests e)-.165 E(xamine ho)-.165 E
2.75(wm)-.275 G(ultiple competing FR \215o)-2.75 E(ws)-.275 E
(share the a)115.2 444.6 Q -.275(va)-.22 G(ilable capacity among them.)
.275 E 10.6(3. Response)93.6 461.2 R(to transient e)2.75 E -.165(ve)
-.275 G(nts: The simulation tests e).165 E(xamine ho)-.165 E 2.75(waF)
-.275 G 2.75<528d>-2.75 G .55 -.275(ow r)-2.75 H(eacts to a).275 E
(sudden congestion e)115.2 474.2 Q -.165(ve)-.275 G(nt.).165 E 10.6
(4. Beha)93.6 490.8 R(viour in a range of en)-.22 E(vironments: T)-.44 E
(ests assess a range of bandwidths, R)-.77 E(TTs, and)-.66 E -.275(va)
115.2 503.8 S(rying idle periods.).275 E 2.75(Al)93.6 520.4 S(ater v)
-2.75 E(ersion of this draft will pro)-.165 E
(vide more discussion on these results in the appendix and)-.165 E
(implications will be noted here.)93.6 533.4 Q F1(7.)72 559.4 Q F3
(Implementation Issues)5.5 E F0(TB)93.6 576 Q(A)-.385 E F1(8.)72 602 Q
F3(Security Considerations)5.5 E F0
(DCCP security considerations are discussed in [RFC4340].)93.6 618.6 Q
-.165(Fa)5.5 G(ster Restart adds no additional).165 E
(security considerations.)93.6 631.6 Q F1(9.)72 657.6 Q F3(IAN)5.5 E 3.5
(AC)-.28 G(onsiderations)-3.5 E F0(There are no IAN)93.6 674.2 Q 2.75
(Ac)-.385 G(onsiderations.)-2.75 E -.385(Ko)72 733 S(hler).385 E 2.75
(,e)-.44 G 2.75(ta)-2.75 G 330.349(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E/F1 11/Times-Bold@0 SF(10.)72
85 Q/F2 14/Times-Bold@0 SF(Thanks)5.5 E F0 1.76 -.88(We t)93.6 101.6 T
(hank the DCCP W).88 E
(orking Group for feedback and discussions; we particularly thank)-.88 E
(Gorry F)93.6 114.6 Q 2.75(airhurst. W)-.165 F 2.75(et)-.88 G
(hank Vlad Balan for pointing out problems with the mechanisms)-2.75 E
(discussed in pre)93.6 127.6 Q(vious v)-.275 E(ersions of the draft.)
-.165 E 1.76 -.88(We a)5.5 H(lso thank Gerrit Renk).88 E
(er for feedback.)-.11 E F2(Normati)72 153.6 Q .28 -.14(ve R)-.14 H
(efer).14 E(ences)-.252 E F0 29.634([RFC2119] Bradner)98.7 170.2 R 2.75
(,S)-.44 G(., "K)-2.75 E .33 -.165(ey w)-.275 H
(ords for use in RFCs to Indicate Requirement).055 E(Le)181.2 183.2 Q
-.165(ve)-.275 G(ls", BCP 14, RFC 2119, March 1997.).165 E 29.634
([RFC3448] Handle)98.7 199.8 R 1.43 -.715(y, M)-.165 H(., Flo).715 E
(yd, S., P)-.11 E(adh)-.165 E(ye, J., and J. W)-.055 E(idmer)-.44 E 2.75
(,")-.44 G(TCP Friendly Rate)-2.75 E(Control \(TFRC\): Protocol Speci\
\214cation", RFC 3448, Proposed Standard,)181.2 212.8 Q(January 2003.)
