One document matched: draft-irtf-tmrg-tools-02.ps

%!PS-Adobe-3.0
%%Creator: groff version 1.19
%%CreationDate: Thu Jun  1 22:27:33 2006
%%DocumentNeededResources: font Courier-Bold
%%+ font Times-Bold
%%+ font Times-Roman
%%DocumentSuppliedResources: procset grops 1.19 0
%%Pages: 14
%%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
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
/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
(S. Floyd)198 E 288(INTERNET-DRAFT E.)72 98 R(Kohler)6 E 222
(draft-irtf-tmrg-tools-02.ps Editors)72 111 R(Expires: December 2006)72
124 Q 6(1J)234 G(une 2006)-6 E/F1 14/Times-Bold@0 SF -1.288(To)105.958
149 S(ols f)1.288 E(or the Ev)-.35 E(aluation of Simulation and T)-.14 E
(estbed Scenarios)-1.288 E/F2 11/Times-Bold@0 SF(Status of this Memo)72
207 Q/F3 11/Times-Roman@0 SF
(By submitting this Internet-Draft, each author represents that an)100.8
223.6 Q 2.75(ya)-.165 G(pplicable patent or other)-2.75 E
(IPR claims of which he or she is a)100.8 236.6 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)100.8 249.6
Q -.11(wa)-.165 G
(re will be disclosed, in accordance with Section 6 of BCP 79.).11 E
(Internet-Drafts are w)100.8 266.2 Q
(orking documents of the Internet Engineering T)-.11 E(ask F)-.88 E
(orce \(IETF\), its)-.165 E(areas, and its w)100.8 279.2 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.)100.8 292.2 Q
(Internet-Drafts are draft documents v)100.8 308.8 Q
(alid for a maximum of six months and may be)-.275 E
(updated, replaced, or obsoleted by other documents at an)100.8 321.8 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)
100.8 334.8 Q(ork in progress.")-.11 E
(The list of current Internet-Drafts can be accessed at http://www)100.8
351.4 Q(.ietf.or)-.715 E(g/ietf/1id-)-.198 E(abstracts.txt.)100.8 364.4
Q(The list of Internet-Draft Shado)100.8 381 Q 2.75(wD)-.275 G
(irectories can be accessed at)-2.75 E(http://www)100.8 394 Q(.ietf.or)
-.715 E(g/shado)-.198 E -.715(w.)-.275 G(html.).715 E
(This Internet-Draft will e)100.8 410.6 Q(xpire on December 2006.)-.165
E F2(Abstract)72 436.6 Q F3(This document describes tools for the e)
100.8 453.2 Q -.275(va)-.275 G
(luation of simulation and testbed scenarios used).275 E
(in research on Internet congestion control mechanisms.)100.8 466.2 Q
1.76 -.88(We b)5.5 H(elie).88 E .33 -.165(ve t)-.275 H(hat research in)
.165 E(congestion control mechanisms has been seriously hampered by the\
 lack of good models)100.8 479.2 Q
(underpinning analysis, simulation, and testbed e)100.8 492.2 Q
(xperiments, and that tools for the)-.165 E -.275(eva)100.8 505.2 S(lua\
tion of simulation and testbed scenarios can help in the construction o\
f better).275 E(scenarios, based on better underlying models.)100.8
518.2 Q(One use of the tools described in this)5.5 E
(document is in comparing k)100.8 531.2 Q .33 -.165(ey c)-.11 H
(haracteristics of test scenarios with kno).165 E(wn characteristics)
-.275 E(from the di)100.8 544.2 Q -.165(ve)-.275 G(rse and e).165 E
-.165(ve)-.275 G -.22(r-).165 G(changing real w).22 E 2.75(orld. T)-.11
F(ools characterizing the aggre)-.88 E -.055(ga)-.165 G(te traf).055 E
<8c63>-.275 E(on a link include the distrib)100.8 557.2 Q(ution of per)
-.22 E(-pack)-.22 E(et round-trip times, the distrib)-.11 E
(ution of pack)-.22 E(et)-.11 E(sequence numbers, and the lik)100.8
570.2 Q 2.75(e. T)-.11 F
(ools characterizing end-to-end paths include drop rates)-.88 E
(as a function of pack)100.8 583.2 Q(et size and of b)-.11 E
(urst size, the synchronization ratio between tw)-.22 E 2.75(oe)-.11 G
(nd-to-)-2.75 E(end TCP \215o)100.8 596.2 Q(ws, and the lik)-.275 E 2.75
(e. F)-.11 F(or each characteristic, we describe what aspects of the)
-.165 E(scenario determine this characteristic, ho)100.8 609.2 Q 2.75
(wt)-.275 G(he characteristic can af)-2.75 E(fect the results of)-.275 E
(simulations and e)100.8 622.2 Q(xperiments for the e)-.165 E -.275(va)
-.275 G(luation of congestion control mechanisms, and).275 E
(what is kno)100.8 635.2 Q(wn about this characteristic in the real w)
-.275 E 2.75(orld. W)-.11 F 2.75(ea)-.88 G(lso e)-2.75 E(xplain wh)-.165
E 2.75(yt)-.055 G(he use of)-2.75 E(such tools can add considerable po)
100.8 648.2 Q(wer to our understanding and e)-.275 E -.275(va)-.275 G
(luation of simulation).275 E(and testbed scenarios.)100.8 661.2 Q(Flo)
72 769 Q(yd, K)-.11 E 331.229(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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 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. Introduction)100.8 110 Q F0 11(.......................)11.01 G F2(3)
11.5 E(2. Con)100.8 122 Q -.15(ve)-.4 G(ntions).15 E F0 11
(.......................)10.44 G F2(3)11.5 E(3. T)100.8 134 Q(ools)-.8 E
F0 11(.........................)10.97 G F2(3)11.5 E
(3.1. Characterizing Aggre)115.8 146 Q -.05(ga)-.15 G(te T).05 E(raf)
-.35 E(\214c on a Link)-.25 E F0 11(.............)2.69 G F2(3)11.5 E
(3.2. Characterizing an End-to-End P)115.8 158 Q(ath)-.15 E F0 11
(...............)7.85 G F2(4)11.5 E(3.3. Other Characteristics)115.8 170
Q F0 11(...................)8.25 G F2(4)11.5 E(4. Distrib)100.8 182 Q
(ution of per)-.2 E(-pack)-.2 E(et round-trip times)-.1 E F0 11
(..............)13.61 G F2(4)11.5 E(5. Distrib)100.8 194 Q
(ution of pack)-.2 E(et sequence numbers)-.1 E F0 11(...............)
6.89 G F2(5)11.5 E(6. The Distrib)100.8 206 Q(ution of P)-.2 E(ack)-.15
E(et Sizes)-.1 E F0 11(.................)13.41 G F2(6)11.5 E
(7. The Ratio Between F)100.8 218 Q(orw)-.15 E(ard-path and Re)-.1 E
-.15(ve)-.25 G(rse-path T).15 E(raf)-.35 E<8c63>-.25 E F0 11(..........)
4.4 G F2(6)11.5 E(8. The Distrib)100.8 230 Q(ution of Per)-.2 E(-P)-.2 E
(ack)-.15 E(et Peak Flo)-.1 E 2.5(wR)-.25 G(ates)-2.5 E F0 11
(.............)6.09 G F2(7)11.5 E(9. The Distrib)100.8 242 Q(ution of T)
-.2 E(ransport Protocols.)-.35 E F0 11(...............)9.62 G F2(7)11.5
E(10. The Synchronization Ratio)100.8 254 Q F0 11(...................)
2.68 G F2(8)11.5 E(11. Drop or Mark Rates as a Function of P)100.8 266 Q
(ack)-.15 E(et Size)-.1 E F0 11(............)11.15 G F2(9)11.5 E
(12. Drop Rates as a Function of Burst Size.)100.8 278 Q F0 11
(...............)7.13 G F2(10)6.5 E
(13. Drop Rates as a Function of Sending Rate.)100.8 290 Q F0 11
(..............)8.66 G F2(11)6.5 E
(14. Congestion Control Mechanisms for T)100.8 302 Q(raf)-.35 E
(\214c, along with Sender and)-.25 E F0 11(.......)5.8 G F2(11)6.5 E
(15. Characterization of Congested Links in T)100.8 314 Q
(erms of Bandwidth and T)-.7 E(ypical Le)-.8 E -.15(ve)-.25 G
(ls of Congestion).15 E F0(.)-2.08 E F2(12)8.92 E
(16. Characterization of Challenging Lo)100.8 326 Q(wer Layers.)-.25 E
F0 11(.............)4.08 G F2(12)6.5 E(17. Characterization of Netw)
100.8 338 Q(ork Changes Af)-.1 E(fecting Congestion)-.25 E F0 11
(.........)8.07 G F2(12)6.5 E(18. Using the T)100.8 350 Q
(ools Presented in this Document)-.8 E F0 11(..............)3.34 G F2
(12)6.5 E(19. Related W)100.8 362 Q(ork)-.8 E F0 11
(.......................).43 G F2(12)6.5 E(20. Conclusions)100.8 374 Q
F0 11(.......................)6 G F2(12)6.5 E
(21. Security Considerations)100.8 386 Q F0 11(....................).87
G F2(12)6.5 E(22. IAN)100.8 398 Q 2.5(AC)-.35 G(onsiderations)-2.5 E F0
11(....................)9.56 G F2(12)6.5 E(23. Ackno)100.8 410 Q
(wledgements)-.25 E F0 11(.....................)3.77 G F2(12)6.5 E
(Informati)100.8 422 Q .3 -.15(ve R)-.25 H(eferences).15 E F0 11
(.....................)4.22 G F2(12)6.5 E(Editors' Addresses)100.8 434 Q
F0 11(......................)8.38 G F2(13)6.5 E(Full Cop)100.8 446 Q
(yright Statement)-.1 E F0 11(....................)10.12 G F2(14)6.5 E
(Intellectual Property)100.8 458 Q F0 11(......................)2.27 G
F2(14)6.5 E F0(Flo)72 769 Q(yd, K)-.11 E 331.229(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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 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-irtf-tmr)103.55 111 Q
(g-tools-01.txt:)-.198 E 2.75(*A)103.55 124 S
(dded section on "Drop Rates as a Function of Sending Rate.")-2.75 E
2.75(*A)103.55 150 S(dded a number of ne)-2.75 E 2.75(wr)-.275 G
(eferences.)-2.75 E/F1 11/Times-Bold@0 SF(1.)72 189 Q/F2 14/Times-Bold@0
SF(Intr)5.5 E(oduction)-.252 E F0
(This document discusses tools for the e)100.8 205.6 Q -.275(va)-.275 G
(luation of simulation and testbed scenarios used).275 E
(in research on Internet congestion control mechanisms.)100.8 218.6 Q
(These tools include b)5.5 E(ut are not)-.22 E(limited to measurement t\
ools; the tools discussed in this document are lar)100.8 231.6 Q(gely w)
-.198 E(ays of)-.11 E(characterizing aggre)100.8 244.6 Q -.055(ga)-.165
G(te traf).055 E
(\214c on a link, or characterizing the end-to-end path.)-.275 E
(One use of)5.5 E(these tools is for understanding k)100.8 257.6 Q .33
-.165(ey c)-.11 H(haracteristics of test scenarios; man).165 E 2.75(yc)
-.165 G(haracteristics,)-2.75 E(such as the distrib)100.8 270.6 Q
(ution of per)-.22 E(-pack)-.22 E(et round-trip times on the link, don')
-.11 E 2.75(tc)-.198 G(ome from a single)-2.75 E(input parameter b)100.8
283.6 Q(ut are determined by a range of inputs.)-.22 E 2.75(As)5.5 G
(econd use of the tools is to)-2.75 E(compare k)100.8 296.6 Q .33 -.165
(ey c)-.11 H(haracteristics of test scenarios with what is kno).165 E
(wn of the same)-.275 E(characteristics of the past and current Interne\
t, and with what can be conjectured about)100.8 309.6 Q
(these characteristics of future netw)100.8 322.6 Q 2.75(orks. This)-.11
F(paper follo)2.75 E(ws the general approach from)-.275 E
("Internet Research Needs Better Models" [FK02].)100.8 335.6 Q(As an e)
100.8 352.2 Q(xample of the po)-.165 E
(wer of tools for characterizing scenarios, a great deal is kno)-.275 E
(wn)-.275 E(about the distrib)100.8 365.2 Q
(ution of connection sizes on a link, or equi)-.22 E -.275(va)-.275 G
(lently).275 E 2.75(,t)-.715 G(he distrib)-2.75 E(ution of per)-.22 E(-)
-.22 E(pack)100.8 378.2 Q(et sequence numbers.)-.11 E
(It has been conjectured that a hea)5.5 E(vy-tailed distrib)-.22 E
(ution of)-.22 E(connection sizes is an in)100.8 391.2 Q -.275(va)-.44 G
(riant feature of Internet traf).275 E 2.75(\214c. A)-.275 F
(test scenario with mostly long-)2.75 E(li)100.8 404.2 Q -.165(ve)-.275
G 2.75(dt).165 G(raf)-2.75 E(\214c, or with a mix with only long-li)
-.275 E -.165(ve)-.275 G 2.75(da).165 G(nd v)-2.75 E(ery short \215o)
-.165 E(ws, does not ha)-.275 E .33 -.165(ve a)-.22 H(realistic distrib)
100.8 417.2 Q(ution of connection sizes, and can gi)-.22 E .33 -.165
(ve u)-.275 H(nrealistic results in simulations or).165 E -.165(ex)100.8
430.2 S(periments e).165 E -.275(va)-.275 G
(luating congestion control mechanisms.).275 E -.165(Fo)5.5 G 2.75(ri)
.165 G(nstance, the distrib)-2.75 E(ution of)-.22 E(pack)100.8 443.2 Q
(et sequence numbers mak)-.11 E(es clear the fraction of traf)-.11 E
(\214c on a link from medium-sized)-.275 E(connections, e.g., with pack)
100.8 456.2 Q(et sequence numbers from 100 to 1000.)-.11 E
(These medium-sized)5.5 E(connections can slo)100.8 469.2 Q
(w-start up to a lar)-.275 E(ge congestion windo)-.198 E 1.43 -.715
(w, p)-.275 H(ossibly coming to an abrupt).715 E(stop soon afterw)100.8
482.2 Q(ards, contrib)-.11 E(uting signi\214cantly to the b)-.22 E
(urstiness of the aggre)-.22 E -.055(ga)-.165 G(te traf).055 E(\214c,)
-.275 E(and to the problems f)100.8 495.2 Q(acing congestion control.)
