etri Netbased Mo del for eb Service Comp osition Rac hid Hamadi Boualem Benatallah Sc ho ol of Computer Science and Engineering The Univ ersit of New South ales Sydney NSW  Australia rhamadibouale cs

etri Netbased Mo del for eb Service Comp osition Rac hid Hamadi Boualem Benatallah Sc ho ol of Computer Science and Engineering The Univ ersit of New South ales Sydney NSW Australia rhamadibouale cs - Description

uns w edu a Abstract The In ternet is going through sev eral ma jor hanges It has ecome ehicle of eb services rather than just rep osi tory of information Man organizations are putting their core business comp etencies on the In ternet as collection ID: 29121 Download Pdf

239K - views

etri Netbased Mo del for eb Service Comp osition Rac hid Hamadi Boualem Benatallah Sc ho ol of Computer Science and Engineering The Univ ersit of New South ales Sydney NSW Australia rhamadibouale cs

uns w edu a Abstract The In ternet is going through sev eral ma jor hanges It has ecome ehicle of eb services rather than just rep osi tory of information Man organizations are putting their core business comp etencies on the In ternet as collection

Similar presentations


Tags : uns edu
Download Pdf

etri Netbased Mo del for eb Service Comp osition Rac hid Hamadi Boualem Benatallah Sc ho ol of Computer Science and Engineering The Univ ersit of New South ales Sydney NSW Australia rhamadibouale cs




Download Pdf - The PPT/PDF document "etri Netbased Mo del for eb Service Comp..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.



Presentation on theme: "etri Netbased Mo del for eb Service Comp osition Rac hid Hamadi Boualem Benatallah Sc ho ol of Computer Science and Engineering The Univ ersit of New South ales Sydney NSW Australia rhamadibouale cs"— Presentation transcript:


Page 1
etri Net-based Mo del for eb Service Comp osition Rac hid Hamadi Boualem Benatallah Sc ho ol of Computer Science and Engineering The Univ ersit of New South ales Sydney NSW 2052, Australia rhamadi,bouale cs e. uns w. edu .a Abstract The In ternet is going through sev eral ma jor hanges. It has ecome ehicle of eb services rather than just rep osi- tory of information. Man organizations are putting their core business comp etencies on the In ternet as collection of eb services. An imp ortan hallenge is to in tegrate them to cre- ate new alue-added eb services in ys that could

nev er foreseen forming what is kno wn as Business-to-Business (B2B) services. Therefore, there is need for mo deling tec hniques and to ols for reliable eb service comp osition. In this pap er, prop ose etri net-based algebra, used to mo del con trol o ws, as necessary constituen of reliable eb service com- osition pro cess. This algebra is expressiv enough to capture the seman tics of complex eb service com binations. Keywor ds: eb services, etri net, eb service com- osition. In tro duction In order to surviv the massiv comp etition created the new online econom man organizations are

rushing to put their core business comp etencies on the In ternet as collection of eb services for more automation and global visibilit The concept of Web servic has ecome recen tly ery opular, ho ev er, there is no clear agreed up on denition et. ypical examples of eb services include on-line tra el reser- ations, pro curemen t, customer relationship manage- men (CRM), billing, accoun ting, and supply hain. In this pap er, eb service (or simply service) mean an autonomous soft are application or comp o- nen t, i.e., seman tically ell dened functionalit uniquely iden

tied Uniform Resource Lo cator (URL). The abilit to ecien tly and eectiv ely share ser- vices on the eb is critical step to ards the de- elopmen of the new online econom driv en the Business-to-Business (B2B) e-commerce. Existing en terprises ould form alliances and in tegrate their services to share costs, skills, and resources in oer- ing alue-added service to form what is kno wn as B2B servic es Briey stated, B2B service is con- glomeration of mostly outsourced services orking in tandem to ac hiev the business goals of the desired en- terprise. An