181.2 225.8 Q 16.797([RFC3448bis] Handle)98.7 242.4 R 1.43 -.715(y, M)
-.165 H(., Flo).715 E(yd, S., P)-.11 E(adh)-.165 E(ye, J., and J. W)
-.055 E(idmer)-.44 E 2.75(,")-.44 G(TCP Friendly Rate)-2.75 E(Control \
\(TFRC\): Protocol Speci\214cation", internet draft draft-ietf-dccp-)
181.2 255.4 Q(rfc3448bis-02.txt, w)181.2 268.4 Q
(ork-in-progress, July 2007.)-.11 E 29.634([RFC4340] K)98.7 285 R(ohler)
-.385 E 2.75(,E)-.44 G(., Handle)-2.75 E 1.43 -.715(y, M)-.165 H
(., and S. Flo).715 E(yd, "Datagram Congestion Control)-.11 E
(Protocol \(DCCP\)", RFC 4340, March 2006.)181.2 298 Q 29.634
([RFC4342] Flo)98.7 314.6 R(yd, S., K)-.11 E(ohler)-.385 E 2.75(,E)-.44
G(., and J. P)-2.75 E(adh)-.165 E
(ye, "Pro\214le for Datagram Congestion)-.055 E
(Control Protocol \(DCCP\) Congestion Control ID 3: TCP-Friendly Rate)
181.2 327.6 Q(Control \(TFRC\)", RFC 4342, March 2006.)181.2 340.6 Q F2
(Inf)72 366.6 Q(ormati)-.35 E .28 -.14(ve R)-.14 H(efer).14 E(ences)
-.252 E F0 40.645([CCID4] Flo)98.7 383.2 R(yd, S., and E. K)-.11 E
(ohler)-.385 E 2.75(,")-.44 G(Pro\214le for Datagram Congestion Control)
-2.75 E
(Protocol \(DCCP\) Congestion ID 4: TCP-Friendly Rate Control for Small)
181.2 396.2 Q -.165(Pa)181.2 409.2 S(ck).165 E
(ets \(TFRC-SP\)", Internet-Draft draft-\215o)-.11 E
(yd-dccp-ccid4-01.txt, w)-.11 E(ork)-.11 E(in progress, June 2007.)181.2
422.2 Q 41.866([JCH84] R.K.)98.7 438.8 R(Jain, Dah-Ming Chiu, and W)2.75
E(illian R. Ha)-.44 E(we, A Quantitati)-.165 E -.165(ve)-.275 G
(Measure of F)181.2 451.8 Q
(airness and Discrimination for Resource Allocation in)-.165 E
(Shared Systems, DEC T)181.2 464.8 Q
(echnical tleport TR-301, Digital Equipment)-.77 E
(Corporation, September 1984.)181.2 477.8 Q 51.029([JS00] W)98.7 494.4 R
2.75(.J)-1.012 G(iang and H. Schulzrinne, Analysis of On-Of)-2.75 E 2.75
(fP)-.275 G(atterns in V)-2.915 E(oIP and)-1.419 E(Their Ef)181.2 507.4
Q(fect on V)-.275 E(oice T)-1.419 E(raf)-.385 E(\214c Aggre)-.275 E
-.055(ga)-.165 G(tion, Proceedings of the Ninth).055 E
(Conference on Computer Communications and Netw)181.2 520.4 Q
(orks \(ICCCN\),)-.11 E(October 2000.)181.2 533.4 Q 29.634
([RFC2581] Allman,)98.7 550 R(M., P)2.75 E(axson, V)-.165 E(., and W)
-1.419 E 2.75(.S)-1.012 G(te)-2.75 E -.165(ve)-.275 G
(ns, "TCP Congestion Control",).165 E(RFC 2581, April 1999.)181.2 563 Q
29.634([RFC2861] Handle)98.7 579.6 R 1.43 -.715(y, M)-.165 H(., P).715 E
(adh)-.165 E(ye, J., and S. Flo)-.055 E(yd, "TCP Congestion W)-.11 E
(indo)-.44 E(w)-.275 E -1.221(Va)181.2 592.6 S
(lidation", RFC 2861, June 2000.)1.221 E 29.634([RFC3390] Allman,)98.7
609.2 R(M., Flo)2.75 E(yd, S., and C. P)-.11 E
(artridge, "Increasing TCP')-.165 E 2.75(sI)-.605 G(nitial)-2.75 E -.44
(Wi)181.2 622.2 S(ndo).44 E(w", RFC 3390, October 2002.)-.275 E 29.634
([RFC3819] Karn,)98.7 638.8 R -1.221(P.)2.75 G 2.75(,E)1.221 G
(d., Bormann, C., F)-2.75 E(airhurst, G., Grossman, D., Ludwig, R.,)
-.165 E(Mahda)181.2 651.8 Q(vi, J., Montene)-.22 E(gro, G., T)-.165 E
(ouch, J., and L. W)-.88 E(ood, "Advice for)-.88 E(Internet Subnetw)