-.11 E(In the sections belo)100.8 511.8 Q 2.75(ww)-.275 G 2.75(ew)-2.75
G(ill discuss a number of tools for describing and e)-2.75 E -.275(va)
-.275 G(luating).275 E(scenarios, sho)100.8 524.8 Q 2.75(wh)-.275 G .55
-.275(ow t)-2.75 H(hese characteristics can af).275 E
(fect the results of research on congestion)-.275 E
(control mechanisms, and summarize what is kno)100.8 537.8 Q
(wn about these characteristics in real-)-.275 E -.11(wo)100.8 550.8 S
(rld netw).11 E(orks.)-.11 E F1(2.)72 576.8 Q F2(Con)5.5 E -.14(ve)-.56
G(ntions).14 E F0(The k)100.8 593.4 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
606.4 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].)100.8 619.4
Q F1(3.)72 645.4 Q F2 -1.288(To)5.5 G(ols)1.288 E F0
(The tools or characteristics that we discuss are the follo)100.8 662 Q
(wing.)-.275 E F1(3.1.)72 688 Q/F3 13/Times-Bold@0 SF
(Characterizing Aggr)5.5 E(egate T)-.234 E(raf\214c on a Link)-.962 E F0
12.65<8344>100.8 704.6 S(istrib)-12.65 E(ution of per)-.22 E(-pack)-.22
E(et round-trip times.)-.11 E(Flo)72 769 Q(yd, K)-.11 E 273.479
(ohler Section)-.385 F 2.75(3.1. [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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E 12.65<8344>100.8 85 S(istrib)
-12.65 E(ution of pack)-.22 E(et sequence numbers.)-.11 E 12.65<8344>
100.8 101.6 S(istrib)-12.65 E(ution of pack)-.22 E(et sizes.)-.11 E
12.65<8352>100.8 118.2 S(atio between forw)-12.65 E(ard-path and re)-.11
E -.165(ve)-.275 G(rse-path traf).165 E(\214c.)-.275 E 12.65<8344>100.8
134.8 S(istrib)-12.65 E(ution of peak \215o)-.22 E 2.75(wr)-.275 G
(ates.)-2.75 E 12.65<8344>100.8 151.4 S(istrib)-12.65 E
(ution of transport protocols.)-.22 E/F1 11/Times-Bold@0 SF(3.2.)72
177.4 Q/F2 13/Times-Bold@0 SF(Characterizing an End-to-End P)5.5 E(ath)
-.13 E F0 12.65<8353>100.8 194 S(ynchronization ratio.)-12.65 E 12.65
<8344>100.8 210.6 S(rop rates as a function of pack)-12.65 E(et size.)
-.11 E 12.65<8344>100.8 227.2 S(rop rates as a function of b)-12.65 E
(urst size.)-.22 E 12.65<8344>100.8 243.8 S
(rop rates as a function of sending rate.)-12.65 E 12.65<8344>100.8
260.4 S -.165(eg)-12.65 G(ree of pack).165 E(et drops.)-.11 E 12.65
<8352>100.8 277 S(ange of queueing delay)-12.65 E(.)-.715 E F1(3.3.)72
303 Q F2(Other Characteristics)5.5 E F0 12.65<8343>100.8 319.6 S
(ongestion control mechanisms for traf)-12.65 E
(\214c, along with sender and recei)-.275 E -.165(ve)-.275 G 2.75(rb)
.165 G(uf)-2.97 E(fer sizes.)-.275 E 12.65<8343>100.8 336.2 S(haracteri\
zation of congested links in terms of bandwidth and typical le)-12.65 E
-.165(ve)-.275 G(ls of).165 E(congestion \(in terms of pack)117.3 349.2
Q(et drop rates\).)-.11 E 12.65<8343>100.8 365.8 S
(haracterization of congested links in terms of b)-12.65 E(uf)-.22 E
(fer size.)-.275 E 12.65<8343>100.8 382.4 S
(haracterization of challenging lo)-12.65 E
(wer layers in terms of reordering, delay v)-.275 E(ariation,)-.275 E
(pack)117.3 395.4 Q(et corruption, and the lik)-.11 E(e.)-.11 E 12.65
<8343>100.8 412 S(haracterization of netw)-12.65 E(ork changes af)-.11 E
(fecting congestion, such as routing changes or)-.275 E(link outages.)
117.3 425 Q(Belo)100.8 441.6 Q 2.75(ww)-.275 G 2.75(ew)-2.75 G
(ill discuss each characteristic in turn, gi)-2.75 E
(ving the de\214nition, the f)-.275 E(actors)-.11 E
(determining that characteristic, the ef)100.8 454.6 Q
(fect on congestion control metrics, and what is kno)-.275 E(wn)-.275 E
(so f)100.8 467.6 Q(ar from measurement studies in the Internet.)-.11 E
F1(4.)72 493.6 Q/F3 14/Times-Bold@0 SF(Distrib)5.5 E(ution of per)-.28 E
(-pack)-.518 E(et r)-.14 E(ound-trip times)-.252 E F0
(De\214nition: The distrib)100.8 510.2 Q(ution of per)-.22 E(-pack)-.22
E(et round-trip times on a link is de\214ned formally by)-.11 E
(assigning to each pack)100.8 523.2 Q
(et the most recent round trip time measured for that end-to-end)-.11 E
2.75(connection. In)100.8 536.2 R
(practice, coarse-grained information is generally suf)2.75 E
(\214cient, e)-.275 E -.165(ve)-.275 G 2.75(nt).165 G(hough it)-2.75 E
(has been sho)100.8 549.2 Q(wn that there is signi\214cant v)-.275 E
(ariability in round-trip times within a TCP)-.275 E
(connection [AKSJ03], and it is suf)100.8 562.2 Q
(\214cient to assign to each pack)-.275 E
(et the \214rst round-trip time)-.11 E(measurement for that connection,\
 or to assign the current round-trip time estimate)100.8 575.2 Q
(maintained by the TCP connection.)100.8 588.2 Q(Determining f)100.8
617.8 Q(actors: The distrib)-.11 E(ution of per)-.22 E(-pack)-.22 E
(et round-trip times on a link is determined)-.11 E
(by end-to-end propag)100.8 630.8 Q
(ation delays, by queueing delays along end-to-end paths, and by the)
-.055 E(congestion control mechanisms used by the traf)100.8 643.8 Q
2.75(\214c. F)-.275 F(or e)-.165 E(xample, for a scenario using TCP)
-.165 E(,)-1.221 E
(TCP connections with smaller round-trip times will recei)100.8 656.8 Q
.33 -.165(ve a p)-.275 H(roportionally lar).165 E(ger fraction)-.198 E
(of traf)100.8 669.8 Q(\214c than competing TCP connections with lar)
-.275 E(ger round-trip times, all else being)-.198 E
(equal, due to the dynamics of TCP f)100.8 682.8 Q -.22(avo)-.11 G
(ring \215o).22 E(ws with smaller round-trip times.)-.275 E(This will)
5.5 E(generally shift the distrib)100.8 695.8 Q(ution of per)-.22 E
(-pack)-.22 E(et R)-.11 E(TTs lo)-.66 E(wer relati)-.275 E .33 -.165
(ve t)-.275 H 2.75(ot).165 G(he distrib)-2.75 E(ution of per)-.22 E(-)
-.22 E(connection R)100.8 708.8 Q(TTs, since short-R)-.66 E
(TT connections will ha)-.66 E .33 -.165(ve m)-.22 H(ore pack).165 E
(ets.)-.11 E(Flo)72 769 Q(yd, K)-.11 E 281.729(ohler Section)-.385 F
2.75(4. [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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E(Ef)100.8 85 Q
(fect on congestion control metrics: The distrib)-.275 E(ution of per)
-.22 E(-pack)-.22 E(et round-trip times on a)-.11 E(link af)100.8 98 Q
(fects the b)-.275 E(urstiness of the aggre)-.22 E -.055(ga)-.165 G
(te traf).055 E(\214c, and therefore can af)-.275 E(fect congestion)
-.275 E(control performance in a range of areas such as delay/throughpu\
t tradeof)100.8 111 Q 2.75(fs. The)-.275 F(distrib)100.8 124 Q
(ution of per)-.22 E(-pack)-.22 E(et round-trip times can also af)-.11 E
(fect metrics of f)-.275 E(airness, de)-.11 E(gree of)-.165 E
(oscillations, and the lik)100.8 137 Q 2.75(e. F)-.11 F(or e)-.165 E
(xample, long-term oscillations of queueing delay are more)-.165 E(lik)
100.8 150 Q(ely to occur in scenarios with a narro)-.11 E 2.75(wr)-.275
G(ange of round-trip times [FK02].)-2.75 E(Measurements: The distrib)
100.8 166.6 Q(ution of per)-.22 E(-pack)-.22 E
(et round-trip times for TCP traf)-.11 E(\214c on a link can)-.275 E
(be measured from a pack)100.8 179.6 Q(et trace with the passi)-.11 E
.33 -.165(ve T)-.275 H(CP round-trip time estimator from Jiang).165 E
(and Do)100.8 192.6 Q(vrolis [JD02].)-.165 E
([Add pointers to other estimators, such as ones mentioned in JD02.)5.5
E(Add a pointer to Mark Allman')100.8 205.6 Q 2.75(sl)-.605 G
(oss detection tool.])-2.75 E(Their paper sho)5.5 E(ws the distrib)-.275
E(ution of)-.22 E(per)100.8 218.6 Q(-pack)-.22 E
(et round-trip times for TCP pack)-.11 E(ets for a number of dif)-.11 E
(ferent links.)-.275 E -.165(Fo)5.5 G 2.75(rt).165 G(he links)-2.75 E
(measured, the percent of pack)100.8 231.6 Q
(ets with round-trip times at most 100 ms ranged from 30% to)-.11 E
(80%, and the percent of pack)100.8 244.6 Q
(ets with round-trip times at most 200 ms ranged from 55% to)-.11 E
(90%, depending on the link.)100.8 257.6 Q(In the NS simulator)100.8
274.2 Q 2.75(,t)-.44 G(he distrib)-2.75 E(ution of per)-.22 E(-pack)-.22
E(et round-trip times for TCP pack)-.11 E(ets on a)-.11 E
(link can be reported by the queue monitor)100.8 287.2 Q 2.75(,u)-.44 G
(sing TCP')-2.75 E 2.75(se)-.605 G(stimated round-trip time added to)
