Tree Graph Views for a Distributed Pervasive Environment

(1)

HAL Id: hal-00733447

https://hal.archives-ouvertes.fr/hal-00733447

Submitted on 18 Sep 2012

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Tree Graph Views for a Distributed Pervasive

Environment

Tuyet-Tram Dang-Ngoc, Nicolas Travers

To cite this version:

Tuyet-Tram Dang-Ngoc, Nicolas Travers. Tree Graph Views for a Distributed Pervasive Environment.

1st International Conference on Network-Based Information Systems (NBIS), 2007, Regensburg,

Ger-many. �hal-00733447�

(2)

Environment

TuyêtTrâmDangNgo

1

andNi olasTravers

2

1

ETISLaboratory-UniversityofCergy-Pontoise,Fran e.

Tuyet-Tram.Dang-Ngo u- ergy.f r

2

PRiSMLaboratory-UniversityofVersailles, Fran e.

Ni olas.Traversprism.uvsq.fr

Abstra t. ThepervasiveInternetandthemassivedeploymentofsensor

devi eshaveleadtoahugeheterogeneousdistributedsystem onne ting

millionsofdatasour esand ustomerstogether[Fra01 ℄.Ontheonehand,

mediationsystems[BGL

+

99 ,DNJT05 ℄ usingXMLas anex hange

lan-guagehavebeenproposedtofederatedataa rossdistributed

heteroge-neousdatasour es.Ontheotherhand,work[MSFC02,AML05 ,BGS01 ,NDK

+

03 ℄

havebeendonetointegratedatafromsensors.The hallengeisnowto

in-tegratedata omingfromboth" lassi al"data(DBMS,Websites,XML

les)and"dynami "data(sensors)inthe ontextofanad-ho network,

andnally,toadaptqueriesandresulttomat hthe lientprole.

We propose to use the TGV model [TDNL06 ,TDNL07a ℄ as a mobile

agenttoquerysour esa rossdevi es(sour esandterminal)inthe

on-text of a res ue oordination system. This work is integrated in the

PADAWANproje t.

Keywords: XQueryevaluation,Tree GraphView(TGV), Pervasive

envi-ronment,Res ueCoordination

1 Introdu tion

ThepervasiveInternet andthemassivedeploymentofsensordevi es havelead

to ahugeheterogeneousdistributed system onne tingmillionsofdatasour es

and ustomerstogether.

Ononeside,datasour esareheterogeneousasthey anbeofdierenttypes

(relational,text,XML,streamingvalue,et .)and anhavedierentupdate

fre-quen ies(from "never" for sometext do umentto "always" for sensorsvalue)

andtheirautonomy(from non-manageableobfus ated bla kboxthat just

pro-videvaluesto afulla ess managementonaDBMS).On theotherside, there

anbedierentprolesof lients:a esspermission,terminal apabilities,user

preferen es,et .

Todealwithdistributedheterogeneousandautonomousdatasour es,

media-tionsystemshavebeenwidelystudied[Wie92,MFK01,NGT98,BGL

+

(3)

heterogeneity.

XML[BPSM98℄,hasbe omethepreferredformattorepresentsemi-stru tured

data[Abi97℄and anee tivewaytodene anytypeofdatathat anbe

repre-sentedas atree.

Moreover, XQuery [W3C05℄ has proved to be an expressive and powerful

querylanguagetoqueryXMLdatabothonstru tureand ontent,andtomake

transformation on the data. In addition, its query fun tionalities ome from

both the database ommunity (ltering, join, sele tion, aggregation), and the

text ommunity(supportinganddeningfun tion astextsear h).

TGV(TreeGraphView)[Tra06,TDNL06,TDNL07a℄isaTreePattern-based

model(su hasTPQ[CJLP03℄andGTP[AYCLS01℄)tomodelXQueryqueries.

