Global fairness of additive-increase and multiplicative-decrease with heterogeneous round-trip times
Milan Vojnovic,J.-Y. Le Boudec,Catherine Boutremans +2 more
- Vol. 3, pp 1303-1312
TLDR
It is shown that the source rates tend to be distributed in order to maximize an objective function called F/sub A//sup h/ ("F/ sub A//Sup h/ fairness"), which provides some insight into the distribution of rates, and hence of packet loss ratios, which can be expected in a given network with a number of competing TCP or TCP-friendly sources.Abstract:
Consider a network with an arbitrary topology and arbitrary communication delays, in which congestion control is based on additive-increase and multiplicative-decrease. We show that the source rates tend to be distributed in order to maximize an objective function called F/sub A//sup h/ ("F/sub A//sup h/ fairness"). We derive this result under the assumption of rate proportional negative feedback and for the regime of rare negative feedback. This applies to TCP in moderately loaded networks, and to those TCP implementations that are designed to interpret multiple packet losses within one RTT as a single congestion indication and do not rely on re-transmission timeout. This result provides some insight into the distribution of rates, and hence of packet loss ratios, which can be expected in a given network with a number of competing TCP or TCP-friendly sources. We validate our findings by analyzing a multiple-bottleneck scenario, and comparing with previous results (Floyd, 1991, Mathis et al, 1997) and an extensive numerical simulation with realistic parameter settings. We apply F/sub A//sup h/ fairness to gain a more accurate understanding of the bias of TCP against long round-trip times.read more
Globalfairnessofadditive–increaseand
multiplicative–decreasewithheterogeneous
round–triptimes
MilanVojnovi´c,Jean-YvesLeBoudec,andCatherineBoutremans
InstituteforComputerCommunicationsandApplications
SwissFederalInstituteofTechnologyatLausanne(EPFL)
CH-1015Lausanne,Switzerland
mvojnovi@epfl.ch SSC Technical Report SSC/1999/024
Abstract—Consideranetworkwithanarbitrarytopology
andarbitrarycommunicationdelays,inwhichcongestion
controlisbasedonadditive–increaseandmultiplicative–
decrease.Weshowthatthesourceratestendtobedis-
tributedinordertomaximizeanobjectivefunctioncalled
F
h
A
(“
F
h
A
fairness”).Wederivethisresultundertheas-
sumptionofrateproportionalnegativefeedbackandforthe
regimeofrarenegativefeedback.ThisappliestoTCPin
moderatelyloadednetworks,andtothoseTCPimplemen-
tationsthataredesignedtointerpretmultiplepacketlosses
withinoneRTTasasinglecongestionindicationanddonot
relyonre-transmissiontimeout.Thisresultprovidessome
insightintothedistributionofrates,andhenceofpacket
lossratios,whichcanbeexpectedinagivennetworkwith
anumberofcompetingTCPorTCP-friendlysources.We
validateourfindingsbyanalyzingtheparkinglotscenario,
andcomparingwithpreviousresults[1],[2],andanexten-
sivenumericalsimulationwithrealisticparametersettings.
Weapply
F
h
A
fairnesstogainamoreaccurateunderstanding
ofthebiasofTCPagainstlongroundtriptimes.
Keywords—Additive–Increase,Multiplicative–Decrease,
Fairness,Best–Effort,TCP,TCP–Friendly,TCPthroughput-
lossformula,RTT,parking–lot,StochasticApproximation,
ODE,Lyapunov.
I.INTRODUCTION
Thereisacontinuinginterestonthroughputandfair-
nessissuesofTCP[3]congestionavoidance.Thisinterest
isparticularlynourishedbytheproliferationofreal–time
“stream”applicationsovertheInternet(e.g.voice,video)
forwhichitisrequiredtobeTCP–Friendly,i.e.tofairly
coexistwithalreadyexistingTCPapplications.
Inoneofthepioneeringworks,ChiuandJain[4]formu-
latedasetofbasicprinciplesoftheadditive–increaseand
multiplicative–decreasecongestionavoidancetoachieve
efficiencyandfairness,byanalyzingthesimplemodelof
asinglebottleneck.
In[5]Kelly,Maulloo,andTanshowedthatalarge-
scalenetworkdeployingsomespecificformofadditive–
increaseandmultiplicative–decreasecongestionavoid-
ancetendstodistributeratesaccordingtoproportionalfair-
ness.Thisresultiscommonlymisinterpretedasbeingap-
plicabletocongestionavoidanceintheInternetwithTCP.