example of an in tegrated B2B service is nancial managemen system that uses pa yroll, tax preparation, and cash managemen as comp onen ts. The comp onen services migh all outsourced to business partners. Cop yrigh 2003, Australian Computer So ciet Inc. This pa- er app eared at ourteen th Australasian Database Conference (ADC2003), Adelaide, Australia. Conferences in Researc and Practice in Information ec hnology ol. 17. Xiaofang Zhou and Klaus-Dieter Sc hew e, Ed. Repro duction for academic, not- for prot purp oses ermitted pro vided this text is included. date, the dev elopmen

of B2B services has een largely ad-ho c, time-consuming, and requiring enor- mous eort of lo w-lev el programming (Benatallah, Medjahed, Bouguetta a, Elmagarmid Beard 2000, Benatallah, Dumas, Sheng Ngu 2002). This task ould ob viously tedious and hardly scalable e- cause of the olatilit and size of the eb. As ser- vices are most lik ely autonomous and heterogeneous, building B2B service with appropriate in ter-service co ordination ould dicult. More imp ortan tly the fast and dynamic comp osition of services is an essen tial requiremen for organisations to adapt their business

practices to the dynamic nature of the eb. As oin ted out efore, In ternet and eb tec hnolo- gies ha op ened new ys of doing business more heaply and ecien tly Ho ev er, for B2B e-commerce to really tak o, there is need for eectiv and e- cien means to abstract, comp ose, analyze, and ev olv eb services in an appropriate time-frame. Ad-ho and proprietary solutions on the one hand, and lac of canonical mo del for mo deling and managing eb services on the other hand, ha largely hamp ered faster pace in deplo ying B2B services. Curren tec h- nologies based on Univ

ersal Description, Disco ery and In tegration (UDDI) eb Service Description Language (WSDL) and Simple Ob ject Access Pro- to col (SO AP) do not realize complex eb service com binations, hence pro viding limited supp ort in ser- vice comp osition. SO AP is standard for exc hanging XML-formatted messages er HTTP et een appli- cations. WSDL is general purp ose XML language for describing what eb service do es, where it re- sides, and ho to in ok it. UDDI is standard for publishing information ab out eb services in global registry as ell as for eb service disco ery In this pap er, prop ose etri

net-based al- gebra for mo deling eb services con trol o ws. The mo del is expressiv enough to capture the seman tics of complex service com binations and their resp ectiv sp ecicities. The obtained framew ork enables declar- ativ comp osition of eb services. sho that the dened algebra caters for the creation of dynamic and transien relationships among services. The remainder of this pap er is organized as fol- lo ws. eb service mo deling and sp ecication using etri nets are presen ted in Section 2. Section is de- oted to the algebra for comp osing eb services

and its etri net-based formal seman tics. Section dis- cusses the analysis and erication of eb services. Section giv es brief erview of related ork. Fi- nally Section pro vides some concluding remarks. ttp://www.uddi.org/. ttp://www.w3.org/TR/wsdl/. ttp://www.w3.org/TR/soap12-part1/.
Page 2
eb Services as etri Nets etri nets (P etri 1962, eterson 1981) are ell- founded pro cess mo deling tec hnique that ha for- mal seman tics. They ha een used to mo del and analyze sev eral yp es of pro cesses including proto- cols, man ufacturing systems, and business pro cesses (Aalst

1999). etri net is directed, connected, and bipartite graph in whic eac no de is either plac or tr ansition ok ens ccup places. When there is at least one tok en in ev ery place connected to transition, sa that the transition is enable An enabled transition ma r remo ving one to- en from ev ery input place, and dep ositing one to- en in eac output place. or more elab orate in- tro duction to etri nets, the reader is referred to (Murata 1989, Reisig 1985, eterson 1981). The use of visual mo deling tec hniques suc as etri nets in the design of complex eb services is justied man

reasons. or example, visual represen tations pro vide high-lev el et precise language whic allo ws to express and reason ab out concepts at their natural lev el of abstraction. eb service eha vior is basically partially ordered set of op erations. Therefore, it is straigh t- forw ard to map it in to etri net. Op erations are mo deled transitions and the state of the service is mo deled places. The arro ws et een places and transitions are used to sp ecify causal relations. can categorise eb services in to material ser- vices (e.g., deliv ery of ph ysical pro ducts), informa- tion services