Thismodelissuitableto ourneedssin e:

it supports the omplexity of the full untyped-XQuery spe i ation:

rela-tionalandsetoperator,aggregation,ordering,nestedre onstru tion,

ondi-tionalpredi ate,et .

itis designedfor amediation ontext a essing to distributed autonomous

andheterogeneousdatasour es:itsstru tureidentiesdata olle tionsand

dependen ies between them. An annotation model in layers allows to

an-notateanypie e of information(lo ation ofthe sour e(s), ost model, et )

that anbeusefulaevaluationtime.Finaly,transformationruleshavebeen

denedtooptimizeandevaluatetheTGVtoprodu etheresult.

Therestofthispaperisorganizedasfollows.Tostartowith,wemotivate

theneedforamobilesemi-stru turedmodelinapervasiveenvironmentin

Se -tion2and we expressissuesand related works inSe tion 3.Further,were all

theTGVmodeland fun tionalitiesin Se tion4and showhowitissuitable to

our ontext. We then present some extensions to the TGV model that would

makeit more suitable to a pervasiveenvironment (Se tion 5). Inthe end, we

on ludeinSe tion6andpresentfuturedire tionsofourwork.

2 Context

Inthe global ontext of ourwork,dierent typesof datasour esand terminal

lientaredissiminatedalloveranetwork onsistedoftraditionalIProutingand

addressing(e.g. theInternet),andad-ho routings heme.

2.1 Motivational S enario

Theappli ations enarioisthedeploymentofaRes ueCoordinationCenterafter

orduring adisaster(re,earthquake,ood,et .)Atru k(Figure1) arriesthe

PADAWANproxy,somea esspoints(wire,wirelessWI-FI,SINK,et .)andan

InternetA ess.

(4)

Primergy

Sink

Doctor

Officer

Expert

802.11 AP

IP

Router

Ad−Hoc Network

Sensor Network

Internet

Fireman

00

11

11 Proxy

Proxy

Emergency Unit

PADAWAN

Emergency Unit

PADAWAN

Directories

DBMS

Web Server

Broadcast News

Fig.1.DeploymentofaRes ueCoordinationCenter

Wireless ad-ho Sensor Network [Toh01℄ (ad-ho WSN): Sensors are

de-ployed(eg. from heli opterorembedded in theres ue team equipment)in

the monitoring area to form a WSN. The entry point is a sink (a node

with largeresour esthat olle tsstatisti s from nodes in its overage and

generallyremainsinastati lo ation).

privatenetwork and/or ad-ho network: used bythe res ue team (remen,

o ers,do tors,emergen yunit)

theInternet:toa essdatabases,website,dire tories,et .

Using su h asystem,dependingof theirprole anda ess rights, notonlythe

members of the res ue team, but the experts and the press an a ess to an

integratedviewonallsour esrelatedtothedisasterandqueryit.

2.2 PADAWAN

Theres ues enarioreliesonthePADAWAN(aProxyforAllDevi esA essing

theWorldAndNeighborhood)infrastru ture.ThePADAWANinfrastru tureis

seen asa graphwhere links are networks links (radio orwire), and nodes are

eitherdata sour es, lientterminalsorthePADAWANproxyitself.

DataSour es Wedenebysour e,anydevi ethatprovidesdata:DBMS,

sen-sors,websites,RSSfeeds,lega yappli ations,dire tories,les,et .We onsider

dierentkindsof(non-ex lusive)heterogeneity:

datatype:sour es anhavearelational,text-based,semi-stru tured,

unstru -turedmodel.Thequerylanguage andier: SQL,XQuery,OQL, ontains

fun tion,httprequest,et . andso theresultformat :tuple, XML, obje ts,

textlines,textdo uments,et .

autonomy:ex ept for somedatabasesmanaged by the PADAWAN

admin-istratorsthemselves,themajornumberofsour esareautonomousandjust

ommuni ate what their owner want them to ommuni ate. They mainly

(5)

stati sour es where data donot hange frequently, eg. Webpages, LDAP

dire tories,somedatabases.Otherdatasour esare onsideredasdynami ,

astheir information update very frequentlyor on ea h request (RSS feed,

sensorsmeasure).