Recently,Hurley,LeBoudec,andThiran[6]showed
thatinanetworkemployingadditive–increaseandmulti-
plicative–decrease,thesourceratestendtobedistributed
inordertomaximizeanobjectivefunctioncalled
F
A
.The
authorscallthis“
F
A
fairness”.Thisresultisobtained
bythelimitmeanordinarydifferentialequation(ODE)
method,foranetworkoperatingintheregimeofrareneg-
ativefeedback.Thepivotalassumptionofthatworkis
therateproportionalnegativefeedback,whichtheauthors
claimtobemorerealisticthanonewhichdependsexclu-
sivelyontheoverallload[5].However,theresultisre-
strictedtothehomogeneousround–triptime(RTT)case
wheretheratesareupdatedsynchronously.
Inthispaper,weextendthemodelingof[6]tothehet-
erogeneousRTTcase.Ourresultisageneralizationof
F
A
fairness,whichwecall
F
h
A
fairness.Itgivesthedistribu-
tionofratesinaarbitrarynetworkemployingtheadditive–
increaseandmultiplicative–decreasemethodforconges-
tioncontrol,withtheassumptionthatnegativefeedback
israre.Weallowtheround–triptimestodifferfromone
sourcetoanother.Theratestendtomaximizeanobjec-
tivefunctioncalled
F
h
A
,whoseparametersreflecttherate
adaptationalgorithm.Tothebestofourknowledge,this
isthefirstgeneralresultencompassingmanyoftherele-
vantsystemparameters,applicabletoanarbitrarynetwork
topologywithmultiplebottlenecks.Ourresultsallowsto
findafirstorderapproximationofratedistributions;com-
binedwithaloss-throughputformulasuchas[2],[7],this
givesapredictionofthelossrates.Extensivesimulation
resultsconfirmthesepredictions.
Thenoveltyofourapproachisanapplicationofthe
recentweakconvergenceresultsofdecentralizedasyn-
chronousstochasticapproximationalgorithms[8].Our
1
modelessentiallydiffersfrom[6]inthatwedonotas-
sumethatrate–adaptationisperformedsynchronouslyby
allsources;incontrast,weuseanasynchronousmodel
whereeverysourceupdatesitsratebasedonitsownround
triptimeinterval.Unlikethesynchronousmodelin[4],
[5]or[6],thisallowsustoaddressthecasewithdifferent
roundtriptimes.Buteveninthecasewhereallroundtrip
timesareequal,thisgivesamoreaccuratemodel.Indeed,
withthesynchronousmodel,rateadjustmentisbasedon
themostrecentpreviousrates.Inreality,thefeedbackre-
ceivedbyonesourceattheendofoneroundtriptimein-
tervaldependsontheratesduringthepreviousinterval,
shiftedintimebythedelayrequiredforfeedbacktoreach
thesources.Thesynchronousmodelassumesimplicitly
thatfeedbackreachessourcesinstantaneously.Wecallthis
assumption“stolenlag”.Weshowwithourmodelingthat
thestolenlagassumptiondoesnotaffectthedistribution
ofaveragerates;bysimulation,weseehoweverthatitaf-
fectstheamplitudeofoscillations.Notethatourmodel
explicitlyconsidersallcommunicationdelays.
Weassumeinthispaperthatthenegativefeedbackre-
ceivedbysourcesisrare,andisproportionaltothesource
rate.Therarenegativefeedbackassumptionisvalidina
reasonablyloadednetwork;theproportionalassumption
shouldbetruewithactivequeuemanagement[9](e.g.
RED[10])appliedtootherwiseFIFOqueues.Inaddition,
ourmodelassumesasinglerateupdatingperRTT;thisfits
withTCPimplementationsdesignedtocopewithmulti-
plepacketlosseswithinsingleRTT,i.e.thattreatmultiple
packetlosseswithinoneRTTasasinglecongestionsignal,
andavoidre-transmissiontimeouts.
Ourmodeldoesnotincorporatetheeffectofthevari-
ationofRTTforonegivensourcefromonefeedbackin-
tervaltotheother.Itisknownthat,foranetworkwith
fixedwindows[11],thevariationofroundtriptimesdueto
queuesbuildinguphasinitselfacongestionavoidanceef-
fect,whichisnotcapturedbyourmodeling.Anotherlimi-
tationisthatweassumetheratestobepiecewiseconstant,
i.e.tobeadjustedonlyonceperroundtriptime.Thus,the
effectofburstinessatthetimescaleoftheroundtriptimeis
nottakenintoaccount.Incontrast,ourstudycapturesthe
effectofthewindoworrateadaptationmechanismfound
forexamplewithTCPorABR.Ourresultsmaybeusedas
areferencefairnessmeasureinperformanceevaluationsof
TCP–friendlyrateadjustmentalgorithms.