181.2 664.8 Q(ork Designers", RFC 3819, July 2004.)-.11 E 29.634
([RFC4782] Flo)98.7 681.4 R(yd, S., Allman, M., Jain, A., and P)-.11 E
2.75(.S)-1.221 G(arolahti, "Quick-Start for TCP)-2.75 E
(and IP", RFC 4782, June 2006.)181.2 694.4 Q -.385(Ko)72 733 S(hler).385
E 2.75(,e)-.44 G 2.75(ta)-2.75 G 330.349(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E 29.634([RFC4828] Flo)98.7 85
R(yd, S., and E. K)-.11 E(ohler)-.385 E 2.75(,")-.44 G
(TCP Friendly Rate Control \(TFRC\): the)-2.75 E(Small-P)181.2 98 Q(ack)
-.165 E(et \(SP\) V)-.11 E(ariant", RFC 4828, April 2007.)-1.221 E/F1 11
/Times-Bold@0 SF(A.)72 124 Q/F2 14/Times-Bold@0 SF -.35(Ap)5.5 G
(pendix: Simulations).35 E F0(This appendix describes a set of initial \
test case scenarios for simulation analysis of F)93.6 140.6 Q(aster)
-.165 E(Restart. The topology will be the classic dumb-bell topology us\
ed in man)93.6 153.6 Q 2.75(ys)-.165 G(imulations of)-2.75 E(TCP)93.6
166.6 Q(.)-1.221 E(Se)93.6 183.2 Q -.165(ve)-.275 G(ral types of \215o)
.165 E(ws are considered:)-.275 E 12.65<8342>93.6 199.8 S(ulk TCP Flo)
-12.65 E(ws.)-.275 E 12.65<8349>93.6 216.4 S(nteracti)-12.65 E .33 -.165
(ve \()-.275 H(short\) TCP Flo).165 E(ws.)-.275 E 12.65<8354>93.6 233 S
(FRC Flo)-12.65 E(ws with and without F)-.275 E(aster Restart.)-.165 E
12.65<8354>93.6 249.6 S(FRC-SP Flo)-12.65 E(ws with and without F)-.275
E(aster Restart.)-.165 E(The implications on other \215o)93.6 266.2 Q
(ws \(e.g. using UDP\) may be e)-.275 E(xtrapolated from this.)-.165 E
-.165(Fo)93.6 282.8 S 2.75(rt).165 G
(hese simulations, we consider three application-limited rates.)-2.75 E
12.65<8354>93.6 299.4 S
(he \214rst resembles Constant Bit Rate \(CBR\) v)-12.65 E(oice o)-.22 E
-.165(ve)-.165 G 2.75(rI).165 G 2.75(Pw)-2.75 G
(ith a media bit rate of 64)-2.75 E(Kbps \(using pack)110.1 312.4 Q
(ets of size 160 bytes and a nominal transmit rate of 8 Kbps\).)-.11 E
12.65<8354>93.6 329 S
(he second resembles Constant Bit Rate \(CBR\) medium quality video o)
-12.65 E -.165(ve)-.165 G 2.75(rI).165 G 2.75(Pw)-2.75 G(ith a)-2.75 E
(media bit rate of 512 Kbps \(using pack)110.1 342 Q
(ets of size 1000 bytes and a nominal transmit rate)-.11 E
(of 64 Kbps\).)110.1 355 Q 12.65<8354>93.6 371.6 S(he third class uses \
an unspeci\214ed upper limit on the sending rate, b)-12.65 E(ut e)-.22 E
(xperiences)-.165 E(period of idleness.)110.1 384.6 Q
(These are intended to be illustrati)93.6 401.2 Q -.165(ve)-.275 G 2.75
(,r).165 G(ather than e)-2.75 E(xact models of the application beha)
-.165 E(viour)-.22 E(.)-.605 E(The simulations will model the ef)93.6
417.8 Q(fect of an idle period in which the application does not)-.275 E
(attempt to send an)93.6 430.8 Q 2.75(yd)-.165 G
(ata for a period of time, then resumes transmission.)-2.75 E
(In the \214rst case, we shall e)93.6 447.4 Q
(xamine periods of idleness of 1s, 2s, and 4s with a path R)-.165 E
(TT of)-.66 E(50ms, 300ms.)93.6 460.4 Q(The scenarios to be e)93.6 477 Q
(xamined are:)-.165 E 12.65<8350>93.6 493.6 S(erformance of a long-li)
-12.65 E -.165(ve)-.275 G 2.75(d\().165 G -.22(bu)-2.75 G