Clients All kindof lients ana ess toPADAWAN toquerythesour es:from

asimplepager to a omplexappli ationmanipulating hugevolumeof sour es.

Available data and results are pro essed using spe i views based on lient

prole.Theseviews are omputeddependingon:

user preferen e: eg. the remen o er want to know about the

tempera-turemeasured by ea h sensors, the pressjust needs to a ess the average

temperatureofthesite.

user a ess rights: eg. the res ue team has a ess to the personal medi al

informationofinjuriedpeople, othersdon't.

lient terminal apabilities: eg. the ell-phone used by the reghter does

nothavethesamedisplay apabilitiesasthelaptop omputeroftheres ue

ooordination o er, so large volume of information, images and video is

notapplyablein every ase.

The PADAWAN Proxy The ore ofthe systemis the PADAWAN proxy (that

is lo atedin the res uetru k). This proxyis amediator system olle tingand

requestingallavailabledata omingfrom deployedsensors,and embedded

sys-temsusedbytheres uersre eivedviathea esspoints,andalsofromdatabases

andRSSfeedsfromtheInternet.

TheFigure 2showsa queryQand itsXQuery representation.We suppose

that, forexample, information onbuilding o upation is stored in arelational

DBMSlo atedonanInternetsite,andthatsensorsaredeployedoveranad-ho

sensornetworkrea hablebyasink a esspoint.

LetQbethequerythat"listeverybuildingso upiedbymorethan100inhabitants,andforea h,get

thedistri tandthelistofmaximumtemperaturemeasuredbythesensorslo atedinthesamedistri t."

XQueryRequest XMLResult

for$ain/buildings/building <distri tMonitoring>

where$a/des ription/inhabitant>100 <lo ation>YellowLake</lo ation>

return <temperatures>

<distri tMonitoring> <temperature>14</temperature> <lo ation>{$a/distri t}</lo ation> </temperatures>

{for$bin//sensor <distri tMonitoring>

where <lo ation>GreenValley</lo ation>

$b/deploymentArea/distri t=$a/distri t <temperatures>

return <temperature>163 </temperature>

</temperatures> <temperature>43</temperature>

</distri tMonitoring> </temperatures>

</distri tMonitoring>

(6)

Mediationsystems[BGL

+

99,DNJT05℄basedonmediator/wrappersar hite ture

[Wie92℄ using XML as an ex hange language have been proposed to federate

data a ross distributed heterogeneous data sour es. The heterogeneity of the

datatypeishandledbywrappersthata tas"translators"fromthesour enative

querylanguageandresulttothe ommonquerylanguageandmodelusedbythe

mediator. Themediatorde omposestheuserinto subqueriessenttowrappers,

andre omposesthenalresult,andthus,managesthedistributedaspe tofthe

system.

Manywork[MSFC02,AML05,BGS01,NDK

+

03℄ havebeendonetointegrate

datafromsensors. Thersttypeofapproa h onsidersthesensornetworkasa

virtual[MSFC02℄ ormaterialized[BGS01℄ relationaltable.These ond typeof

approa hasintheIrisNet[NDK

+

03℄, onsidersthewebasa"huge"XML

do -ument, using DNS extension to lo atenodes of the XML do ument.However,

these worksare notdesignedtobeintegratedwith otherDBMS ina

heteroge-neousenvironment,usingquerieswith omplexfun tionalities.

The hallengeishowtoevaluateXQuerya rossthegraph omposedof

het-erogeneoussour es,heterogeneous lientsandheterogeneousinfrastru ture.

WeproposetousetheTGVmodeltodealwiththisproblem.Were allthe

TGVbasisinthenextse tion,andshowhowitissuitabletoourneeds.