Inthenextsubsectionweoutlineourmainresults.
A.SummaryoftheMainResults
Weconsideranetworkwithmultiplebottlenecksand
heterogeneousround–triptimes.Then,underthecondi-
tionthatthereisnosubstantialqueuingdelayvariation,
andthenetworkisoperatingintheregimeoftherareneg-
ativefeedback,thecollectionofrates
x
=(
x
1
;:::;x
i
;:::
)
isdistributedsuchthat
x
maximizestheobjectivefunction
F
h
A
(
x
)=
X
i
2S
1
i
log
x
i
r
i
+
i
x
i
;
subjecttotheconstraints
P
j
2S
A
l
;
j
x
j
c
l
,
8
l
2L
.Inthe
formula,
S
isthesetofsources,
L
thesetoflinks,
A
l;i
the
routingmatrix(
A
l;i
is
0
or
1
),
c
l
thecapacityoflink
l
,and
i
istheRTTforfloworsource
i
.Thereisoneflowper
source.Therateadaptationparametersare
r
i
(additive–
increaseelement)and
i
(multiplicative–decreasefactor);
theymaydependonsource
i
.
Theaboveresultisappliedtotheparking–lotnetwork
topology;weobtainaclosed-formforthedistributionof
rates.Thisallowsustoverifytheconsistencyofourre-
sultwithexistingworkandwithconductedsimulations.
Wefindthattheresultsin[1]areanasymptoticcase
of
F
h
A
fairnessforsmalladditive–increase/multiplicative–
decreaseratiorelativetoconnectionthroughput.
Wealsogainamoreaccurateunderstandingofthebias
ofTCPagainstlongroundtriptimes.Wepointoutthatit
isimportanttomakethedifferencebetweenabiasagainst
longRTTs(perhapsanundesirablefeature)andabias
againstflowswithmanyhops(perhapsadesiredfeature).
Weseethatthebiasagainstflowswithmanyhopsisinthe
natureofanyrateadaptationalgorithmbasedonadditive–
increaseandmultiplicative–decrease.Incontrast,abias
againstlongRTTscanbeattenuatedwithcorrectionssuch
asmentionedin[1]and[12].Finally,wealsoconfirm
throughputlossformulas,withinthelimitationsofour
modeling.
B.OutlineofthePaper
Thepaperisorganizedasfollows.InSectionII,the
mainresultsarederived.Followingthebasicmodeldef-
initions,feedbackmodelingisdescribedinmoredetail.
Then,asymptoticconvergenceresultsofthedecentralized
asynchronousstochasticapproximationalgorithms[8]are
sketched.Intherestofthesection,objectivefunction
F
h
A
ofthealgorithmofconcernisderivedandanalyzed.In
SectionIII,
F
h
A
resultisappliedtotheparking–lotnetwork
topologyforwhichaclosed-formratedistributioniscom-
puted,andresultsareverifiedthroughnumericalsimula-
tion.InSectionIV,theresultsarediscussedandcompared
totherelatedpreviouswork.Implicationsoftheresultto
theInternetareaddressedinSectionV.InSectionVI,
concludingremarksaregiven.InAppendixAwegivethe
maintheoremoftheunderlyingtheory[8].
2
l;i
l
:
c
l
;
l
i;l
S
i
D
i
Fig.1.Anillustrationofthedefineddelays.
II.DERIVATIONOFTHEMAINRESULTS
A.ModelSetup
Thenotationisdevelopedasfollows.Letset
L
contain
networklinks.Thenlet
c
l
and
l
bethecapacityanddelay
oflink
l
2L
,respectively.Letset
S
comprisesources
(flows)thatareactiveonthegivennetwork.Therout-
ingsettingwedescribebyroutingmatrix
A
=(
A
l;i
;l
2
L
;
i
2S
)
,suchthat
A
l;i
=1
,ifflow
i
traverseslink
l
,and
A
l;i
=0
,otherwise.
1
Further,wedefinecommunicationdelays.Let
i;l
de-
notedelayfromsource
i
tolink
l
,andlet
l;i
bethedelay
fromlink
l
tosource
i
.Then,set
i;j
=
j;l
+
l;i
.Defin-
ing
i
astheRTTofsource
i
,clearly,
i
=
i;l
+
l;i
,for
all
l
suchthat
A
l;i
>
0
.InFig.1,asamplenetworkillus-
tratesdefineddelays.