(create, pro cess, manage, and pro vide information), and material/information services, the mixture of oth. assume that etri net, whic represen ts the eha vior of service, con tains one input place (i.e., place with no incoming arcs) and one output place (i.e., place with no outgoing arcs). etri net with one input place, for absorbing information, and one output place, for emitting information, will facilitate the denition of the comp osition op erators (see Sec- tion 3.2) and the analysis as ell as the erication of certain prop erties (e.g, reac habilit deadlo k, and liv

eness). an giv en time, eb service can in one of the follo wing states: NotInstantiated Ready Running Suspended or Completed (similar to the ones dened in (Sc uster, Georgak op oulos, Cic ho ki Bak er 2000)). When eb service is in the Ready state, this means that tok en is in its corresp onding input place, whereas the Completed state means that there is tok en in the corresp onding output place. Denition 2.1 (Service Net) servic net is lab ele Plac e/T ansition net, i.e., tuple i; o; wher e: is nite set of plac es, is nite set of tr ansitions epr esenting the

op er ations of the servic e, is set of dir cte ar cs (ow elation), is the input plac with x; is the output plac with o; and is lab eling function wher is set of op er ation names. We assume that and denotes silent op er ation. The o relation can also in terpreted as function will use oth in terpretations throughout this pap er. Silen op erations are transition rings that cannot observ ed. They are used to distinguish et een ex- ternal and in ternal eha vior of the service. Note that the service net, dened ab e, is an or dinary etri net, that is, there is at

most one directed arc linking place to transition or transition to place. giv no formal denition of eb service. Denition 2.2 (W eb Service) Web servic is tuple ameS; esc; Loc; L; S; wher e: ameS is the name of the servic e, use as its unique identier, esc is the description of the servic pr ovide d. It summarizes what the servic oers, Loc is the server the servic is lo ate in, is the invo ation of the Web servic e, is set of its omp onent servic es. If ameS then is asic servic e. Otherwise is omp osite servic e, and i; o; is the servic net mo deling the dynamic

ehavior of the servic e. The place is considered as the initial marking of the service (i.e., only con tains tok en). The execution of starts when tok en is in the place and terminates when tok en reac hes the place Comp osing eb Services eb service has sp ecic task to erform and ma dep end on other eb services, hence eing comp osite. or example, compan that is in terested in selling oks could fo cus on this asp ect while outsourcing other asp ects suc as pa ymen and shipmen t. The comp osition of or more services generates new service pro viding oth the original individual eha v-

ioral logic and new collab orativ eha vior for car- rying out new comp osite task. This means that existing services are able to co op erate although the co op eration as not designed in adv ance. Service comp osition could static (service comp onen ts in- teract with eac other in pre-negotiated manner) or dynamic (they disco er eac other and negotiate on the y). In this section presen an algebra that allo ws the creation of new alue-added eb services using existing ones as building blo ks. Sequence, alterna- tiv e, iteration, and arbitrary sequence are ypical con- structs sp

ecied in the con trol o w. More elab orate op erators, dealt with in this pap er, are parallel with comm unication, discriminator, selection, and rene- men t. also giv formal seman tics to the pro- osed algebra in terms of etri nets as ell as some nice algebraic prop erties. 3.1 eb Service Algebra describ elo the syn tax and informal seman tics of the service algebra op erators. The constructs ere hosen to allo common and adv anced eb service com binations. The set of services can dened the follo wing grammar in BNF-lik notation: ::= S p; p; )] ef S; a;

where: represen ts an empty service, i.e, service whic erforms no op eration. represen ts service onstant used as an atomic or basic service in this con text.
Page 3
represen ts comp osite service that er- forms the service follo ed the service i.e., is an op erator of se quenc represen ts comp osite service that e- ha es as either service or service Once one of them executes its rst op eration the sec- ond service is discarded, i.e., is an alternative (or choic op erator. represen ts comp osite service that er- forms either the service follo ed the service or follo ed i.e.,