4 Tree Graph View (TGV)

TGV (Tree Graph View) [Tra06,TDNL06,TDNL07a℄ is a Tree Pattern-based

model (su h as TPQ [CJLP03℄ and GTP [AYCLS01℄) designed to represent

XQueryrequest and itsevaluation. TheTGV supports allthe fun tionnalities

of untyped-XQuery, uses anintuitive representation ompliantwith mediation

issues,andprovidesasupportforoptimizationandinformation.

4.1 TGV example

ThisTGVrepresentationoftheXQueryQ(Figure2)isshownonFigure3(a).

Thetree patternsofthe twodata olle tionsof thequeryareshownin ir les.

The intermediate and nal result onstru tion are represented within boxes.

Dependen iesareshownby hyperlinklines bindingthepatternsorthepattern

nodes:joinonthetwosour etreepatternsandproje tionfromonenodeortree

patterntoanother.Thenestedtemperatureintemperaturesisalsosupported

(the $tbox).

Thistype ofrepresentation issuitable to amediationsystemassubpart of

(7)

$a

$b

> 100

temperatures

location

districtMonitoring

description

building

buildings

inhabitant

sensor

district

deploymentArea

district

temperature

$t

max_temp

=

4) Logical TGV

description

building

inhabitant

temperatures

location

deploymentArea

deploymentArea max_temp

location

$a

$b

sensor

$t

> 100

buildings

=

district

annotation

3) Physical TGV

annotation

with cost

2) Physical TGV

annotation

1) Physical TGV

with evaluation

with location

description

building

district

deploymentArea

$a

$b

sensor

buildings

=

deploymentArea

location

temperatures

> 100

inhabitant

sensor

$t

max_temp

location

description

building

inhabitant

district

temperatures

location

deploymentArea max_temp

location

$a

$b

sensor

$t

> 100

buildings

=

deploymentArea

temperatures

location

$t

buildings

=

deploymentArea

sensor

max_temp

deploymentArea

$b

sensor

location

> 100

inhabitant

$a

building

district

description

Yellow Lake | 1678

Rhode Forest | 1986

Green Valley | 82761

Yellow Lake | 14

Green Valley| 163

Green Valley| 25

Green Valley| 43

**C2=cost_card*0.02**

cp=max(C1, C2)

cost=cp+Op

CS1=Cost(S1)

**C1=CS1selCS2*IO**

Cost(S2)

Source1

Mediator

Source2, Source 3

Mediator

Fig.3.(a)TGVoftheQueryQ-(b)TGVAnnotationLayersView

4.2 Annotation

Set of elements ofthe TGV an be annotatedfor (a) any granularity of

infor-mation and (b) any type of information ( ost models and statisti s, lo ation,

onstraint,a ura y,se urity,ruletra ability).

Using annotation, a TGV an be viewed on any type of annotation that

hasbeendened onit. OnFigure 3(b),theoriginal logi alTGV 4

hasthree annotatedviews:

1

lo ation annotation: ea hlo ationof theexe utionofsubparts oftheTGV isreportedontheasso iatesetofTGVelementsontheTGV.Inour

exam-ple,the informationon buildings anbe retrieved froma sour e Sour e1

that isa DBMSa essible from the Internet. The information on sensors

areretrievedfromSour e2 andSour e3thatarelo atedon twoSINKS

re-spe tivelylo atedon theYellow Lakeandon theGreen Valley.Theother

parts of the TGV areevaluatedby the mediator,lo atedon the PADAWAN

proxy.

2

time ost annotation [LDNL07℄: the time ost exe utionare annotatedon subpartsoftheTGV.Thetime ostsareevaluatedusing ostmodels.

3

evaluation annotation: This annotationlayeris used toevaluate theTGV. Theevaluationannotationsareintermediateornalresultsthathavebeen

evaluated on subparts of the TGV. Theevaluation pro ess is des ribedin

thesubse tion4.4("evaluation").

Theannotationspe i ation[TDNL06,LDNL07℄aregeneri enoughtoannotate

anysubpartoftheTGV,in ludingmoney ost,energy ost(batteryforsensors),

a ura y,et .