Let
f
i;n
g
n
0
beanon-decreasing
[0
;
1
)
-valuedse-
quenceofrateupdatingtimesofsource
i
.Then,anum-
berofrateupdatesofsource
i
ontheinterval
[0
;t
)
is
N
i
(
t
)=
P
1
n
=1
1
f
i;n
<t
g
.
Let
f
x
i;n
g
n
0
bea
[0
;
1
)
-valuedstochasticprocess,
where
x
i;n
isarateofsource
i
atthe
n
-thupdate.Then,
defineacontinuoustimeinterpolationonrealtimeas
x
i
(
t
)=
x
i;n
,for
t
2
[
i;n
;
i;n
+1
)
.
For
b
a
0
definea
-algebraoftheform
F
i
;
l
[
a
;
b
)
=
(
x
j
;
k
:
A
l
;
j
;
A
l
;
i
>
0
;
N
j
(
a
,
i
;
j
)
k
<
N
j
(
b
,
i
;
j
))
.
Finally,let
F
i
[
a
;
b
)
=
[
l
:
A
l
;
i
>
0
F
i
;
l
[
a
;
b
)
.
Anadditive–increaseandmultiplicative–decreasealgo-
rithmhasthefollowingform
x
i;n
+1
=
x
i;n
+
r
i
(1
,
I
i;n
)
,
i
I
i;n
x
i;n
;
(1)
where
r
i
and
i
aretheadditive–increaseelementandthe
multiplicative–decreasefactor,respectively.Therandom
sequence
f
I
i;n
g
n
0
isanegativefeedbackindicationwith
valuesin
f
0
;
1
g
.Weassumethatthenegativefeedback
indication
I
i;n
isbasedonthefeedbackreceivedbetween
n
-thand
n
+1
-thrateupdating.Consequently,itturnsout
that
I
i;n
ismeasurableon
-algebra
F
i
[
i
;
n
;
i
;
n
+1
)
.
1
Wedefine
A
l;i
on
f
0
;
1
g
whichcanbeextendedto
[0
;
1]
toaccom-
modateforinstanceloadsharing,etc.[6]
(
n
+1 )
n
(
n
,
1)
s
t
n
s<
(
n
+1 )
x
i;n
+1
x
i;n
,
1
s
,
x
j
(
s
,
)
;j
:
A
l;i
;A
l;j
>
0
x
i;n
x
i
(
t
)
I
i;n
Fig.2.FeedbackmodelingfortheHOMRTTcase.
Nowletusbrieflycommentonthespecialcasead-
dressedin[6],whereitisassumedthatallround–triptimes
areequal,andtheratesareupdatedsynchronously.Fol-
lowingthedefinitionof
I
i;n
inthefullextent,itisrather
easytoseethat
I
i;n
isafunctionof
x
j;n
,
x
j;n
,
1
,and
x
j;n
,
2
,forall
j
2S
suchthat
A
l
;
i
;
A
l
;
j
>
0
,depending
onvaluesof
j;l
.Intherelatedwork[4]–[6],itiscom-
monlyassumedthat
I
i;n
iscomputedbasedon
x
j;n
,forall
j
2S
suchthat
A
l;i
;A
l;j
>
0
,whichisindeedanunreal-
isticassumption.
Letusassumethatallflowstraversinglink
l
haveequal
accessdelaytothatlink.Formally,
i;l
=
j;l
,forall
i;j
2S
,suchthat
A
l;i
;A
l;j
>
0
.Then,itfollowsthat
I
i;n
dependson
x
j;n
,
1
,forall
j
2S
suchthat
A
l;i
;A
l;j
>
0
.
WerefertothisassumptionasaHOMRTTassumption.In
addition,whenever
x
i;n
isusedwhereitshouldbe
x
i;n
,
1
werefertothisasastolenlag.Feedbackmodelingisil-
lustratedinFig.2.
B.FeedbackModeling
First,weintroduceanotionofthelinkcostfunction
g
l
(
): [0
;
1
)
!
[0
;
1]
,where
l
2L
.Atagiventime
t
,thelinkcostisafunctionofthelinkload
f
l
(
t
)=
X
i
2S
A
l
;
i
x
i
(
t
,
i
;
l
)
:
(2)
Onecaninterpret
g
l
(
f
l
(
t
))
asaprobabilityofmarking
asinglepacketattime
t
.Weareconcernedwithneg-
ativefeedbackindication,
I
i;n
,basedonfeedbackre-
ceivedwithin
[
i;n
;
i;n
+1
)
.Letuspartitiontheinterval
[
i;n
;
i;n
+1
)
intonon-overlappingintervals
[
a
k
;b
k
)
such
that
x
j
(
s
)
,isconstantfor
s
2
[
a
k
,
i;j
;b
k
,
i;j
)
,for
all
j
2S
suchthat
A
l;i
;A
l;j
>
0
.