is an unor der se quenc (or an arbitr ary se quenc op erator. S represen ts service that erforms certain um er of times the service i.e., represen ts an iter ation op erator. represen ts comp osite service that er- forms the services and indep enden tly from eac other with ossibilities of comm unication er the set of pairs of op erations, that is, is ar al lel op erator with ommunic ation represen ts comp osite service that aits for the execution of one service (among the services and efore activ ating the subse- quen service i.e., is discriminator op- erator. Note that and are

erformed in parallel and without comm unication. )] is comp osite service that dynamically selects one service pro vider among ailable services and exe- cutes it. It eha es as follo ws: rst request is sen comp oser to ailable service pro viders of giv en trading comm unit through their en try access oin ts Then based on the receiv ed resp onses, from their exit access oin ts and according to giv en rank- ing criteria (e.g. price, deliv ery date/time, or com bination of oth) the est service pro vider is hosen. Finally the needed op erations are er- formed. [:] is an op erator of sele

ction ef a; represen ts comp osite service that eha es as except for op erations in with lab el that are replaced the non empt service ef is enement op erator. The prop osed algebra eries the closure prop ert It guaran tees that eac result of an op eration on ser- vices is service to whic can again apply algebra op erators. are th us able to build more complex services aggregating and reusing existing services through declarativ expressions of service algebra. 3.2 ormal Seman tics In this section, giv formal denition, in terms of etri nets, of the comp osition op

erators. Let ameS esc Loc with for ::; eb services suc that and for It is imp ortan to note that service comp osition, as will describ ed elo w, applies to syn tactically dif- feren services. This is due to the fact that the places and transitions of the comp onen services ust disjoin for prop er comp osition. Ho ev er, service ma comp osed with itself. ypically this situation ccurs when services describ arian ts of the same op- eration (e.g., normal execution and exceptional situ- ations) or, for instance, if single supplier oers dieren go ds, the requests ma handled indep en-

den tly as though they ere from dieren sup- pliers. In this case, the erlapping ust resolv ed prior to comp osition. This can accomplished re- naming the sets and of one of the equal services. The services remain equal up to isomorphism on the names of transitions and places. Note also that, in case of silen op erations, represen graphically the corresp onding transitions as blac rectangles. 3.2.1 Basic Constructs Empt Service. The empt service is service that erforms no op eration. It is used for tec hnical and theoretical reasons. Denition 3.1 The empty servic is dene

as ameS; esc; Loc; L; S; wher e: ameS Empty esc \Empty Web Servic e", Loc Null stating that ther is no server for the servic e, Null stating that ther is no URL for Empty and p; p; Graphically is represen ted the etri net of Figure con taining only one place. o2 . . . . . . o3         i2 . . . i1 . . . . . . i3 . . . . . . . . . . . . . . . o1 . . . . . . . . . . . . . . . S2 S3 S1 Figure 1: Services and Sequence. The sequence op erator allo ws the execu- tion of services and in sequence, that is, one after another. ust completed efore can start. This is ypically the

case when service dep ends on the output of the previous service. or example, the service Payment is executed after the completion of the service Delivery Denition 3.2 The servic is dene as ameS; esc; Loc; L; S; wher e: ameS is the name of the new servic e, esc is the description of the new servic e, Loc is the lo ation of the new servic (may at the same server as one of the two omp onent servic es), is the invo ation of the new servic e, i; o; wher e: t; and t; Graphically giv en and (see Figure 1), is represen ted the etri net sho wn in Figure 2. Alternativ e. The alternativ

op erator ermits, giv en services and to mo del the exe- cution of either or but not oth. or in- stance, the assess claim service is follo ed ei- ther the service indemnify customer or the service convoke customer
Page 4
                . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S2 i2 o = o2 o1 S1 i = i1 Figure 2: Service Denition 3.3 The servic is dene as ameS; esc; Loc; L; S; wher e: ameS is the name of the new servic e, esc is the description of the new servic e, Loc is the

lo ation of the new servic e, is the invo ation of the new servic e, i; o; wher e: i; i; i; and Graphically giv en and (see Figure 1), is represen ted the etri net sho wn in Figure 3(a).                                                                                                                                                          (b) (a) i1 . . . . . . . . . . . . . . . i2 . . . . . . . . . . . . . . . . . . S1 S2 p3 p1 p2 o1 o2 p5 p4