(lk\) TCP \215o).22 E 2.75(w\()-.275 G
(e.g. FTP\) with TFRC \(with and without)-2.75 E
(FR\): The test scenario w)110.1 506.6 Q(ould in)-.11 E -.22(vo)-.44 G
(lv).22 E 2.75(eas)-.165 G(ingle lar)-2.75 E(ge FTP \215o)-.198 E 2.75
(ww)-.275 G(ith v)-2.75 E(arying number of)-.275 E(CBR \215o)110.1 519.6
Q 2.75(ws. Each)-.275 F(CBR \215o)2.75 E 2.75(wb)-.275 G
(ecomes idle for 1s and then restarts.)-2.75 E(The FTP \215o)5.5 E 2.75
(ws)-.275 G(tarts)-2.75 E(during the idle period.)110.1 532.6 Q
(The throughput performance of the single FTP \215o)5.5 E 2.75(ww)-.275
G(ould be)-2.86 E(plotted for v)110.1 545.6 Q
(arying number of CBR \215o)-.275 E 2.75(ws. Simulations)-.275 F -.11
(wo)2.75 G(uld be performed by v).11 E(arying)-.275 E
(parameters such as CBR rate and number of silence periods.)110.1 558.6
Q(Does the single FTP \215o)5.5 E(w)-.275 E(get at least 1/n share of t\
he bandwidth, where 'n' is the total number of \215o)110.1 571.6 Q 2.75
(ws? Does)-.275 F(the)2.75 E(use of F)110.1 584.6 Q
(aster Restart by the TFRC \215o)-.165 E
(ws decrease the bandwidth recei)-.275 E -.165(ve)-.275 G 2.75(db).165 G
2.75(yt)-2.75 G(he TCP)-2.75 E<8d6f>110.1 597.6 Q(w?)-.275 E 12.65<8346>
93.6 614.2 S(airness test: The test scenario w)-12.815 E(ould in)-.11 E
-.22(vo)-.44 G(lv).22 E(ed 'n' CBR and long li)-.165 E -.165(ve)-.275 G
2.75(dT).165 G(CP \215o)-2.75 E 2.75(ws. The)-.275 F(CBR \215o)110.1
627.2 Q(ws become idle for 1s and then restart.)-.275 E
(During the silence period, the FTP \215o)5.5 E(ws)-.275 E(arri)110.1
640.2 Q -.165(ve)-.275 G 5.5(.D).165 G 2.75(oa)-5.5 G(ll \215o)-2.75 E
(ws get at least a 1/n share of the bandwidth?)-.275 E(Jain')5.5 E 2.75
(sF)-.605 G(airness Inde)-2.915 E(x)-.165 E([JCH84] w)110.1 653.2 Q
(ould be an appropriate measure.)-.11 E 12.65<8350>93.6 669.8 S
(erformance of small TCP \215o)-12.65 E
(ws \(HTTP\) with TFRC with and without F)-.275 E(aster Restart:)-.165 E
(The test scenario w)110.1 682.8 Q(ould in)-.11 E -.22(vo)-.44 G(lv).22
E 2.75(eas)-.165 G(ingle CBR \215o)-2.75 E 2.75(ww)-.275 G
(hich runs for 10s, becomes idle)-2.75 E
(between 2s and 3s and then restarts.)110.1 695.8 Q
(At 3s, a number of HTTP \215o)5.5 E(ws are started.)-.275 E(The)5.5 E
-.385(Ko)72 733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G 330.349
(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E
(min, max and median of the request/response time of these HTTP \215o)
110.1 85 Q(ws w)-.275 E(ould be)-.11 E 2.75(plotted. Simulations)110.1
98 R -.11(wo)2.75 G(uld be performed by v).11 E(arying se)-.275 E -.165
(ve)-.275 G(ral parameters such as CBR).165 E
(rate, bottleneck bandwidth, delay and queue size.)110.1 111 Q
(Do the request/response times of these)5.5 E(HTTP \215o)110.1 124 Q
(ws dif)-.275 E(fer? If so, by ho)-.275 E 2.75(wm)-.275 G(uch?)-2.75 E
12.65<8348>93.6 140.6 S(igh-congestion test: In a w)-12.65 E
(orst-case scenario with high congestion, all \215o)-.11 E(ws use TFRC,)
-.275 E(with a range of arri)110.1 153.6 Q -.275(va)-.275 G 2.75(lt).275