Wedefine
M
i;l
a;b
asanamountofnegativefeedbackre-
ceivedbysource
i
fromlink
l
withininterval
[
a;b
)
,which
isequaltoanumberofmarkedpacketsoftheflow
i
bythe
link
l
withininterval
[
a
,
i
;b
,
i
)
.
Let
P
i;l
a;b
(
)
and
P
i
a;b
(
)
beconditionalprobabilitiesgiven
F
i
;
l
[
a
;
b
)
and
F
i
[
a
;
b
)
,respectively.Analogously,let
E
i;l
a;b
[
]
and
E
i
a;b
[
]
berespectiveconditionalexpectations.
3
Admittingtheinterpretationof
g
l
(
)
asaprobabilityof
markingasinglepacket,itiseasytoseethatwedohavea
binomialconditionalprobability
P
i;l
a
k
;b
k
(
M
i;l
a
k
;b
k
=
m
)=
d
x
i;N
i
(
a
k
,
i
)
(
b
k
,
a
k
)
e
m
!
g
l
(
f
l
(
a
i
k
))
m
[1
,
g
l
(
f
l
(
a
i
k
))]
d
x
i;N
i
(
a
k
,
i
)
(
b
k
,
a
k
)
e,
m
;
(3)
where
a
i
k
iswritteninlieuof
a
k
,
l;i
.Notethat
d
x
i;N
i
(
a
k
,
i
)
(
b
k
,
a
k
)
e
correspondstoanumberofpackets
offlow
i
thatarepresentonlink
l
within
[
a
k
,
i
;b
k
,
i
)
.
Clearly,theexpectedamountofnegativefeedbackis
E
i;l
a
k
;b
k
[
M
i;l
a
k
;b
k
]=
d
x
i;N
i
(
a
k
,
i
)
(
b
k
,
a
k
)
e
g
l
(
f
l
(
a
i
k
))
:
Similarly,let
M
i
a;b
beanamountofnegativefeedbackre-
ceivedwithin
[
a;b
)
bysource
i
fromalllinks
l
suchthat
A
l;i
>
0
.Itfollows
P
i
a
k
;b
k
(
M
i
a
k
;b
k
=0 )=
Y
l
:
A
l;i
>
0
P
i;l
a
k
;b
k
(
M
i;l
a
k
;b
k
=0 )
;
andfrom(3)follows
P
i
a
k
;b
k
(
M
i
a
k
;b
k
=0 )=
=
Q
l
:
A
l;i
>
0
[1
,
g
l
(
f
l
(
a
i
k
))]
d
x
i;N
i
(
a
k
,
i
)
(
b
k
,
a
k
)
e
:
(4)
Bydefinitionof
[
a
k
;b
k
)
wehave
P
i
i;n
;
i;n
+1
(
M
i
i;n
;
i;n
+1
=0 )=
Y
k
P
i
a
k
;b
k
(
M
i
a
k
;b
k
=0 )
:
(5)
Finally,
I
i;n
=1
,ifsource
i
hasreceivedanindica-
tion,within
[
i;n
;
i;n
+1
)
,thatatleastonepackethasbeen
marked,thus
P
i
i;n
;
i;n
+1
(
I
i;n
=1 )=
P
i
i;n
;
i;n
+1
(
M
i
i;n
;
i;n
+1
1)=
=1
,
P
i
i;n
;
i;n
+1
(
M
i
i;n
;
i;n
+1
=0 )
:
(6)
Letusexamine(6)fortheHOMRTTcase.Herewehave
asinglepartitionof
[
i;n
;
i;n
+1
)
,hence,from(4)–(6)fol-
lows
P
i
i;n
;
i;n
+1
(
I
i;n
=1 )=1
,
Y
l
:
A
l;i
>
0
[1
,
g
l
(
f
l
)]
d
x
i;n
,
1
e
;
(7)
where
i;n
+1
,
i;n
=
,forall
i
2S
,
n
0
,and
f
l
standsfor
f
l
(
i;n
,
l;i
)=
P
j
2S
A
l
;
j
x
j
;
n
,
1
.In
thelimitcase
g
l
(
)
!