-2.75 E(pack)100.8 300.2 Q(et headers.)-.11 E
(This is illustrated in the v)5.5 E
(alidation test "./test-all-simple stats3" in the)-.275 E
(directory tcl/test.)100.8 313.2 Q(Scenarios: [FK02] sho)100.8 329.8 Q
(ws a relati)-.275 E -.165(ve)-.275 G
(ly simple scenario, with a dumbbell topology with four).165 E
(access links on each end, that gi)100.8 342.8 Q -.165(ve)-.275 G 2.75
(saf).165 G(airly realistic range of round-trip times.)-2.86 E
([Look for)5.5 E(the other citations to add.])100.8 355.8 Q/F1 11
/Times-Bold@0 SF(5.)72 381.8 Q/F2 14/Times-Bold@0 SF(Distrib)5.5 E
(ution of pack)-.28 E(et sequence numbers)-.14 E F0
(De\214nition: The distrib)100.8 398.4 Q(ution of pack)-.22 E
(et sequence numbers on a link is de\214ned by gi)-.11 E(ving each)-.275
E(pack)100.8 411.4 Q(et a sequence number)-.11 E 2.75(,w)-.44 G
(here the \214rst pack)-2.75 E
(et in a connection has sequence number 1,)-.11 E(the second pack)100.8
424.4 Q(et has sequence number 2, and so on.)-.11 E(The distrib)5.5 E
(ution of pack)-.22 E(et sequence)-.11 E(numbers can be deri)100.8 437.4
Q -.165(ve)-.275 G 2.75(di).165 G 2.75(nas)-2.75 G(traightforw)-2.75 E
(ard manner from the distrib)-.11 E(ution of connection)-.22 E
(sizes, and vice v)100.8 450.4 Q 2.75(ersa; ho)-.165 F(we)-.275 E -.165
(ve)-.275 G .88 -.44(r, t).165 H(he distrib).44 E
(ution of connection sizes is more suited for traf)-.22 E<8c63>-.275 E
(generators, and the distrib)100.8 463.4 Q(ution of pack)-.22 E
(et sequence numbers is more suited for measuring)-.11 E
(and illustrating the pack)100.8 476.4 Q(ets actually seen on a link o)
-.11 E -.165(ve)-.165 G 2.75(ra\214).165 G -.165(xe)-2.75 G 2.75(di).165
G(nterv)-2.75 E(al of time.)-.275 E(There has)5.5 E
(been a considerably body of research o)100.8 489.4 Q -.165(ve)-.165 G
2.75(rt).165 G(he last ten years on the hea)-2.75 E(vy-tailed distrib)
-.22 E(ution)-.22 E(of connection sizes for traf)100.8 502.4 Q
(\214c on the Internet.)-.275 E([CBC95] [Add citations.])5.5 E
(Determining f)100.8 519 Q(actors: The distrib)-.11 E
(ution of connection sizes is lar)-.22 E(gely determined by the)-.198 E
(traf)100.8 532 Q(\214c generators used in a scenario.)-.275 E -.165(Fo)
5.5 G 2.75(re).165 G(xample, is there a single traf)-2.915 E
(\214c generator)-.275 E(characterized by a distrib)100.8 545 Q
(ution of connection sizes?)-.22 E 2.75(Am)5.5 G(ix of long-li)-2.75 E
-.165(ve)-.275 G 2.75(da).165 G(nd web traf)-2.75 E(\214c,)-.275 E
(with the web traf)100.8 558 Q(\214c characterized by a distrib)-.275 E
(ution of connection sizes?)-.22 E(Or something)5.5 E(else?)100.8 571 Q
(Ef)100.8 587.6 Q(fect on congestion control metrics: The distrib)-.275
E(ution of pack)-.22 E(et sequence numbers af)-.11 E(fects)-.275 E
(the b)100.8 600.6 Q(urstiness of aggre)-.22 E -.055(ga)-.165 G(te traf)
.055 E(\214c on a link, thereby af)-.275 E
(fecting all congestion control metrics)-.275 E(for which this is a f)
100.8 613.6 Q(actor)-.11 E 5.5(.A)-.605 G 2.75(sa)-5.5 G 2.75(ne)-2.75 G
(xample, [FK02] illustrates that the traf)-2.915 E(\214c mix can af)
-.275 E(fect)-.275 E(the queue dynamics on a congested link.)100.8 626.6
Q([Find more to cite, about the ef)5.5 E(fect of the)-.275 E(distrib)
100.8 639.6 Q(ution of pack)-.22 E
(et sequence numbers on congestion control metrics.])-.11 E
([Add a paragraph about the impact of medium-size \215o)100.8 656.2 Q
(ws.])-.275 E([Add a paragraph about the impact of \215o)100.8 672.8 Q
(ws starting and stopping.])-.275 E([Add a w)100.8 689.4 Q
(arning about scenarios that use only long-li)-.11 E -.165(ve)-.275 G
2.75<648d>.165 G -.275(ow)-2.75 G(s, or a mix of long-li).275 E -.165
(ve)-.275 G 2.75(da).165 G(nd)-2.75 E -.165(ve)100.8 702.4 S
(ry short \215o).165 E(ws.])-.275 E(Flo)72 769 Q(yd, K)-.11 E 281.729
(ohler Section)-.385 F 2.75(5. [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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E
(Measurements: [Cite some of the literature.])100.8 85 Q -.385(Tr)100.8
101.6 S(af).385 E(\214c generators: Some of the a)-.275 E -.275(va)-.22
G(ilable traf).275 E(\214c generators are listed on the web site for)
-.275 E("T)100.8 114.6 Q(raf)-.385 E(\214c Generators for Internet T)
-.275 E(raf)-.385 E(\214c" [TG].)-.275 E(This includes pointers to traf)
5.5 E(\214c generators)-.275 E(for peer)100.8 127.6 Q(-to-peer traf)-.22
E(\214c, traf)-.275 E(\214c from online g)-.275 E(ames, and traf)-.055 E
(\214c from Distrib)-.275 E(uted Denial of)-.22 E
(Service \(DDoS\) attacks.)100.8 140.6 Q(In the NS simulator)100.8 157.2
Q 2.75(,t)-.44 G(he distrib)-2.75 E(ution of pack)-.22 E
(et sequence numbers for TCP pack)-.11 E(ets on a link)-.11 E
(can be reported by the queue monitor at a router)100.8 170.2 Q 5.5(.T)
-.605 G(his is illustrated in the v)-5.5 E(alidation test)-.275 E
("./test-all-simple stats3" in the directory tcl/test.)100.8 183.2 Q/F1
11/Times-Bold@0 SF(6.)72 209.2 Q/F2 14/Times-Bold@0 SF(The Distrib)5.5 E
(ution of P)-.28 E(ack)-.14 E(et Sizes)-.14 E F0
(De\214nition: The distrib)100.8 225.8 Q(ution of pack)-.22 E
(et sizes is de\214ned in a straightforw)-.11 E(ard w)-.11 E(ay)-.11 E
2.75(,u)-.715 G(sing)-2.75 E(pack)100.8 238.8 Q(et sizes in bytes.)-.11
E(Determining f)100.8 255.4 Q(actors: The distrib)-.11 E(ution of pack)
-.22 E(et sizes is determined by the traf)-.11 E(\214c mix, the)-.275 E
(path MTUs, and by the pack)100.8 268.4 Q
(et sizes used by the transport-le)-.11 E -.165(ve)-.275 G 2.75(ls).165
G(enders.)-2.75 E(The distrib)100.8 285 Q(ution of pack)-.22 E
(et sizes on a link is also determined by the mix of forw)-.11 E
(ard-path)-.11 E(TCP traf)100.8 298 Q(\214c and re)-.275 E -.165(ve)
-.275 G(rse-path TCP traf).165 E
(\214c in that scenario, for a scenario characterized by a)-.275 E
(`forw)100.8 311 Q
(ard path' \(e.g., left to right on a particular link\) and a `re)-.11 E
-.165(ve)-.275 G(rse path' \(e.g., right to left).165 E
(on the same link\).)100.8 324 Q -.165(Fo)5.5 G 2.75(rs).165 G
(uch a scenario, the forw)-2.75 E(ard-path TCP traf)-.11 E
(\214c contrib)-.275 E(utes data)-.22 E(pack)100.8 337 Q
(ets to the forw)-.11 E(ard link and ackno)-.11 E(wledgment pack)-.275 E
(ets to the re)-.11 E -.165(ve)-.275 G(rse link, while the).165 E(re)
100.8 350 Q -.165(ve)-.275 G(rse-path TCP traf).165 E(\214c contrib)
-.275 E(utes small ackno)-.22 E(wledgment pack)-.275 E(ets to the forw)
-.11 E(ard link.)-.11 E(The ratio between TCP data and TCP A)100.8 363 Q
(CK pack)-.44 E(ets on a link can be used as some)-.11 E
(indication of the ratio between forw)100.8 376 Q(ard-path and re)-.11 E
-.165(ve)-.275 G(rse-path TCP traf).165 E(\214c.)-.275 E(Ef)100.8 392.6
Q(fect on congestion control metrics: The distrib)-.275 E(ution of pack)
-.22 E(et sizes on a link is an)-.11 E(indicator of the ratio of forw)
100.8 405.6 Q(ard-path and re)-.11 E -.165(ve)-.275 G(rse-path TCP traf)
.165 E(\214c in that netw)-.275 E 2.75(ork. The)-.11 F(amount of re)
100.8 418.6 Q -.165(ve)-.275 G(rse-path traf).165 E
(\214c determines the loss and queueing delay e)-.275 E(xperienced by)
-.165 E(ackno)100.8 431.6 Q(wledgement pack)-.275 E(ets on the re)-.11 E
-.165(ve)-.275 G(rse path, signi\214cantly af).165 E(fecting the b)-.275
E(urstiness of the)-.22 E(aggre)100.8 444.6 Q -.055(ga)-.165 G(te traf)
.055 E(\214c on the forw)-.275 E(ard path.)-.11 E([In what other w)5.5 E
(ays does the distrib)-.11 E(ution of pack)-.22 E(et)-.11 E(sizes af)
100.8 457.6 Q(fect congestion control metrics?])-.275 E
(Measurements: There has been a wealth of measurements o)100.8 474.2 Q
-.165(ve)-.165 G 2.75(rt).165 G(ime on the pack)-2.75 E(et size)-.11 E
(distrib)100.8 487.2 Q(ution of traf)-.22 E(\214c [A00], [HMTG01].)-.275
E(These measurements are generally consistent)5.5 E(with a model of rou\
ghly 10% of the TCP connections using an MSS of roughly 500 bytes,)100.8
500.2 Q
(and with the other 90% of TCP connections using an MSS of 1460 bytes.)