4.3 Transformations

In[TDN07℄, apattern-basedlanguagefor extensible ruleshasbeendened for

(8)

trans-phase, ostannotation anbe usedto generate betterplan: the rule ondition

anexpressthatiftheestimated ardinalityof leftsideofanbind-join ismu h

lesserthanthe ardinalityoftherightside,thantherulewillinvertea hsideof

thebind-joinforbetterperforman eduring theevaluation.

4.4 Evaluation

A parti ular ategoryof transformation rules is the evaluation rules ategory,

whi h evaluate subparts of TGV using evaluation annotation. A TGV with

emptyevaluationannotationis onsideredasanexe utionplan.Mat hingsour es

lltheevalutionannotationwithdatamat hingthere ognizedpatterns.Then,

using iterativelyevaluation rules,the annotationsarepropagated in theTGV.

At the end of the evaluation, pro ess, the whole TGV is annotated with the

result of the query. The Figure 4 shows the evaluation pro ess steps.

Start-ing from the TGVQuery onFigure 3 withempty evaluation annotations,the

Sour e Tree Patterns are annotated (7a)with mat hing information retrieved

fromtheappropriatesour es(usinglo ationannotation).Thenthe

transforma-tionrulemat hingthejoinhyperlinkapply(7b), thentheaggregationrule(7 )

and nally, the proje tion rule annotates thewhole TGV with the evaluation

annotation ontaining the nal result (7d) that an be returned as an XML

do ument(Figure2(left olumn)).

A similar evaluation approa hhas beendevelopped in Mutant QueryPlan

[PMT03℄.

5 TGV in a Pervasive Environment

5.1 TGV Mobile Agent

TheTGVevaluationbytransformationrulesonannotationsisverywelladapted

toamobileagentplatform.TheTGVMobileAgentmovebetweennodesa ross

thegraph.Ea h nodeofthegraph:

appliesevaluationrulesonthemat hedTGV,tollevaluationannotations

readslo ationannotationstoroutetheTGVtothenext on ernednode

Querypro essing using Mobile Agent -but in relational ontext- has been

developped in the work of [MHMM05℄. Using the TGV model, mobile agent

wouldbeabletoqueryonsemi-stru tureddistributeddata,andthusbeusedin

aheterogeneousenvironment.

5.2 Views on TGV

A lient prole is represented by a view. A view is a request, and so an be

representedbyaTGV.Ea h lienthasaview onstru tedfromtheuser

prefer-en e,theusera esspermissionandthe lientterminal apabilities.Theprole

ismodeledasaview(a TGVrequest)thatisapplied tothe lientrequest

(9)

0000000000000000000000

1111111111111111111111

00000000000000000

11111111111111111

00000000000000000

11111111111111111

0000000000000000000000

1111111111111111111111

$a

$b

> 100

temperatures

location

districtMonitoring

description

building

buildings

inhabitant

sensor

district

deploymentArea

district

temperature

$t

max_temp

=

$a

$b

> 100

temperatures

location

districtMonitoring

description

building

buildings

inhabitant

sensor

district

deploymentArea

district

temperature

$t

max_temp

=

$a

$b

> 100

temperatures

location

districtMonitoring

description

building

buildings

inhabitant

sensor

district

deploymentArea

district

temperature

$t

max_temp

=

$a

$b

> 100

temperatures

location

districtMonitoring

description

building

buildings

inhabitant

sensor

district

deploymentArea

district

temperature

$t

max_temp

=

Green Valley | 82761 | 163

| | 25

| | 43

Yellow Lake | 678 | 14

Green Valley | 82761 | 163

Green Valley | 82761 | 25

Green Valley | 82761 | 43

Yellow Lake | 678 | 14

14

163

25

43 Green Valley | 82761 | 163

Green Valley | 82761 | 25

Green Valley | 82761 | 43

Yellow Lake | 678 | 14

Yellow Lake | 14

Green Valley | 163

Green Valley | 25

Green Valley | 43

Yellow Lake | 678

Rhode Forrest| 1986

Green Valley | 82761

(c)