0
,limiteddevelopmentyields
[1
,
g
l
(
f
l
)]
d
x
i;n
,
1
e
'
1
,d
x
i;n
,
1
e
g
l
(
f
l
)
,thenreplacing
thisin(7),andneglectingthehigherorderproducts,yield
P
i
i;n
;
i;n
+1
(
I
i;n
=1 )
'
X
l
2L
A
l
;
i
g
l
(
f
l
)
d
x
i
;
n
,
1
e
:
(8)
Therefore,itisshownthat,undertherarenegativefeed-
backassumption,(7)degeneratestotherateproportional
feedbackasisimplicitlyassumedin[6].However,note
that(8)dependson
x
i;n
,
1
andnoton
x
i;n
.
C.AsymptoticConvergence
Traditionaltheoryofthestochasticapproximational-
gorithms[13]–[14]isconcernedwithanalgorithmofthe
generalform
x
i;n
+1
=
x
i;n
+
n
H
i;n
(
x
i;n
;
i;n
)
;i
2S
;
where
x
i;n
isdefinedon
R
,
H
i;n
(
):
R
R
!
R
,
i;n
:
R
!
R
isarandomnoise,and
n
astepsize.It
isassumedthatcomponents
x
i;
,
i
2S
,areupdatedsyn-
chronously,facilitatingtheassociationofcontinuousinter-
polation
x
i
(
t
)
todiscreteprocess
f
x
i;n
g
n
0
onthe“natu-
ral”commoniteratetime
f
n
g
n
0
.However,itfollows
thatforanasynchronousupdatingonehastoworkinreal
time,oratleastanappropriatelyscaledrealtime[8].In
general,fordecreasing
n
,inrespectto
n
,convergence
withprobabilityonecanbeobtained,whileforconstant
small
n
,onlyconvergenceinprobabilitycanbe
proven(theweakconvergence).Intherestofthissec-
tionwebrieflysketchresultsof[8]thatareappliedinour
work.Foracompletetreatmentoftheunderlyingtheory
thereaderisreferredto[8].
Let
f
i;k
g
k
0
bearandomsequenceofupdatingin-
tervalsof
f
x
i;k
g
k
0
,
i
2S
.Thendenote(scaled)real
updatingtimeof
x
i;n
as
i;n
=
n
,
1
X
k
=0
i;k
;
(9)
andacontinuousinterpolationontheiteratetime
i
(
t
)=
i;n
,for
t
2
[
n;
(
n
+1 )
)
.Further,let
N
i
(
t
)
isanumber
ofupdatesof
f
x
i;k
g
k
0
before
t=
.Formally,
N
i
(
t
)=
1
X
n
=1
1
f
i;n
<t=
g
:
(10)
Fromthedefinitions,itturnsoutthat
N
i
(
i
(
t
))=
n
,
t
2
[
n;
(
n
+1 )
)
,i.e.
N
i
(
)
isinverseof
i
(
)
.
Then,let
^
x
i
(
t
)=
x
i;n
,
t
2
[
i;n
;
i;n
+1
)
,isacontinuous
interpolationonscaledrealtime,and
^
x
(
)=( ^
x
i
(
)
;i
=
1
;
2
;:::;S
)
,
S
=
jSj
.Fromdefinitionsof“time”pro-
cesses(9)and(10)itfollows
x
i
(
t
)=^
x
i
(
i
(
t
))
,and
^
x
i
=
x
i
(
N
i
(
t
))
.
Furthermore,let
i;j;n
beanon-negativerandomvari-
ablerepresentingascaled(multipliedby
)communica-
tiondelaybetweensource
i
andsource
j
atthe
n
-thrate
updatingofsource
i
.
4
Finally,thedecentralizedasynchronousalgorithmcan
bewrittenintheform
x
i;n
+1
=
[
a
i
;b
i
]
x
i;n
+
H
i;n
(^
x
j
(
i;n
+1
,
i;j;n
))
=
=
x
i;n
+
H
i;n
(^
x
j
(
i;n
+1
,
i;j;n
))+
Z
i;n
;i
2S
;
(11)
where
[
a
i
;b
i
]
(
)
denotesprojectionoftheargumenton
[
a
i
;b
i
]
,fortheconstrained
x
on
C
=[
a
1
;b
1
]
[
a
2
;b
2
]
:::
[
a
S
;b
S
]
,and
Z
i;n
isareflectionterm.
Letforall
i
,
F
i
;
n
and
F
;
+
i
;
n
benon-decreasing
-
algebrasmeasuringthepastdata(including
x
i;
0
,
H
j;k
,and
j;k
,
j
2S
)availableon
[0
;
i;n
+1
)
,and
[0
;
i;n
+1
]
,re-
spectively.Then,with
P
i;n
and
P
;
+
i;n
denoterespective
conditionalprobabilities,andanalogously
E
i;n
and
E
;
+
i;n
conditionalexpectations.