100.8 513.2 Q F1(7.)72 539.2 Q F2(The Ratio Between F)5.5 E
(orward-path and Re)-.35 E -.14(ve)-.21 G(rse-path T).14 E(raf\214c)
-1.036 E F0(De\214nition: F)100.8 555.8 Q
(or a scenario characterized by a `forw)-.165 E
(ard path' \(e.g., left to right on a)-.11 E
(particular link\) and a `re)100.8 568.8 Q -.165(ve)-.275 G
(rse path' \(e.g., right to left on the same link\), the ratio between)
.165 E(forw)100.8 581.8 Q(ard-path and re)-.11 E -.165(ve)-.275 G
(rse-path traf).165 E
(\214c can be de\214ned as the ratio between the forw)-.275 E(ard-path)
-.11 E(traf)100.8 594.8 Q(\214c in bps, and the re)-.275 E -.165(ve)
-.275 G(rse-path traf).165 E(\214c in bps.)-.275 E(Determining f)100.8
611.4 Q(actors: The ratio between forw)-.11 E(ard-path and re)-.11 E
-.165(ve)-.275 G(rse-path traf).165 E(\214c is de\214ned)-.275 E(lar)
100.8 624.4 Q(gely by the traf)-.198 E(\214c mix.)-.275 E(Ef)100.8 641 Q
(fect on congestion control metrics: Zhang, Shenk)-.275 E
(er and Clark ha)-.11 E .33 -.165(ve s)-.22 H(ho).165 E(wn in 1991 that)
-.275 E(for TCP)100.8 654 Q 2.75(,t)-1.221 G(he amount of re)-2.75 E
-.165(ve)-.275 G(rse-path traf).165 E(\214c af)-.275 E(fects the A)-.275
E(CK compression and pack)-.44 E(et drop)-.11 E(rate for TCP ackno)100.8
667 Q(wledgement pack)-.275 E(ets, signi\214cantly af)-.11 E
(fecting the b)-.275 E(urstiness of TCP traf)-.22 E<8c63>-.275 E
(on the forw)100.8 680 Q(ard path [ZSC91].)-.11 E
(The queueing delay on the re)5.5 E -.165(ve)-.275 G(rse path also af)
.165 E(fects the)-.275 E(performance of delay-based congestion control \
mechanisms, if the delay is computed)100.8 693 Q
(based on round-trip times.)100.8 706 Q(This has been sho)5.5 E
(wn by Grieco and Mascolo in [GM04] and by)-.275 E(Prasad, Jain, and Do)
100.8 719 Q(vrolis in [PJD04].)-.165 E(Flo)72 769 Q(yd, K)-.11 E 281.729
(ohler Section)-.385 F 2.75(7. [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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E(Measurements: There is a need\
 for measurements on the range of ratios between forw)100.8 85 Q(ard-)
-.11 E(path and re)100.8 98 Q -.165(ve)-.275 G(rse-path traf).165 E
(\214c for congested links.)-.275 E(In particular)5.5 E 2.75(,f)-.44 G
(or TCP traf)-2.75 E(\214c tra)-.275 E -.165(ve)-.22 G(rsing).165 E
(congested link X, what is the lik)100.8 111 Q(elihood that the ackno)
-.11 E(wledgement traf)-.275 E(\214c will encounter)-.275 E
(congestion \(i.e., queueing delay)100.8 124 Q 2.75(,p)-.715 G(ack)-2.75
E(et drops\) some)-.11 E(where on the re)-.275 E -.165(ve)-.275 G
(rse path as well?).165 E(As discussed in Section 6, the distrib)100.8
140.6 Q(ution of pack)-.22 E(et sizes on a link can be used as an)-.11 E
(indicator of the ratio of forw)100.8 153.6 Q(ard-path and re)-.11 E
-.165(ve)-.275 G(rse-path TCP traf).165 E(\214c in that netw)-.275 E
(ork.)-.11 E/F1 11/Times-Bold@0 SF(8.)72 179.6 Q/F2 14/Times-Bold@0 SF
(The Distrib)5.5 E(ution of P)-.28 E(er)-.28 E(-P)-.518 E(ack)-.14 E
(et P)-.14 E(eak Flo)-.28 E 3.5(wR)-.14 G(ates)-3.5 E F0
(De\214nition: The distrib)100.8 196.2 Q(ution of peak \215o)-.22 E 2.75
(wr)-.275 G(ates is de\214ned by assigning to each pack)-2.75 E(et the)
-.11 E(peak sending rate in bytes per second of that connection, where \
the peak sending rate is)100.8 209.2 Q(de\214ned o)100.8 222.2 Q -.165
(ve)-.165 G 2.75(r0).165 G(.1-second interv)-2.75 E 2.75(als. The)-.275
F(distrib)2.75 E(ution of peak \215o)-.22 E 2.75(wr)-.275 G(ates gi)
-2.75 E -.165(ve)-.275 G 2.75(ss).165 G(ome)-2.75 E
(indication of the ratio of "alpha" and "beta" traf)100.8 235.2 Q
(\214c on a link, where alpha traf)-.275 E(\214c on a)-.275 E
(congested link is de\214ned as traf)100.8 248.2 Q
(\214c with that link at the main bottleneck, while the beta)-.275 E
(traf)100.8 261.2 Q(\214c on the link has a primary bottleneck else)
-.275 E(where along its path [RSB01].)-.275 E(Determining f)100.8 277.8
Q(actors: The distrib)-.11 E(ution of peak \215o)-.22 E 2.75(wr)-.275 G
(ates is determined by \215o)-2.75 E(ws with)-.275 E(bottlenecks else)
100.8 290.8 Q(where along their end-to-end path, e.g., \215o)-.275 E
(ws with lo)-.275 E(w-bandwidth access)-.275 E 2.75(links. The)100.8
303.8 R(distrib)2.75 E(ution of peak \215o)-.22 E 2.75(wr)-.275 G
(ates is also af)-2.75 E(fected by applications with limited)-.275 E
(sending rates.)100.8 316.8 Q(Ef)100.8 333.4 Q
(fect on congestion control metrics: The distrib)-.275 E
(ution of peak \215o)-.22 E 2.75(wr)-.275 G(ates af)-2.75 E(fects the)
-.275 E -.22(bu)100.8 346.4 S(rstiness of aggre).22 E -.055(ga)-.165 G
(te traf).055 E(\214c, with lo)-.275 E(w-peak-rate traf)-.275 E
(\214c decreasing the aggre)-.275 E -.055(ga)-.165 G(te).055 E -.22(bu)
100.8 359.4 S(rstiness, and adding to the traf).22 E(\214c')-.275 E 2.75
(st)-.605 G(ractability)-2.75 E(.)-.715 E(Measurements: [RSB01].)100.8
376 Q(The distrib)5.5 E(ution of peak rates can be e)-.22 E
(xpected to change o)-.165 E -.165(ve)-.165 G(r).165 E(time, as there i\
s an increasing number of high-bandwidth access links to the home, and \
of)100.8 389 Q(high-bandwidth Ethernet links at w)100.8 402 Q
(ork and at other institutions.)-.11 E(Simulators: [F)100.8 418.6 Q(or \
NS, add a pointer to the DelayBox, "http://dirt.cs.unc.edu/delaybox/", \
for)-.165 E(more easily simulating lo)100.8 431.6 Q
(w-bandwidth access links for \215o)-.275 E(ws.])-.275 E -.77(Te)100.8
448.2 S
(stbeds: In testbeds, Dummynet [Dummynet] and NISTNet [NISTNet] pro).77
E(vide)-.165 E(con)100.8 461.2 Q -.165(ve)-.44 G(nient w).165 E
(ays to emulate paths with dif)-.11 E(ferent limited peak rates.)-.275 E
F1(9.)72 487.2 Q F2(The Distrib)5.5 E(ution of T)-.28 E(ransport Pr)
-1.036 E(otocols.)-.252 E F0(De\214nition: The distrib)100.8 503.8 Q
(ution of transport protocols on a congested link is straightforw)-.22 E
(ard,)-.11 E(with each pack)100.8 516.8 Q(et gi)-.11 E -.165(ve)-.275 G
2.75(ni).165 G(ts associated transport protocol \(e.g., TCP)-2.75 E 2.75
(,U)-1.221 G 2.75(DP\). The)-2.75 F(distrib)2.75 E(ution)-.22 E
(is often gi)100.8 529.8 Q -.165(ve)-.275 G 2.75(nb).165 G
(oth in terms of pack)-2.75 E(ets and in terms of bytes.)-.11 E -.165
(Fo)100.8 546.4 S 2.75(rU).165 G(DP pack)-2.75 E(ets, it might be more \
helpful to classify them in terms of the port number)-.11 E 2.75(,o)-.44
G(r)-2.75 E(the assumed application \(e.g., DNS, RIP)100.8 559.4 Q 2.75
(,g)-1.221 G(ames, W)-2.805 E(indo)-.44 E(ws Media, RealAudio, RealV)
-.275 E(ideo,)-.66 E 2.75(etc.\) [MA)100.8 572.4 R 2.75(WI]\). Other)
-.99 F(traf)2.75 E(\214c includes ICMP)-.275 E 2.75(,I)-1.221 G
(PSEC, and the lik)-2.75 E 2.75(e. In)-.11 F(the future there could)2.75
E(be traf)100.8 585.4 Q(\214c from SCTP)-.275 E 2.75(,D)-1.221 G(CCP)
-2.75 E 2.75(,o)-1.221 G 2.75(rf)-2.75 G(rom other transport protocols.)
-2.75 E(Ef)100.8 602 Q(fect on congestion control metrics: The distrib)
-.275 E(ution of transport protocols af)-.22 E(fects metrics)-.275 E
(relating to the ef)100.8 615 Q(fecti)-.275 E -.165(ve)-.275 G
(ness of A).165 E(QM mechanisms on a link.)-.605 E
(Measurements: In the past, TCP traf)100.8 631.6 Q
(\214c has typically consisted of 90% to 95% of the bytes)-.275 E
(on a link [UW02], [U)100.8 644.6 Q 2.75(A01]. [Get)-.44 F
(updated citations for this.])2.75 E(Measurement studies sho)5.5 E(w)
-.275 E(that TCP traf)100.8 657.6 Q(\214c from web serv)-.275 E
(ers almost al)-.165 E -.11(wa)-.11 G
(ys uses conformant TCP congestion control).11 E(procedures [MAF05].)