(d)

(a)

(b)

Fig.4.EvaluationofaTGV 5.3 TGV* (TGV Star)

Whenasour e onne tstothePADAWAN proxy,theasso iatedwrappersend

thesour edes riptiontotheproxy.Thesour edes riptionisalsoaTGVwithas

manyannotationviewsasdierenttypesofinformationprovided.AsthisTGV

is aTGVwhere allpathsof supportedtree patterns(as indataguide [GW97℄)

are annotated, this TGV is alled TGV* (TGV star). All TGV* are sent by

wrappers to the PADAWAN proxy and merged to make a big TGV* on the

proxy.ThisTGV*isthenusedbytheproxytoannotateTGVuserrequestwith

informationaboutlo ation, ost,et .

6 Con lusion

Inthis arti le, wehavepresentedtheTGVas amodel suitablefordistributed

evaluation of anXQuery requestovera pervasive environment,using ares ue

oordinations enario.

asarepresentation[TDNL07b℄ofthefull-untypedXQueryspe i ation,the

TGVinherits of allthe power of the XQuery language(relationaland set

operator,aggregation,ordering,nestedre onstru tion, onditionalpredi ate,

et .)

(10)

mergingrulesmakeviewsonrequesteasytoevaluate,andthus, lientprole

easyto onsider

itsevaluation rules by using annotation and transformation make the

dis-tribuatedTGVevaluationfeasableonautonomousnodes.Thus,withrouting

onsideration,movingTGVtomobile agent anbedone.

itsextensible optimisationrules,allowsustopro essTGVe iently.

Future Works In this paper, we onsider that sour es des ription (metadata)

are entralizedonthePADAWANproxyasakindofaYellowPageservi e.The

s enario ontextofthe oordinationofres ueteamslegitimatesthisapproa h.

Inamoregeneral ontext,itwouldbeinterestingtodistributethemetadata

all overthe network.This raisethe problem ofmaintaining su h adistributed

indexofmetadata.Weare urrentlystudyingwhetheraP2PDHT approa his

suitabletoourneeds,andhowtodistributetheTGV*in thisway.

If we hoose to distribute the data sour es des ription, there won't be a

entralPADAWANproxyanymore,andanynode ouldthenbe onsideredasa

PADAWANproxyforothernodes,withmoreorless apabilities.Thisapproa h

will transformthewhole PADAWANar hite tureinto aP2Pnetworkthat will

beapplied inthe ontextwhereno riti al oordinationisneeded.

ReexionaboutroutingTGVasamobileagentwouldthenhavetobedone.

Tooptimizequeriesexe ution,nodes analso a hedataandsour edes ription

fromTGVtheyforwardtoothernodes.

A knowledgement

ThisworkisdoneaspartaofthePADAWAN proje tsupportedbytheANR.

Referen es

[Abi97℄ S.Abiteboul. Querying Semistru turedData. In Pro eeding of the 6th

InternationalConferen e onDatabaseTheory,Delphi,Gree e,1997.

[AML05℄ D.J.Abadi,S.Madden,andW.Lindner. Reed:Robust,e ientltering

andeventdete tioninsensornetworks. InVLDB,pages769780,2005.

[AYCLS01℄ S.Amer-Yahia,S.Cho,L.V.S.Lakshmanan,andD.Srivastava.

Minimiza-tionofTreePatternQueries. InSIGMODConferen e,2001.

[BGL

+

99℄ C.Baru,A.Gupta,B.Ludas her,R.Mar iano,Y.Papakonstantinou,and

P. Velikhov. XML-Based Information Mediation with MIX. In

ACM-SIGMOD,Philadelphia,USA,1999.

[BGS01℄ P.Bonnet,J.Gehrke,and P.Seshardi. Towardsensordatabasesystems.

InConferen eonMobileDataManagement,2001.