G(imes and idle times.)-2.75 E(The simulations are run both with and)5.5
E(without F)110.1 166.6 Q(aster Restart.)-.165 E(Ho)5.5 E 2.75(wd)-.275
G(oes the use of F)-2.75 E(aster Restart af)-.165 E(fect the aggre)-.275
E -.055(ga)-.165 G(te pack).055 E(et)-.11 E(drop rate?)110.1 179.6 Q
12.65<8354>93.6 196.2 S(ransient changes: The \214rst w)-13.035 E
(orst-case scenario with transient changes includes a routing)-.11 E
(change, where the ne)110.1 209.2 Q 2.75(wp)-.275 G
(ath has less bandwidth than the old path.)-2.75 E(The second scenario)
5.5 E(with transient changes includes transient congestion from a sudde\
n increase in traf)110.1 222.2 Q(\214c.)-.275 E(This increase in traf)
110.1 235.2 Q(\214c could be from long-li)-.275 E -.165(ve)-.275 G 2.75
(dT).165 G(CP traf)-2.75 E(\214c, or from higher)-.275 E(-priority)-.22
E(traf)110.1 248.2 Q(\214c, or from man)-.275 E 2.75(yn)-.165 G .55
-.275(ew T)-2.75 H(FRC sessions.).275 E
(The transient congestion could be particularly)5.5 E(se)110.1 261.2 Q
-.165(ve)-.275 G
(re if the congested link is an access link instead of a backbone link.)
.165 E(The third)5.5 E
(scenario with transient changes could include a wireless link with v)
110.1 274.2 Q(ariable bandwidth, as)-.275 E
(discussed earlier in Section 5.)110.1 287.2 Q
(In all cases, the simulations are run both with and without)5.5 E -.165
(Fa)110.1 300.2 S(ster Restart.).165 E(Ho)5.5 E 2.75(wd)-.275 G
(oes the use of F)-2.75 E(aster Restart af)-.165 E(fect the aggre)-.275
E -.055(ga)-.165 G(te pack).055 E(et drop rate?)-.11 E 12.65<8341>93.6
316.8 S 2.75(ni)-12.65 G(deal scenario sho)-2.75 E
(wing the bene\214ts of F)-.275 E
(aster Restart: A scenario with an uncongested)-.165 E(netw)110.1 329.8
Q(ork, just a fe)-.11 E 2.75(wT)-.275 G(FRC \215o)-2.75 E
(ws, comparing the per)-.275 E(-pack)-.22 E(et delay distrib)-.11 E
(ution with and)-.22 E(without F)110.1 342.8 Q(aster Restart.)-.165 E
-.44(Wi)5.5 G(thout F).44 E(aster Restart, there should be a fe)-.165 E
2.75(wp)-.275 G(ack)-2.75 E(ets in each)-.11 E<8d6f>110.1 355.8 Q 2.75
(ww)-.275 G(ith v)-2.75 E(ery lar)-.165 E(ge delay times, from w)-.198 E
(aiting at the sender until the)-.11 E 2.75(yc)-.165 G(an be sent.)-2.75
E 12.65<8341>93.6 372.4 S(scenario sho)-9.9 E
(wing the bene\214ts \(to the \215o)-.275 E 1.43 -.715(w, n)-.275 H
(ot to competing traf).715 E(\214c\) of padding during)-.275 E
(idle periods: Are there an)110.1 385.4 Q 2.75(ys)-.165 G
(cenarios where F)-2.75 E(aster Restart *increases* a \215o)-.165 E(w')
-.275 E 2.75(si)-.605 G(ncenti)-2.75 E -.165(ve)-.275 G(s).165 E
(to pad its sending rate during idle or under)110.1 398.4 Q
(-utilized periods?)-.22 E/F1 14/Times-Bold@0 SF -.7(Au)72 437.4 S
(thors' Addr).7 E(esses)-.252 E F0(Eddie K)93.6 454 Q(ohler <k)-.385 E
(ohler@cs.ucla.edu>)-.11 E(4531C Boelter Hall)93.6 467 Q
(UCLA Computer Science Department)93.6 480 Q(Los Angeles, CA 90095)93.6
493 Q(USA)93.6 506 Q(Sally Flo)93.6 522.6 Q(yd <\215o)-.11 E(yd@icir)
-.11 E(.or)-.605 E(g>)-.198 E(ICSI Center for Internet Research)93.6