100.8 670.6 Q(Flo)72 769 Q(yd, K)-.11 E 281.729(ohler Section)-.385 F
2.75(9. [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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E/F1 11/Times-Bold@0 SF(10.)72
85 Q/F2 14/Times-Bold@0 SF(The Synchr)5.5 E(onization Ratio)-.252 E F0
(De\214nition: The synchronization ratio is de\214ned as the de)100.8
101.6 Q(gree of synchronization of loss)-.165 E -2.365 -.275(ev e)100.8
114.6 T(nts between tw).275 E 2.75(oT)-.11 G(CP \215o)-2.75 E
(ws on the same path.)-.275 E
(Thus, the synchronization ratio is de\214ned)5.5 E
(as a characteristic of an end-to-end path.)100.8 127.6 Q
(When one TCP \215o)5.5 E 2.75(wo)-.275 G 2.75(fap)-2.75 G
(air has a loss e)-2.75 E -.165(ve)-.275 G(nt, the).165 E
(synchronization ratio is gi)100.8 140.6 Q -.165(ve)-.275 G 2.75(nb).165
G 2.75(yt)-2.75 G(he fraction of those loss e)-2.75 E -.165(ve)-.275 G
(nts for which the second \215o).165 E(w)-.275 E(has a loss e)100.8
153.6 Q -.165(ve)-.275 G(nt within one round-trip time.).165 E
(Each connection in a \215o)5.5 E 2.75(wp)-.275 G(air has a separate)
-2.75 E(synchronization ratio, and the o)100.8 166.6 Q -.165(ve)-.165 G
(rall synchronization ratio of the pair of \215o).165 E
(ws is the higher)-.275 E(of the tw)100.8 179.6 Q 2.75(or)-.11 G 2.75
(atios. When)-2.75 F
(measuring the synchronization ratio, it is preferable to start the)2.75
E(tw)100.8 192.6 Q 2.75(oT)-.11 G(CP \215o)-2.75 E(ws at slightly dif)
-.275 E(ferent times, with lar)-.275 E(ge recei)-.198 E .33 -.165(ve w)
-.275 H(indo).165 E(ws.)-.275 E(Determining f)100.8 209.2 Q
(actors: The synchronization ratio is determined lar)-.11 E
(gely by the traf)-.198 E(\214c mix on)-.275 E
(the congested link, and by the A)100.8 222.2 Q
(QM mechanism \(or lack of A)-.605 E(QM mechanism\).)-.605 E(Dif)100.8
238.8 Q(ferent types of TCP \215o)-.275 E(ws are also lik)-.275 E
(ely to ha)-.11 E .33 -.165(ve d)-.22 H(if).165 E
(ferent synchronization measures.)-.275 E(E.g., T)100.8 251.8 Q .22 -.11
(wo H)-.88 H(ighSpeed TCP \215o).11 E(ws might ha)-.275 E .33 -.165
(ve h)-.22 H(igher synchronization measures that tw).165 E(o)-.11 E
(Standard TCP \215o)100.8 264.8 Q
(ws on the same path, because of their more aggressi)-.275 E .33 -.165
(ve w)-.275 H(indo).165 E 2.75(wi)-.275 G(ncrease)-2.75 E 2.75
(rates. Raina,)100.8 277.8 R -1.265 -.88(To w)2.75 H(sle).88 E 1.43
-.715(y, a)-.165 H(nd W).715 E(ischik [R)-.44 E(TW05] ha)-.66 E .33
-.165(ve d)-.22 H(iscussed the relationships between).165 E
(synchronization and TCP')100.8 290.8 Q 2.75(si)-.605 G
(ncrease and decrease parameters.)-2.75 E(Ef)100.8 307.4 Q
(fect on congestion control metrics: The synchronization ratio af)-.275
E(fects con)-.275 E -.165(ve)-.44 G -.198(rg).165 G(ence times).198 E
(for high-bandwidth TCPs.)100.8 320.4 Q(Con)5.5 E -.165(ve)-.44 G -.198
(rg).165 G(ence times are kno).198 E(wn to be poor for some high-)-.275
E(bandwidth protocols in en)100.8 333.4 Q(vironments with high le)-.44 E
-.165(ve)-.275 G(ls of synchronization.).165 E([Cite the papers)5.5 E
(by Leith and Shorten.])100.8 346.4 Q(Ho)5.5 E(we)-.275 E -.165(ve)-.275
G .88 -.44(r, i).165 H 2.75(ti).44 G 2.75(sn)-2.75 G
(ot clear if these en)-2.75 E(vironments with high le)-.44 E -.165(ve)
-.275 G(ls of).165 E(synchronization are realistic.)100.8 359.4 Q -.44
(Wi)100.8 376 S(schik and MeKweon [WM05] ha).44 E .33 -.165(ve s)-.22 H
(ho).165 E(wn that the le)-.275 E -.165(ve)-.275 G 2.75(lo).165 G 2.75
(fs)-2.75 G(ynchronization af)-2.75 E(fects the)-.275 E -.22(bu)100.8
389 S -.275(ff).22 G(er requirements at congested routers.).275 E
(Baccelli and Hong [BH02] ha)5.5 E .33 -.165(ve a m)-.22 H(odel).165 E
(sho)100.8 402 Q(wing the ef)-.275 E
(fect of the synchronization ratio on aggre)-.275 E -.055(ga)-.165 G
(te throughput.).055 E(Measurements: Gren)100.8 418.6 Q
(ville Armitage and Qiang Fu ha)-.44 E .33 -.165(ve p)-.22 H
(erformed initial e).165 E(xperiments of)-.165 E
(synchronization in the Internet, using Standard TCP \215o)100.8 431.6 Q
(ws, and ha)-.275 E .33 -.165(ve f)-.22 H(ound v).165 E(ery lo)-.165 E
2.75(wl)-.275 G -2.365 -.275(ev e)-2.75 H(ls).275 E(of synchronization.)
100.8 444.6 Q(In a discussion of the relationship between stability and\
 desynchronization, Raina, T)100.8 461.2 Q -.275(ow)-.88 G(sle).275 E
-.715(y,)-.165 G(and W)100.8 474.2 Q(ischik [R)-.44 E
(TW05] report that "synchronization has been reported ag)-.66 E
(ain and ag)-.055 E(ain in)-.055 E 2.75(simulations". In)100.8 487.2 R
(contrast, synchronization has not been reported ag)2.75 E(ain and ag)
-.055 E(ain in the)-.055 E(real-w)100.8 500.2 Q(orld Internet.)-.11 E
(Appenzeller)100.8 516.8 Q 2.75(,K)-.44 G(eslassy)-3.025 E 2.75(,a)-.715
G(nd McK)-2.75 E(eo)-.275 E(wn in [AKM04] report the follo)-.275 E
(wing: "Flo)-.275 E(ws are not)-.275 E
(synchronized in a backbone router carrying thousands of \215o)100.8
529.8 Q(ws with v)-.275 E(arying R)-.275 E(TTs. Small)-.66 E -.275(va)
100.8 542.8 S(riations in R).275 E(TT or processing time are suf)-.66 E
(\214cient to pre)-.275 E -.165(ve)-.275 G(nt synchronization [QZK01];)
.165 E
(and the absence of synchronization has been demonstrated in real netw)
100.8 555.8 Q(orks [F02,IMD01].")-.11 E
([Appenzeller et al, Sizing Router Buf)100.8 572.4 Q
(fers, reports that synchronization is rare as the)-.275 E
(number of competing \215o)100.8 585.4 Q(ws increases.)-.275 E -.275
(Kev)5.5 G(in Jef).275 E -.11(fa)-.275 G 2.75(yh).11 G
(as some results on synchronization)-2.75 E(also.])100.8 598.4 Q
(Needed: W)100.8 615 Q 2.75(en)-.88 G
(eed measurements of the synchronization ratio for \215o)-2.75 E
(ws that use high-)-.275 E(bandwidth protocols o)100.8 628 Q -.165(ve)
-.165 G 2.75(rh).165 G(igh-bandwidth paths, gi)-2.75 E -.165(ve)-.275 G
2.75(nt).165 G(ypical le)-2.75 E -.165(ve)-.275 G(ls of competing traf)
.165 E<8c63>-.275 E
(and with typical queuing mechanisms at routers \(whate)100.8 641 Q
-.165(ve)-.275 G 2.75(rt).165 G(hese are\), to see if there are)-2.75 E
(higher le)100.8 654 Q -.165(ve)-.275 G(ls of synchronization with high\
-bandwidth protocols such as HighSpeed TCP).165 E(,)-1.221 E -.165(Fa)
100.8 667 S(st TCP).165 E 2.75(,a)-1.221 G(nd the lik)-2.75 E
(e, which are more aggressi)-.11 E .33 -.165(ve t)-.275 H
(han Standard TCP).165 E 5.5(.T)-1.221 G(he assumption)-5.5 E -.11(wo)
100.8 680 S(uld be that in man).11 E 2.75(ye)-.165 G -.44(nv)-2.75 G
(ironments, high-bandwidth protocols ha).44 E .33 -.165(ve h)-.22 H
(igher le).165 E -.165(ve)-.275 G(ls of).165 E
(synchronization than \215o)100.8 693 Q(ws using Standard TCP)-.275 E(.)
-1.221 E(Flo)72 769 Q(yd, K)-.11 E 276.229(ohler Section)-.385 F 2.75
(10. [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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E/F1 11/Times-Bold@0 SF(11.)72
85 Q/F2 14/Times-Bold@0 SF(Dr)5.5 E(op or Mark Rates as a Function of P)
-.252 E(ack)-.14 E(et Size)-.14 E F0
(De\214nition: Drop rates as a function of pack)100.8 101.6 Q
(et size are de\214ned by the actual drop rates for)-.11 E(dif)100.8
114.6 Q(ferent pack)-.275 E
(ets on an end-to-end path or on a congested link o)-.11 E -.165(ve)
-.165 G 2.75(rap).165 G(articular time)-2.75 E(interv)100.8 127.6 Q 2.75
(al. In)-.275 F(some cases, e.g., Drop-T)2.75 E
(ail queues in units of pack)-.88 E(ets, general statements can)-.11 E
(be made; e.g., that lar)100.8 140.6 Q(ge and small pack)-.198 E
(ets will e)-.11 E(xperience the same pack)-.165 E(et drop rates.)-.11 E
(Ho)100.8 153.6 Q(we)-.275 E -.165(ve)-.275 G .88 -.44(r, i).165 H 2.75
(no).44 G(ther cases, e.g., Drop-T)-2.75 E
(ail queues in units of bytes, no such general statement)-.88 E
(can be made, and the drop rate as a function of pack)100.8 166.6 Q
(et size will be determined in part by the)-.11 E(traf)100.8 179.6 Q
(\214c mix at the congested link at that point of time.)-.275 E
(Determining f)100.8 196.2 Q
(actors: The drop rate as a function of pack)-.11 E
(et size is determined in part by the)-.11 E(queue architecture.)100.8
209.2 Q(E.g., is the Drop-T)5.5 E(ail queue in units of pack)-.88 E
(ets, of bytes, of 60-byte)-.11 E -.22(bu)100.8 222.2 S -.275(ff).22 G
(ers, or of a mix of b).275 E(uf)-.22 E(fer sizes?)-.275 E(Is the A)5.5
E(QM mechanism in pack)-.605 E(et mode, dropping each)-.11 E(pack)100.8
235.2 Q(et with the same probability)-.11 E 2.75(,o)-.715 G 2.75(ri)
-2.75 G 2.75(nb)-2.75 G(yte mode, with the probability of dropping or)
-2.75 E(marking a pack)100.8 248.2 Q(et being proportional to the pack)
-.11 E(et size in bytes.)-.11 E(The ef)100.8 264.8 Q(fect of pack)-.275
E(et size on drop rate w)-.11 E(ould also be af)-.11 E
(fected by the presence of preferential)-.275 E
(scheduling for small pack)100.8 277.8 Q(ets, or by dif)-.11 E
(ferential scheduling for pack)-.275 E(ets from dif)-.11 E(ferent \215o)
-.275 E(ws)-.275 E(\(e.g., per)100.8 290.8 Q(-\215o)-.22 E 2.75(ws)-.275
G(cheduling, or dif)-2.75 E(ferential scheduling for UDP and TCP traf)
-.275 E(\214c\).)-.275 E(In man)100.8 307.4 Q 2.75(ye)-.165 G -.44(nv)
-2.75 G(ironments, the drop rate as a function of pack).44 E
(et size will be hea)-.11 E(vily af)-.22 E(fected by)-.275 E(the traf)
100.8 320.4 Q(\214c mix at a particular time.)-.275 E -.165(Fo)5.5 G
2.75(re).165 G(xample, is the traf)-2.915 E(\214c mix dominated by lar)
-.275 E(ge)-.198 E(pack)100.8 333.4 Q(ets, or by smaller ones?)-.11 E
(In some cases, the o)5.5 E -.165(ve)-.165 G(rall pack).165 E
(et drop rate could also af)-.11 E(fect)-.275 E(the relati)100.8 346.4 Q
.33 -.165(ve d)-.275 H(rop rates for dif).165 E(ferent pack)-.275 E
(et sizes.)-.11 E(In wireless netw)100.8 363 Q
(orks, the drop rate as a function of pack)-.11 E
(et size is also determined by the)-.11 E(pack)100.8 376 Q
(et corruption rate as a function of pack)-.11 E(et size.)-.11 E
([Cite Deborah Pinck')5.5 E 2.75(sp)-.605 G(apers on)-2.75 E(Satellite-\
Enhanced Personal Communications Experiments and on Experimental Result\
s)100.8 389 Q(from Internetw)100.8 402 Q(orking Data Applications Ov)
-.11 E(er V)-.165 E(arious W)-1.221 E(ireless Netw)-.44 E
(orks Using a Single)-.11 E(Fle)100.8 415 Q
(xible Error Control Protocol.])-.165 E([Cite the general literature.])