Subsequently,wehavethefollowingconditions.Itis
assumedthat
f
H
i;n
;
i;n
;
;i;n
g
;
(12)
and
i;j;n
,
;
+
i;j;n
areuniformlyintegrable.Weconsider
theMartingaledifferencenoise[8],forwhichwehave
E
i;n
H
i;n
=
h
i;n
(^
x
j
(
i;n
+1
,
i;j;n
)
;j
2S
)+
i
;
n
;
where
h
i;n
(
)
arereal-valuedfunctionscontinuousin
n
and
,
i;n
isasymptoticallynegligiblenoise,and
sup
n
T=
i;j;n
!
0
.Therearereal-valuedfunctions
u
i;n
(
)
thatarestrictlypositive(
inf
n;;x;
u
i;n
(
x;
)
>
0
)
andarecontinuousuniformlyin
n
and
,andnon-negative
randomvariables
;
+
i;j;n
suchthat
E
;
+
i;n
i;n
+1
=
u
i;n
+1
(^
x
j
(
i;n
+1
,
;
+
i;j;n
+1
)
;j
2S
)
;
(13)
where
sup
n
T=
;
+
i;j;n
!
0
;
inprobabilit ya s
!
0
.
Therearecontinuousreal-valuedfunctions
h
i
(
)
suchthat
foreach
x
2
C
,
2
lim
m;n;
1
m
n
+
m
,
1
X
k
=
n
[
h
i;k
(
x
)
,
h
i
(
x
)]=0
:
(14)
Therearecontinuousreal-valuedfunctions
u
i
(
)
suchthat
forall
x
2
C
,
lim
m;n;
1
m
n
+
m
,
1
X
k
=
n
E
;
+
i;n
[
u
i;k
(
x
)
,
u
i
(
x
)]=0
:
(15)
Suppose
lim
m;n;
n
+
m
,
1
X
k
=
n
E
i;n
i;k
=0
;
inmean
:
(16)
2
In(14),(15),and(16),
lim
m;n;
lim
m
!1
;n
!1
;
!
0
,simulta-
neouslyinanyway.
Finally,fromtheTheorem[8](AppendixA)particularly
followsthat,fortheunconstrainedalgorithm,theweak
convergencesubsequence
^
x
isthelimitsetofODE
_
^
x
i
=
h
i
(^
x
)
u
i
(^
x
)
;i
2S
:
(17)
Thus,thelimitmeanODEisthesameasinthesyn-
chronouscase,exceptforanadditionalweightfactorthat
takesintoaccountfrequencyoftheupdating.
D.
F
h
A
Fairness
Weidentify
H
i;n
of(11)inthealgorithm(1)as
H
i;n
=
r
i
,
(
r
i
+
i
x
i;n
)1
f
i;n
<P
i
i;n
+1
(
I
i;n
=1)
g
;
(18)
where
f
i;n
g
n
0
isasequenceofindependentrandom
variablesuniformlydistributedon
[0
;
1]
.
3
Then,
E
i;n
+1
H
i;n
=
r
i
,
(
r
i
+
i
x
i;n
)
P
i;n
+1
(
I
i;n
=1 )
;
where
P
i;n
+1
(
I
i;n
=1 )
isgivenby(6).
Inthelimitcase,as
!
0
,and
n
!1
,weneglect
scaleddelays
i;j
,thentheprobabilityofnegativefeed-
backis
P
i;n
+1
(
I
i;n
=1 )=^
x
i
(
;
,
i;n
+1
)
i;n
X
l
2L
A
l;i
g
l
(
^
f
l
(
;
,
i;n
+1
))
:
For
i;n
=
i
,forall
n>
0
,themeanvectorfieldis
h
i
(^
x
(
t
))=
r
i
,
^
x
i
(
t
)(
r
i
+
i
^
x
i
(
t
))
i
X
l
2L
A
l
;
i
g
l
(
^
f
l
(
t
))
(19)
where
^
f
l
(
)=
P
j
2S
A
l
;
j
^x
j
(
)
.