535.6 Q(1947 Center Street, Suite 600)93.6 548.6 Q(Berk)93.6 561.6 Q
(ele)-.11 E 1.43 -.715(y, C)-.165 H 2.75(A9).715 G(4704)-2.75 E(USA)93.6
574.6 Q(Arjuna Sathiaseelan <arjuna@er)93.6 591.2 Q(g.abdn.ac.uk>)-.198
E(Electronics Research Group)93.6 604.2 Q(Uni)93.6 617.2 Q -.165(ve)
-.275 G(rsity of Aberdeen).165 E(Aberdeen)93.6 630.2 Q(UK)93.6 643.2 Q
-.385(Ko)72 733 S(hler).385 E 2.75(,e)-.44 G 2.75(ta)-2.75 G 330.349
(l. [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 74.337(-DRAFT Expires:)-1.012 F
(March 2008)2.75 E(September 2007)97.766 E/F1 14/Times-Bold@0 SF
(Full Copyright Statement)72 85 Q F0(Cop)93.6 101.6 Q
(yright \(C\) The IETF T)-.11 E(rust \(2007\).)-.385 E(This document is\
 subject to the rights, licenses and restrictions contained in BCP 78, \
and)93.6 118.2 Q -.165(ex)93.6 131.2 S
(cept as set forth therein, the authors retain all their rights.).165 E
(This document and the information contained herein are pro)93.6 147.8 Q
(vided on an "AS IS" basis and)-.165 E(THE CONTRIB)93.6 160.8 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)93.6 173.8 Q 2.75(,T)
-1.419 G(HE IETF TR)-2.75 E(UST AND THE)-.44 E(INTERNET ENGINEERING T)
93.6 186.8 Q(ASK FORCE DISCLAIM ALL W)-1.023 E(ARRANTIES, EXPRESS)-1.32
E(OR IMPLIED, INCLUDING B)93.6 199.8 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)-2.75 E(USE OF THE INFORMA)93.6 212.8 Q
(TION HEREIN WILL NO)-1.221 E 2.75(TI)-.44 G(NFRINGE ANY RIGHTS OR ANY)
-2.75 E(IMPLIED W)93.6 225.8 Q(ARRANTIES OF MERCHANT)-1.32 E
(ABILITY OR FITNESS FOR A)-1.023 E -1.012(PA)93.6 238.8 S -.66(RT)1.012
G(ICULAR PURPOSE.).66 E F1(Intellectual Pr)72 264.8 Q(operty)-.252 E F0
(The IETF tak)93.6 281.4 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)93.6 294.4 Q
(technology described in this document or the e)93.6 307.4 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)93.6 320.4 Q -.275(va)-.22 G
(ilable; nor does it represent that it has made an).275 E 2.75(yi)-.165
G(ndependent ef)-2.75 E(fort)-.275 E(to identify an)93.6 333.4 Q 2.75
(ys)-.165 G(uch rights.)-2.75 E
(Information on the procedures with respect to rights in RFC)5.5 E
(documents can be found in BCP 78 and BCP 79.)93.6 346.4 Q
(Copies of IPR disclosures made to the IETF Secretariat and an)93.6 363
Q 2.75(ya)-.165 G(ssurances of licenses to be)-2.75 E(made a)93.6 376 Q
-.275(va)-.22 G(ilable, or the result of an attempt made to obtain a ge\
neral license or permission for).275 E(the use of such proprietary righ\
ts by implementers or users of this speci\214cation can be)93.6 389 Q
(obtained from the IETF on-line IPR repository at http://www)93.6 402 Q
(.ietf.or)-.715 E(g/ipr)-.198 E(.)-.605 E(The IETF in)93.6 418.6 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)93.6 431.6
Q -.165(ve)-.165 G 2.75(rt).165 G(echnology that may be)-2.75 E
(required to implement this standard.)93.6 444.6 Q
(Please address the information to the IETF at ietf-)5.5 E(ipr@ietf.or)
93.6 457.6 Q(g.)-.198 E -.385(Ko)72 733 S(hler).385 E 2.75(,e)-.44 G
2.75(ta)-2.75 G 330.349(l. [P)-2.75 F(age 17])-.165 E 0 Cg EP
%%Trailer
end
%%EOF