5.5 E(Ef)100.8 431.6 Q(fect on congestion control metrics: The drop rat\
e as a function of pack)-.275 E(et size has a)-.11 E(signi\214cant ef)
100.8 444.6 Q(fect on the performance of congestion control for V)-.275
E(oIP and other small-pack)-1.419 E(et)-.11 E<8d6f>100.8 457.6 Q 2.75
(ws. [Citation:)-.275 F("TFRC for V)2.75 E(oice: the V)-1.419 E(oIP V)
-1.419 E(ariant", draft-ietf-dccp-tfrc-v)-1.221 E(oip-02.txt, and)-.22 E
(earlier papers.])100.8 470.6 Q(The drop rate as a function of pack)5.5
E(et size also has an ef)-.11 E(fect on TCP)-.275 E
(performance, as it af)100.8 483.6 Q(fects the drop rates for TCP')-.275
E 2.75(sS)-.605 G(YN and A)-2.75 E(CK pack)-.44 E 2.75(ets. [Citation:)
-.11 F(Jef)100.8 496.6 Q -.11(fa)-.275 G 2.75(ya).11 G(nd others.])-2.75
E(Measurements: W)100.8 513.2 Q 2.75(en)-.88 G
(eed measurements of the drop rate as a function of pack)-2.75 E
(et size o)-.11 E -.165(ve)-.165 G 2.75(ra).165 G
(wide range of paths, or for a wide range of congested links.)100.8
526.2 Q -.165(Fo)5.5 G 2.75(rt).165 G(ests of relati)-2.75 E .33 -.165
(ve d)-.275 H(rop rates).165 E(on end-to-end pack)100.8 539.2 Q
(ets, one possibility w)-.11 E(ould be to run successi)-.11 E .33 -.165
(ve T)-.275 H(CP connections with).165 E
(200-byte, 512-byte, and 1460-byte pack)100.8 552.2 Q
(ets, and to compare the pack)-.11 E(et drop rates.)-.11 E(The)5.5 E
(ideal test w)100.8 565.2 Q
(ould include running TCP connections on the re)-.11 E -.165(ve)-.275 G
(rse path, to measure the drop).165 E(rates for the small A)100.8 578.2
Q(CK pack)-.44 E(ets on the forw)-.11 E(ard path.)-.11 E(It w)5.5 E
(ould also be useful to)-.11 E(characterize the dif)100.8 591.2 Q
(ference in drop rates for 200-byte TCP pack)-.275 E
(ets and 200-byte UDP)-.11 E(pack)100.8 604.2 Q(ets, e)-.11 E -.165(ve)
-.275 G 2.75(nt).165 G(hough some of this dif)-2.75 E
(ference could be due to the relati)-.275 E .33 -.165(ve b)-.275 H
(urstiness of the)-.055 E(dif)100.8 617.2 Q(ferent connections.)-.275 E
(Ping e)100.8 633.8 Q(xperiments could also be used to get measurements\
 of drop rates as a function size,)-.165 E -.22(bu)100.8 646.8 S 2.75
(ti).22 G 2.75(tw)-2.75 G(ould be necessary to mak)-2.86 E 2.75(es)-.11
G(ure that the ping sending rates were adjusted to be TCP-)-2.75 E
(friendly)100.8 659.8 Q(.)-.715 E([Cite the kno)100.8 676.4 Q
(wn literature on drop rates as a function of pack)-.275 E(et size.])
-.11 E(Our conjecture is that there is a wide range of beha)100.8 693 Q
(viors for this characteristic in the real)-.22 E -.11(wo)100.8 706 S
2.75(rld. Routers).11 F(include Drop-T)2.75 E(ail queues in pack)-.88 E
(ets, bytes, and b)-.11 E(uf)-.22 E(fer sizes in between;)-.275 E
(these will ha)100.8 719 Q .33 -.165(ve q)-.22 H(uite dif).165 E
(ferent drop rates as a function of pack)-.275 E(et size.)-.11 E
(Some routers include)5.5 E(Flo)72 769 Q(yd, K)-.11 E 276.229
(ohler Section)-.385 F 2.75(11. [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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E(RED in byte mode \(the def)
100.8 85 Q(ault for RED in Linux\) and some ha)-.11 E .33 -.165(ve R)
-.22 H(ED in pack).165 E(et mode)-.11 E(\(Cisco, I belie)100.8 98 Q
-.165(ve)-.275 G 2.75(\). This).165 F(also af)2.75 E
(fects drop rates as a function of pack)-.275 E(et size.)-.11 E
(Some routers on congested access links use per)100.8 114.6 Q(-\215o)
-.22 E 2.75(ws)-.275 G 2.75(cheduling. In)-2.75 F
(this case, does the per)2.75 E(-)-.22 E<8d6f>100.8 127.6 Q 2.75(ws)
-.275 G(cheduling ha)-2.75 E .33 -.165(ve t)-.22 H(he goal of f).165 E
(airness in *bytes* per second or in *pack)-.11 E(ets* per second?)-.11
E(What ef)100.8 140.6 Q(fect does the per)-.275 E(-\215o)-.22 E 2.75(ws)
-.275 G(cheduling ha)-2.75 E .33 -.165(ve o)-.22 H 2.75(nt).165 G
(he drop rate as a function of pack)-2.75 E(et size,)-.11 E(for pack)
100.8 153.6 Q(ets in dif)-.11 E(ferent \215o)-.275 E
(ws \(e.g., a small-pack)-.275 E(et V)-.11 E(oIP \215o)-1.419 E 2.75(wc)
-.275 G(ompeting ag)-2.75 E(ainst a lar)-.055 E(ge-)-.198 E(pack)100.8
166.6 Q(et TCP \215o)-.11 E(w\) or for pack)-.275 E
(ets within the same \215o)-.11 E 2.75(w\()-.275 G(small A)-2.75 E
(CK pack)-.44 E(ets and lar)-.11 E(ge data)-.198 E(pack)100.8 179.6 Q
(ets on a tw)-.11 E(o-w)-.11 E(ay TCP connection\).)-.11 E/F1 11
/Times-Bold@0 SF(12.)72 205.6 Q/F2 14/Times-Bold@0 SF(Dr)5.5 E
(op Rates as a Function of Burst Size.)-.252 E F0
(De\214nition: Burst-tolerance, or drop rates as a function of b)100.8
222.2 Q(urst size, can be de\214ned in terms)-.22 E
(of an end-to-end path, or in terms of aggre)100.8 235.2 Q -.055(ga)
-.165 G(te traf).055 E(\214c on a congested link.)-.275 E(The b)100.8
251.8 Q(urst-tolerance of an end-to-end path is de\214ned in terms of c\
onnections with dif)-.22 E(ferent)-.275 E(de)100.8 264.8 Q(grees of b)
-.165 E(urstiness within a round-trip time.)-.22 E(When pack)5.5 E
(ets are sent in b)-.11 E(ursts of N)-.22 E(pack)100.8 277.8 Q
(ets, does the drop rate v)-.11 E(ary as a function of N?)-.275 E -.165
(Fo)5.5 G 2.75(re).165 G(xample, if the TCP sender sends)-2.915 E
(small b)100.8 290.8 Q(ursts of K pack)-.22 E
(ets, for K less than the congestion windo)-.11 E 1.43 -.715(w, h)-.275
H .55 -.275(ow d).715 H(oes the size of K).275 E(af)100.8 303.8 Q
(fect the loss rate?)-.275 E(Similarly)5.5 E 2.75(,f)-.715 G
(or a ping tool sending pings at a certain rate in pack)-2.75 E(ets per)
-.11 E(second, one could see ho)100.8 316.8 Q 2.75(wt)-.275 G
(he clustering of the ping pack)-2.75 E(ets in clusters of size K af)
-.11 E(fects)-.275 E(the pack)100.8 329.8 Q(et drop rate.)-.11 E(As al)
5.5 E -.11(wa)-.11 G(ys with such ping e).11 E(xperiments, it w)-.165 E
(ould be important to)-.11 E
(adjust the sending rate to maintain a longer)100.8 342.8 Q
(-term sending rate that w)-.22 E(as TCP-friendly)-.11 E(.)-.715 E
(Determining f)100.8 359.4 Q(actors: The b)-.11 E
(urst-tolerance is determined lar)-.22 E(gely by the A)-.198 E
(QM mechanisms)-.605 E
(for the congested routers on a path, and by the traf)100.8 372.4 Q
(\214c mix.)-.275 E -.165(Fo)5.5 G 2.75(raD).165 G(rop-T)-2.75 E
(ail queue with only)-.88 E 2.75(as)100.8 385.4 S
(mall number of competing \215o)-2.75 E(ws, the b)-.275 E
(urst-tolerance is lik)-.22 E(ely to be lo)-.11 E 1.43 -.715(w, a)-.275
H(nd for A).715 E(QM)-.605 E(mechanisms where the pack)100.8 398.4 Q
(et drop rate is a function of the a)-.11 E -.165(ve)-.22 G
(rage queue size rather than).165 E(the instantaneous queue size, the b)
100.8 411.4 Q(urst tolerance should be quite high.)-.22 E(Ef)100.8 428 Q
(fect on congestion control metrics: The b)-.275 E
(urst-tolerance of the path or congested link can)-.22 E(af)100.8 441 Q
(fect f)-.275 E(airness between competing \215o)-.11 E(ws with dif)-.275
E(ferent round-trip times; for e)-.275 E(xample,)-.165 E
(Standard TCP \215o)100.8 454 Q(ws with longer round-trip times are lik)
-.275 E(ely to ha)-.11 E .33 -.165(ve a m)-.22 H(ore b).165 E
(ursty arri)-.22 E -.275(va)-.275 G(l).275 E
(pattern at the congested link that that of Standard TCP \215o)100.8 467
Q(ws with shorter round-trip times.)-.275 E(As a result, in en)100.8 480
Q(vironment with lo)-.44 E 2.75(wb)-.275 G
(urst tolerance \(e.g., scenarios with Drop-T)-2.97 E(ail)-.88 E
(queues\), longer)100.8 493 Q
(-round-trip-time TCP connections can see higher pack)-.22 E
(et drop rates than)-.11 E(other TCP connections, and recei)100.8 506 Q
.33 -.165(ve a)-.275 H 2.75(ne).165 G -.165(ve)-3.025 G 2.75(ns).165 G
(maller fraction of the link bandwidth than)-2.75 E(the)100.8 519 Q 2.75
(yw)-.165 G(ould otherwise.)-2.86 E([FJ92] \(Section 3.2\).)5.5 E 1.76
-.88(We n)5.5 H(ote that some TCP traf).88 E(\214c is inherently)-.275 E
-.22(bu)100.8 532 S(rsty).22 E 2.75(,e)-.715 G(.g., Standard TCP withou\
t rate-based pacing, particularly in the presence of)-2.75 E(dropped A)
100.8 545 Q(CK pack)-.44 E(ets or of A)-.11 E(CK compression.)-.44 E
(The b)5.5 E(urst-tolerance of a router can also)-.22 E(af)100.8 558 Q
(fect the delay-throughput tradeof)-.275 E(fs and pack)-.275 E
(et drop rates of the path or of the congested)-.11 E(link.)100.8 571 Q
(Measurements: One could measure the b)100.8 587.6 Q
(urst-tolerance of an end-to-end path by running)-.22 E(successi)100.8
600.6 Q .33 -.165(ve T)-.275 H(CP connections, forcing b).165 E
(ursts of size at least K by dropping an appropriate)-.22 E
(fraction of the A)100.8 613.6 Q(CK pack)-.44 E(ets to the TCP recei)
-.11 E -.165(ve)-.275 G 3.96 -.605(r. A).165 H(lternately).605 E 2.75
(,i)-.715 G 2.75(fo)-2.75 G(ne had control of the TCP)-2.75 E(sender)
100.8 626.6 Q 2.75(,o)-.44 G(ne could modify the TCP sender to send b)
-2.75 E(ursts of K pack)-.22 E(ets when the congestion)-.11 E(windo)
100.8 639.6 Q 2.75(ww)-.275 G(as K or more se)-2.86 E(gments.)-.165 E
([Look at Cro)100.8 656.2 Q -.165(ve)-.165 G(lla').165 E 2.75(sp)-.605 G
(aper on loss pairs.])-2.75 E
([Look at: M. Allman and E. Blanton, "Notes on Burst Mitig)100.8 672.8 Q
(ation for T)-.055 E(ransport Protocols",)-.385 E -.44(AC)100.8 685.8 S
2.75(MC).44 G(omputer Communication Re)-2.75 E(vie)-.275 E 1.43 -.715
(w, v)-.275 H(ol. 35\(2\), \(2005\).]).495 E
(Making inferences about the A)100.8 702.4 Q
(QM mechanism for the congested router on an end-to-end)-.605 E
(path: One potential use of measurement tools for determining the b)
100.8 715.4 Q(urst-tolerance of an end-)-.22 E(Flo)72 769 Q(yd, K)-.11 E
270.729(ohler Section)-.385 F 2.75(12. [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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E(to-end path w)100.8 85 Q
(ould be to mak)-.11 E 2.75(ei)-.11 G
(nferences about the presence or absence of an A)-2.75 E(QM)-.605 E
(mechanism at the congested link or links.)100.8 98 Q
(As a simple test, one could run a TCP)5.5 E
(connection until the connection comes out of slo)100.8 111 Q 2.75
(w-start. If)-.275 F(the recei)2.75 E .33 -.165(ve w)-.275 H(indo).165 E
2.75(wo)-.275 G 2.75(ft)-2.75 G(he TCP)-2.75 E(connection w)100.8 124 Q
(as suf)-.11 E(\214ciently high that the connection e)-.275 E(xited slo)
-.165 E(w-start with pack)-.275 E(et drops or)-.11 E
(marks instead of because of the limitation of the recei)100.8 137 Q .33
-.165(ve w)-.275 H(indo).165 E 1.43 -.715(w, o)-.275 H
(ne could record the).715 E(congestion windo)100.8 150 Q 2.75(wa)-.275 G
2.75(tt)-2.75 G(he end of slo)-2.75 E(w-start, and the number of pack)
-.275 E(ets dropped from this)-.11 E(windo)100.8 163 Q 4.18 -.715(w. A)
-.275 H(high pack)3.465 E
(et drop rate might be more typical of a Drop-T)-.11 E
(ail queue with small-)-.88 E(scale statistical multiple)100.8 176 Q
(xing on the congested link, and a single pack)-.165 E
(et drop coming out of)-.11 E(slo)100.8 189 Q(w-start w)-.275 E
(ould suggest an A)-.11 E(QM mechanism at the congested link.)-.605 E
(The synchronization measure could also add information about the lik)
100.8 205.6 Q(ely presence or)-.11 E(absence of A)100.8 218.6 Q(QM on t\
he congested link\(s\) of an end-to-end path, with paths with higher)
-.605 E(le)100.8 231.6 Q -.165(ve)-.275 G
(ls of synchronization being more lik).165 E(ely to ha)-.11 E .33 -.165
(ve D)-.22 H(rop-T).165 E(ail queues with small-scale)-.88 E
(statistical multiple)100.8 244.6 Q(xing on the congested link\(s\).)