[BPSM98℄ T.Bray,J. Paoli, and C.Sperberg-Ma Queen. Extensible Markup

Lan-guage(XML)1.0(W3CRe ommendation),1998.

[CJLP03℄ Z.Chen,H.V.Jagadish,L.V.S.Laksmanan,andS.Paparizos. FromTree

(11)

e ientevaluationofxqueryinanxmlmediator. InPro .of BDA,2004.

[DNJT05℄ T.-T. Dang-Ngo , C. Jamard, and N. Travers. XLive: An XML Light

IntegrationVirtualEngine. InPro .of BDA,2005.

[Fra01℄ M.J. Franklin. Challenges inubiquitousdatamanagement. Informati s,

pages2433,2001.

[GW97℄ R.GoldmanandJ.Widom. Dataguides:Enablingqueryformulationand

optimizationinsemistru tureddatabases. InVLDB,1997.

[LDNL07℄ T.Liu,T.-T.Dang-Ngo ,andD.Laurent.Costframeworkforadistributed

semi-stru turedenvironment. InDatabase Management and Appli ation

overNetworks(DBMAN),2007. (toappear).

[MFK01℄ I.Manoles u, D. Flores u, and D. Kossmann. Answering XML Queries

overHeterogeneous Data Sour es. In27th Intl Conf VLDB,pages 241

250,Roma,Italy,2001.

[MHMM05℄ N.Marsit,AbdelkaderHameurlain,ZoubirMammeri,andFran kMorvan.

Query pro essing in mobileenvironments:A surveyand openproblems.

InDistributedFrameworks forMultimediaAppli ations(DFMA),2005.

[MSFC02℄ S.Madden,R.Szew zyk,M.J.Franklin,andDavidE.Culler. Supporting

aggregatequeriesoverad-ho wirelesssensornetworks.InWMCSA,2002.

[NDK

+

03℄ S.Nath, A. Deshpande, Y. Ke, P.B. Gibbons, B. Karp,and S. Seshan.

Irisnet:Anar hite tureforinternet-s alesensingservi es.InVLDB,2003.

[NGT98℄ H.Naa ke,G.Gardarin,andA.Tomasi . LeveragingMediatorCost

Mod-elswithHeterogeneousDataSour es. InICDE,pages351360,1998.

[PMT03℄ V.Papadimos,D.Maier,andK.Tufte. Distributedquerypro essing and

atalogsforpeer-to-peersystems. InCIDR,2003.

[TDN07℄ N.TraversandT.-T.Dang-Ngo .Anextensibleruletransformationmodel

forxqueryoptimization.InInternationalConferen eonEnterprise

Infor-mationSystems(ICEIS),2007. (toappear).

[TDNL06℄ N. Travers, T.-T. Dang-Ngo , and T. Liu. Tgv: an e ient model for

xqueryevaluationwithin aninteroperablesystem. InternationalJournal

ofInteroperability inBusiness InformationSystems(IBIS),2,2006.

[TDNL07a℄ N. Travers, T.-T. Dang-Ngo , and T. Liu. Tgv: a tree graph view for

modellinguntypedxquery.InDatabaseSystemsforAdvan edAppli ations

(DASFAA),April2007.

[TDNL07b℄ N.Travers,T.-T.Dang-Ngo ,andT.Liu. Untypedxquery anonization.

InWorkshop on EmergingTrends of Web Te hnologies and Appli ations

(WebETrends),2007. (toappear).

[Toh01℄ C.K.Toh. Ad Ho Wireless Networks: Proto ols and Systems. Prenti e

Hall,NJ,USA,2001.

[Tra06℄ Ni olasTravers. Optimization Extensible dans un Médiateur de Données

XML. PhDthesis,UniversityofVersailles, 2006.

[W3C05℄ W3C. AnXMLQueryLanguage(XQuery1.0),2005.

[Wie92℄ G.Wiederhold. MediatorsintheAr hite tureofFutureInformation