Seemingly,
u
i
(^
x
)=
i
,thenwith(19)thelimitmean
ODE(17)becomes
_
^
x
i
=
r
i
i
,
^
x
i
(
r
i
+
i
^
x
i
)
X
l
2L
A
l
;
i
g
l
(
^
f
l
)
:
(20)
Followingthesamestepsasin[6]weexpressODE(20)as
_
^
x
i
=^
x
i
(
r
i
+
i
^
x
i
)
@J
h
A
(^
x
)
@
^
x
i
;
(21)
where
J
h
A
(^
x
)=
X
i
2S
1
i
log
^x
i
r
i
+
i
^x
i
,
G
(
^x
)
;
(22)
andbydefinition
G
(^
x
)=
P
l
2L
G
l
(
^
f
l
)
,where
G
l
(
)
isa
primitiveof
g
l
(
)
.
3
Notethatinsteadof
r
i
and
i
itshouldbewritten
r
i
and
i
,where
r
i
=
r
i
and
i
=
i
,butweabusethisfornotationsimplicity.
Citations
More filters
Journal ArticleDOI
Fair end-to-end window-based congestion control
Jeonghoon Mo,Jean Walrand +1 more
TL;DR: The existence of fair end-to-end window-based congestion control protocols for packet-switched networks with first come-first served routers is demonstrated using a Lyapunov function.
Journal ArticleDOI
Fluid-based analysis of a network of AQM routers supporting TCP flows with an application to RED
TL;DR: This paper uses jump process driven Stochastic Differential Equations to model the interactions of a set of TCP flows and Active Queue Management routers in a network setting and presents a critical analysis of the RED algorithm.
Journal ArticleDOI
Bandwidth sharing: objectives and algorithms
Laurent Massoulié,James Roberts +1 more
TL;DR: It is proved that fixed-size window control can achieve fair bandwidth sharing according to any of these criteria, provided scheduling at each link is performed in an appropriate manner.
Proceedings ArticleDOI
Impact of fairness on Internet performance
Thomas Bonald,Laurent Massoulié +1 more
TL;DR: It is proved that for a broad class of fair bandwidth allocations, the total number of flows in progress remains finite if the load of every link is less than one.
Proceedings ArticleDOI
Statistical bandwidth sharing: a study of congestion at flow level
TL;DR: The statistics of the realized throughput of elastic document transfers are studied, accounting for the way network bandwidth is shared dynamically between the randomly varying number of concurrent flows.
References
More filters
Journal ArticleDOI
Random early detection gateways for congestion avoidance
Sally Floyd,Van Jacobson +1 more
TL;DR: Red gateways are designed to accompany a transport-layer congestion control protocol such as TCP and have no bias against bursty traffic and avoids the global synchronization of many connections decreasing their window at the same time.
Journal ArticleDOI
Congestion avoidance and control
TL;DR: The measurements and the reports of beta testers suggest that the final product is fairly good at dealing with congested conditions on the Internet, and an algorithm recently developed by Phil Karn of Bell Communications Research is described in a soon-to-be-published RFC.
Journal ArticleDOI
Rate control for communication networks: shadow prices, proportional fairness and stability
TL;DR: This paper analyses the stability and fairness of two classes of rate control algorithm for communication networks, which provide natural generalisations to large-scale networks of simple additive increase/multiplicative decrease schemes, and are shown to be stable about a system optimum characterised by a proportional fairness criterion.
BookDOI
Adaptive Algorithms and Stochastic Approximations
TL;DR: The juxtaposition of these two expressions in the title reflects the ambition of the authors to produce a reference work, both for engineers who use adaptive algorithms and for probabilists or statisticians who would like to study stochastic approximations in terms of problems arising from real applications.
Journal ArticleDOI
Analysis of the increase and decrease algorithms for congestion avoidance in computer networks
Dah Ming Chiu,Raj Jain +1 more
TL;DR: It is shown that a simple additive increase and multiplicative decrease algorithm satisfies the sufficient conditions for con- vergence to an efficient and fair state regardless of the starting state of the network.
Related Papers (5)
Frequently Asked Questions (4)
Q2. What have the authors stated for future works in "Global fairness of additive–increase and multiplicative–decrease with heterogeneous round–trip times" ?
Further research should concentrate on analyzing the impact of RTT variations due to queueing.
Q3. What is the nature of the bias against flows?
The authors see that the bias against flows with many hops is in the nature of any rate adaptation algorithm based on additive– increase and multiplicative–decrease.
Q4. What is the effect of the inverse on the network?
under the condition that there is no substantial queuing delay variation,and the network is operating in the regime of the rare negative feedback, the collection of rates x =( x1;:::;x i;::: ) is distributed such that x maximizes the objective functionF hA(x) = X i2S 1 i logxiri + ixi ;subject to the constraints Pj2S Al ;j xj cl , 8l 2 L.