-.165 E([Cite the rele)100.8 261.2 Q -.275(va)-.275 G
(nt literature about tools for determining the A).275 E
(QM mechanism on an end-to-)-.605 E(end path.])100.8 274.2 Q/F1 11
/Times-Bold@0 SF(13.)72 300.2 Q/F2 14/Times-Bold@0 SF(Dr)5.5 E
(op Rates as a Function of Sending Rate.)-.252 E F0(De\214nition: Drop \
rates as a function of sending rate is de\214ned in terms of the drop b\
eha)100.8 316.8 Q(vior)-.22 E(of a \215o)100.8 329.8 Q 2.75(wi)-.275 G
2.75(nt)-2.75 G(he end-to-end path.)-2.75 E
(That is, does the sending rate of an indi)5.5 E(vidual \215o)-.275 E
2.75(wa)-.275 G -.275(ff)-2.75 G(ect).275 E(its o)100.8 342.8 Q(wn pack)
-.275 E(et drop rate, or its pack)-.11 E(et drop rate lar)-.11 E
(gely independent of the sending rate of)-.198 E(the \215o)100.8 355.8 Q
(w?)-.275 E(Determining f)100.8 372.4 Q
(actors: The sending rate of the \215o)-.11 E 2.75(wa)-.275 G -.275(ff)
-2.75 G(ects its o).275 E(wn pack)-.275 E(et drop rate in an)-.11 E(en)
100.8 385.4 Q(vironment with small-scale statistical multiple)-.44 E
(xing on the congested link.)-.165 E(The pack)5.5 E(et)-.11 E
(drop rate is lar)100.8 398.4 Q
(gely independent of the sending rate in an en)-.198 E
(vironment with lar)-.44 E(ge-scale)-.198 E(statistical multiple)100.8
411.4 Q(xing, with man)-.165 E 2.75(yc)-.165 G(ompeting small \215o)
-2.75 E(ws at the congested link.)-.275 E(Thus, the)5.5 E(beha)100.8
424.4 Q(vior of drop rates as a function of sending rate is a rough mea\
sure of the le)-.22 E -.165(ve)-.275 G 2.75(lo).165 G(f)-2.75 E
(statistical multiple)100.8 437.4 Q
(xing on the congested links of an end-to-end path.)-.165 E(Ef)100.8 454
Q(fect on congestion control metrics: The le)-.275 E -.165(ve)-.275 G
2.75(lo).165 G 2.75(fs)-2.75 G(tatistical multiple)-2.75 E
(xing at the congested)-.165 E(link can af)100.8 467 Q(fect the perform\
ance of congestion control mechanisms in transport protocols.)-.275 E
-.165(Fo)100.8 480 S 2.75(re).165 G
(xample, delay-based congestion control is often better suited to en)
-2.915 E(vironments small-)-.44 E(scale statistical multiple)100.8 493 Q
(xing at the congested link, where the transport protocol responds to)
-.165 E(the delay caused by its o)100.8 506 Q(wn sending rate.)-.275 E
(Measurements: In a simulation or testbed, the le)100.8 522.6 Q -.165
(ve)-.275 G 2.75(lo).165 G 2.75(fs)-2.75 G(tatistical multiple)-2.75 E
(xing on the)-.165 E(congested link can be observ)100.8 535.6 Q
(ed directly)-.165 E 5.5(.I)-.715 G 2.75(nt)-5.5 G(he Internet, the le)
-2.75 E -.165(ve)-.275 G 2.75(lo).165 G 2.75(fs)-2.75 G
(tatistical multiple)-2.75 E(xing)-.165 E(on the congested links of an \
end-to-end path can be inferred indirectly through per)100.8 548.6 Q
(-\215o)-.22 E(w)-.275 E(measurements, by observing whether the pack)
100.8 561.6 Q(et drop rate v)-.11 E(aries as a function of the)-.275 E
(sending rate of the \215o)100.8 574.6 Q -.715(w.)-.275 G F1(14.)72
600.6 Q F2(Congestion Contr)5.5 E(ol Mechanisms f)-.252 E(or T)-.35 E
(raf\214c, along with Sender and)-1.036 E F0(Recei)100.8 613.6 Q -.165
(ve)-.275 G 2.75(rB).165 G(uf)-2.75 E(fer Sizes.")-.275 E(Ef)100.8 630.2
Q(fect on congestion control metrics: Please don')-.275 E 2.75(te)-.198
G -.275(va)-3.025 G(luate A).275 E(QM mechanisms by using)-.605 E
(Reno TCP)100.8 643.2 Q 2.75(,o)-1.221 G 2.75(re)-2.75 G -.275(va)-3.025
G(luate ne).275 E 2.75(wt)-.275 G
(ransport protocols by comparing them with the performance of)-2.75 E
(Reno TCP!)100.8 656.2 Q -.165(Fo)5.5 G 2.75(ra).165 G 2.75(ne)-2.75 G
(xplanation, see [FK02] \(Section 3.4\).)-2.915 E 2.75
(Measurements: See)100.8 672.8 R([MAF05].)2.75 E(Flo)72 769 Q(yd, K)-.11
E 270.729(ohler Section)-.385 F 2.75(14. [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 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E/F1 11/Times-Bold@0 SF(15.)72
85 Q/F2 14/Times-Bold@0 SF(Characterization of Congested Links in T)5.5
E(erms of Band)-1.288 E(width and)-.21 E -1.036(Ty)72 98 S(pical Le)
1.036 E -.14(ve)-.21 G(ls of Congestion).14 E F0
([Pointers to the current state of our kno)100.8 114.6 Q(wledge.])-.275
E F1(16.)72 140.6 Q F2(Characterization of Challenging Lo)5.5 E(wer Lay)
-.14 E(ers.)-.14 E F0
([This will just be a short set of pointers to the rele)100.8 157.2 Q
-.275(va)-.275 G(nt literature, which is quite e).275 E(xtensi)-.165 E
-.165(ve)-.275 G(.]).165 E F1(17.)72 183.2 Q F2
(Characterization of Netw)5.5 E(ork Changes Affecting Congestion)-.14 E
F0([Pointers to the current state of our kno)100.8 199.8 Q(wledge.])
-.275 E F1(18.)72 225.8 Q F2(Using the T)5.5 E(ools Pr)-1.288 E
(esented in this Document)-.252 E F0([T)100.8 242.4 Q 2.75(ob)-.88 G
2.75(ed)-2.75 G(one.])-2.75 E F1(19.)72 268.4 Q F2(Related W)5.5 E(ork)
-1.05 E F0([Cite "On the Ef)100.8 285 Q(fecti)-.275 E .33 -.165(ve E)
-.275 H -.275(va).165 G(luation of TCP" by Allman and F).275 E(alk.])
-.165 E F1(20.)72 311 Q F2(Conclusions)5.5 E F0([T)100.8 327.6 Q 2.75
(ob)-.88 G 2.75(ed)-2.75 G(one.])-2.75 E F1(21.)72 353.6 Q F2
(Security Considerations)5.5 E F0
(There are no security considerations in this document.)100.8 370.2 Q F1
(22.)72 396.2 Q F2(IAN)5.5 E 3.5(AC)-.28 G(onsiderations)-3.5 E F0
(There are no IAN)100.8 412.8 Q 2.75(Ac)-.385 G
(onsiderations in this document.)-2.75 E F1(23.)72 438.8 Q F2(Ackno)5.5
E(wledgements)-.14 E F0(Thanks to Xiaoliang \(Da)100.8 455.4 Q(vid\) W)
-.22 E(ei for feedback and contrib)-.88 E(utions to this document.)-.22
E F2(Inf)72 481.4 Q(ormati)-.35 E .28 -.14(ve R)-.14 H(efer).14 E(ences)
-.252 E F0([RFC 2119] S. Bradner)100.8 498 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.).165 E(RFC 2119.)129.6 511 Q([MA)100.8
527.6 Q(WI] M.W)-.99 E 2.75(.G)-1.012 G(roup, Ma)-2.75 E(wi w)-.165 E
(orking group traf)-.11 E(\214c archi)-.275 E -.165(ve)-.275 G 2.75(,U)
.165 G(RL)-2.75 E("http://tracer)129.6 540.6 Q(.csl.son)-.605 E -.715
(y.)-.165 G(jp/ma).715 E(wi/".)-.165 E([AKM04] B. Appenzeller)100.8
557.2 Q 2.75(,I)-.44 G 2.75(.K)-2.75 G(eslassy)-3.025 E 2.75(,a)-.715 G
(nd N. McK)-2.75 E(eo)-.275 E(wn, Sizing Router Buf)-.275 E(fers,)-.275
E(SIGCOMM 2004.)129.6 570.2 Q([AKSJ03] J. Aikat, J. Kaur)100.8 586.8 Q
2.75(,F)-.44 G(.D. Smith, and K. Jef)-3.63 E -.11(fa)-.275 G 1.43 -.715
(y, V).11 H(ariability in TCP Roundtrip T)-.506 E(imes,)-.385 E -.44(AC)
129.6 599.8 S 2.75(MS).44 G
(IGCOMM Internet Measurement Conference, Maimi, FL, October 2003, pp.)
-2.75 E(279-284.)129.6 612.8 Q([A00] M. Allman, A W)100.8 629.4 Q
(eb Serv)-.88 E(er')-.165 E 2.75(sV)-.605 G(ie)-3.41 E 2.75(wo)-.275 G
2.75(ft)-2.75 G(he T)-2.75 E(ransport Layer)-.385 E 2.75(,C)-.44 G
(omputer Communication)-2.75 E(Re)129.6 642.4 Q(vie)-.275 E 1.43 -.715
(w, 3)-.275 H(0\(5\), October 200.).715 E([BH02] F)100.8 659 Q 2.75(.B)
-.88 G(accelli and D. Hong, AIMD, F)-2.75 E
(airness and Fractal Scaling of TCP T)-.165 E(raf)-.385 E(\214c,)-.275 E
(Infocom 2002.)129.6 672 Q([CBC95] C. Cunha, A. Besta)100.8 688.6 Q
(vros, and M. Cro)-.22 E -.165(ve)-.165 G
(lla, "Characteristics of WWW Client-).165 E(based T)129.6 701.6 Q
(races", B)-.385 E 2.75(UT)-.11 G(echnical Report B)-3.52 E
(UCS-95-010, 1995.)-.11 E(Flo)72 769 Q(yd, K)-.11 E 325.729(ohler [P)
-.385 F(age 12])-.165 E 0 Cg EP
%%Page: 13 13
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E([Dummynet] L. Rizzo, Dummynet\
, URL "http://info.iet.unipi.it/~luigi/ip_dummynet/".)100.8 85 Q
(BIBREF F02 "F02" C. J. Fraleigh, Pro)100.8 101.6 Q
(visioning Internet Backbone Netw)-.165 E(orks to Support)-.11 E(Latenc)
129.6 114.6 Q 2.75(yS)-.165 G(ensiti)-2.75 E .33 -.165(ve A)-.275 H 2.75
(pplications. PhD).165 F(thesis, Stanford Uni)2.75 E -.165(ve)-.275 G
(rsity).165 E 2.75(,D)-.715 G(epartment of)-2.75 E
(Electrical Engineering, June 2002.)129.6 127.6 Q([FJ92] S. Flo)100.8
144.2 Q(yd and V)-.11 E 2.75(.J)-1.419 G(acobson, On T)-2.75 E(raf)-.385
E(\214c Phase Ef)-.275 E(fects in P)-.275 E(ack)-.165 E
(et-Switched Gate)-.11 E -.11(wa)-.275 G(ys,).11 E(Internetw)129.6 157.2
Q(orking: Research and Experience, V)-.11 E
(.3 N.3, September 1992, p.115-156.)-1.419 E([FK02] S. Flo)100.8 173.8 Q
(yd and E. K)-.11 E(ohler)-.385 E 2.75(,I)-.44 G
(nternet Research Needs Better Models, Hotnets-I, October)-2.75 E(2002.)
129.6 186.8 Q([GM04] L. Grieco and S. Mascolo, Performance Ev)100.8
203.4 Q(aluation and Comparison of)-.275 E -.88(We)129.6 216.4 S(stw).88
E(ood+, Ne)-.11 E 2.75(wR)-.275 G(eno, and V)-2.75 E -2.475 -.165(eg a)
-1.221 H 2.75(sT).165 G(CP Congestion Control, CCR, April 2004.)-2.75 E
([HMTG01] C. Hollot, V)100.8 233 Q 2.75(.M)-1.419 G(isra, D. T)-2.75 E
-.275(ow)-.88 G(sle).275 E 1.43 -.715(y, a)-.165 H(nd W).715 E 2.75(.G)
-1.012 G(ong, On Designing Impro)-2.75 E -.165(ve)-.165 G(d).165 E
(Controllers for A)129.6 246 Q(QM Routers Supporting TCP Flo)-.605 E
(ws, IEEE Infocom, 2001.)-.275 E([IMD01] G. Iannaccone, M. May)100.8
262.6 Q 2.75(,a)-.715 G(nd C. Diot, Aggre)-2.75 E -.055(ga)-.165 G(te T)
.055 E(raf)-.385 E(\214c Performance with Acti)-.275 E -.165(ve)-.275 G
(Queue Management and Drop From T)129.6 275.6 Q 2.75(ail. SIGCOMM)-.88 F
(Comput. Commun. Re)2.75 E -.715(v.)-.275 G(,).715 E
(31\(3\):4-13, 2001.)129.6 288.6 Q([JD02] H. Jiang and C. Do)100.8 305.2
Q(vrolis, P)-.165 E(assi)-.165 E .33 -.165(ve E)-.275 H
(stimation of TCP Round-trip T).165 E(imes, Computer)-.385 E
(Communication Re)129.6 318.2 Q(vie)-.275 E 1.43 -.715(w, 3)-.275 H
(2\(3\), July 2002.).715 E([MAF05] A. Medina, M. Allman, and A. Flo)
100.8 334.8 Q 2.75(yd. Measuring)-.11 F(the Ev)2.75 E(olution of T)-.22
E(ransport)-.385 E(Protocols in the Internet.)129.6 347.8 Q
(Computer Communication Re)5.5 E(vie)-.275 E 1.43 -.715(w, A)-.275 H
(pril 2005.).715 E([NISTNet] NIST Net, URL "http://snad.ncsl.nist.go)
100.8 364.4 Q(v/itg/nistnet/".)-.165 E
([PJD04] R. Prasad, M. Jain, and C. Do)100.8 381 Q(vrolis, On the Ef)
-.165 E(fecti)-.275 E -.165(ve)-.275 G(ness of Delay-Based).165 E
(Congestion A)129.6 394 Q -.22(vo)-.814 G
(idance, PFLDnet 2004, February 2004.).22 E([QZK01] L. Qiu, Y)100.8
410.6 Q 2.75(.Z)-1.419 G(hang, and S. K)-2.75 E(esha)-.275 E 1.43 -.715
(v, U)-.22 H(nderstanding the Performance of Man).715 E 2.75(yT)-.165 G
(CP)-2.75 E(Flo)129.6 423.6 Q(ws, Comput. Netw)-.275 E
(orks, 37\(3-4\):277-306, 2001.)-.11 E([RSB01] R. Riedi, S. Sarv)100.8
440.2 Q(otham, and R. V)-.22 E(araniuk, Connection-le)-1.221 E -.165(ve)
-.275 G 2.75(lA).165 G(nalysis and)-2.75 E(Modeling of Netw)129.6 453.2
Q(ork T)-.11 E(raf)-.385 E(\214c, SIGCOMM Internet Measurement W)-.275 E
(orkshop, 2001.)-.88 E([R)100.8 482.8 Q(TW05] G. Raina, D. T)-.66 E
-.275(ow)-.88 G(sle).275 E 1.43 -.715(y, a)-.165 H(nd D. W).715 E
(ischik, Control Theory for Buf)-.44 E(fer Sizing, CCR,)-.275 E
(July 2005.)129.6 495.8 Q([TG] T)100.8 512.4 Q(raf)-.385 E
(\214c Generators for Internet T)-.275 E(raf)-.385 E(\214c W)-.275 E
(eb P)-.88 E(age, URL)-.165 E("http://www)129.6 525.4 Q(.icir)-.715 E
(.or)-.605 E(g/models/traf)-.198 E(\214cgenerators.html".)-.275 E([U)
100.8 542 Q(A01] U. of Auckland, Auckland-vi trace data, June 2001.)-.44
E(URL)5.5 E("http://w)129.6 555 Q(ans.cs.w)-.11 E(aikato.ac.nz/w)-.11 E
(and/wits/auck/6/".)-.11 E([UW02] UW)100.8 571.6 Q(-Madison, Netw)-.715
E(ork Performance Statistics, October 2002.)-.11 E(URL)5.5 E
("http://wwwstats.net.wisc.edu/".)129.6 584.6 Q([WM05] D. W)100.8 601.2
Q(ischik and N. McK)-.44 E(eo)-.275 E(wn, Buf)-.275 E
(fer sizes for Core Routers, CCR, July 2005.)-.275 E
([ZSC91] L. Zhang, S. Shenk)100.8 617.8 Q(er)-.11 E 2.75(,a)-.44 G
(nd D.D. Clark, Observ)-2.75 E(ations and Dynamics of a)-.275 E
(Congestion Control Algorithm: the Ef)129.6 630.8 Q(fects of T)-.275 E
-.11(wo)-.88 G(-w).11 E(ay T)-.11 E(raf)-.385 E(\214c, SIGCOMM 1991.)
-.275 E/F1 14/Times-Bold@0 SF(Editors' Addr)72 669.8 Q(esses)-.252 E F0
(Sally Flo)100.8 686.4 Q(yd <\215o)-.11 E(yd@icir)-.11 E(.or)-.605 E(g>)
-.198 E(ICSI Center for Internet Research)100.8 699.4 Q
(1947 Center Street, Suite 600)100.8 712.4 Q(Berk)100.8 725.4 Q(ele)-.11
E 1.43 -.715(y, C)-.165 H 2.75(A9).715 G(4704)-2.75 E(Flo)72 769 Q
(yd, K)-.11 E 325.729(ohler [P)-.385 F(age 13])-.165 E 0 Cg EP
%%Page: 14 14
%%BeginPageSetup
BP
%%EndPageSetup
/F0 11/Times-Roman@0 SF(INTERNET)72 49 Q 66.092(-DRAFT Expires:)-1.012 F
(December 2006)2.75 E(June 2006)116.406 E(USA)100.8 85 Q(Eddie K)100.8
101.6 Q(ohler <k)-.385 E(ohler@cs.ucla.edu>)-.11 E(4531C Boelter Hall)
100.8 114.6 Q(UCLA Computer Science Department)100.8 127.6 Q
(Los Angeles, CA 90095)100.8 140.6 Q(USA)100.8 153.6 Q/F1 14
/Times-Bold@0 SF(Full Copyright Statement)72 179.6 Q F0(Cop)100.8 196.2
Q(yright \(C\) The Internet Society 2006.)-.11 E
(This document is subject to the rights, licenses)5.5 E
(and restrictions contained in BCP 78, and e)100.8 209.2 Q
(xcept as set forth therein, the authors retain all)-.165 E
(their rights.)100.8 222.2 Q
(This document and the information contained herein are pro)100.8 238.8
Q(vided on an "AS IS" basis and)-.165 E(THE CONTRIB)100.8 251.8 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 AND THE INTERNET)100.8
264.8 Q(ENGINEERING T)100.8 277.8 Q(ASK FORCE DISCLAIM ALL W)-1.023 E
(ARRANTIES, EXPRESS OR)-1.32 E(IMPLIED, INCLUDING B)100.8 290.8 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 USE)-2.75 E(OF THE INFORMA)
100.8 303.8 Q(TION HEREIN WILL NO)-1.221 E 2.75(TI)-.44 G
(NFRINGE ANY RIGHTS OR ANY)-2.75 E(IMPLIED W)100.8 316.8 Q
(ARRANTIES OF MERCHANT)-1.32 E(ABILITY OR FITNESS FOR A)-1.023 E -1.012
(PA)100.8 329.8 S -.66(RT)1.012 G(ICULAR PURPOSE.).66 E F1
(Intellectual Pr)72 355.8 Q(operty)-.252 E F0(The IETF tak)100.8 372.4 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 claimed to pertain to the \
implementation or use of the)100.8 385.4 Q
(technology described in this document or the e)100.8 398.4 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)100.8 411.4 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)100.8 424.4 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.)100.8 437.4 Q
(Copies of IPR disclosures made to the IETF Secretariat and an)100.8 454
Q 2.75(ya)-.165 G(ssurances of licenses to)-2.75 E(be made a)100.8 467 Q
-.275(va)-.22 G(ilable, or the result of an attempt made to obtain a ge\
neral license or).275 E(permission for the use of such proprietary righ\
ts by implementers or users of this)100.8 480 Q(speci\214cation can be \
obtained from the IETF on-line IPR repository at)100.8 493 Q(http://www)
100.8 506 Q(.ietf.or)-.715 E(g/ipr)-.198 E(.)-.605 E(The IETF in)100.8
522.6 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)100.8
535.6 Q -.165(ve)-.165 G 2.75(rt).165 G(echnology that may be)-2.75 E
(required to implement this standard.)100.8 548.6 Q
(Please address the information to the IETF at ietf-)5.5 E(ipr@ietf.or)
100.8 561.6 Q(g.)-.198 E(Flo)72 769 Q(yd, K)-.11 E 325.729(ohler [P)
-.385 F(age 14])-.165 E 0 Cg EP
%%Trailer
end
%%EOF