o1 i1 . . . . . . . . . . . . . . . o2 i2 . . . . . . . . . . . . . . . . . . S1 S2 Figure 3: Services (a) and (b) Arbitrary Sequence. The arbitrary sequence op er- ator sp ecies the execution of services that ust not executed concurren tly that is, giv en ser- vices and ha either follo ed or follo ed Supp ose, for instance, that there are go ds, then acquiring single go is useless unless the rest of the conjuncts can also acquired. Moreo er, without deadline, there is no enet making the requests in parallel, and doing so ma lead to unnecessary costs if one of the conjuncts is

una ailable or unobtainable. Therefore, the optimal execution is necessarily an arbitrary serial ordering of requests to suppliers. Denition 3.4 The servic is dene as ameS; esc; Loc; L; S; wher e: ameS is the name of the omp osite servic e, esc is the description of the omp osite servic e, Loc is the server wher the omp osite servic is lo ate d, is the invo ation of the omp osite servic e, i; o; wher e: i; o; i; and Graphically giv en and (see Figure 1), is represen ted the etri net sho wn in Figure 3(b). Iteration. The iteration op erator mo dels the execu- tion of service

follo ed certain um er of times itself. ypical examples where iteration is required are comm unication and qualit con trol where services are executed more than once. Another example is when ordering sev eral pro cessors, the service Order Processor is executed sev eral times. Denition 3.5 The servic S is dene as S ameS; esc; Loc; L; S; wher e: ameS is the name of the new servic e, esc is the description of the new servic e, Loc is the lo ation of the new servic e, is the invo ation of the new servic e, i; o; wher e: i; i; t; and t; Graphically giv en (see

Figure 1), S is repre- sen ted the etri net sho wn in Figure 4(a).               !!! !!! !!! !!! """" """" """" ### ### ### $$$$ $$$$ $$$$ %%% %%% %%% &&& &&& &&& ''' ''' ''' . . . . . . . . . . . . . . . i1 o1 S1 (a) (b) o1 i1 . . . . . . . . . . . . . . . o2 i2 . . . . . . . . . . . . . . . . . . p1 p2 S1 S2 Figure 4: Services (a) S and (b) 3.2.2 Adv anced Constructs arallelism with Comm unication. The parallel op erator represen ts the concurren execution of services. Concurren services

ma sync hronize and exc hange information. or instance, in an Online Computer Store eb service, after receiving an or- der for computer from customer, parallel services are triggered: Order Monitor and Order Processor
Page 5
Denition 3.6 et ; ; set of ommunic ation elements. The servic is dene as ameS; esc; Loc; L; S; wher e: ameS is the name of the omp osite servic e, esc is the description of the omp osite servic e, Loc is the lo ation of the omp osite servic e, is the invo ation of the omp osite servic e, i; o; wher e: i; i; and Giv en (see Figure

1), and ; is represen ted graphically the etri net sho wn in Figure 4(b). Discriminator. eb services are unreliable; they ha relativ ely high probabilit of failing or of eing unacceptably slo w. Dela ys of only few seconds could result in service pro viders losing signican sums of money or disapp oin ting their customers. Dieren service pro viders ma pro vide the same or similar ser- vices. Therefore, it should ossible to com bine unreliable services to obtain more \reliable" services. The discriminator op erator is used, for instance, to place redundan orders to

dieren suppliers oering the same service to increase reliabilit The rst to erform the requested service triggers the subsequen service and all other late resp onses are ignored for the rest of the comp osite service pro cess. This construct is similar to the discriminator pattern in tro duced in (Aalst, Hofstede, Kiepuszewski Barros 2002) but our implemen tation, in terms of etri nets, is dier- en t. Denition 3.7 The servic is dene as ameS; esc; Loc; L; S; wher e: ameS is the name of the new servic e, esc is the description of the new servic e,

Loc is the lo ation of the new servic e, is the invo ation of the new servic e, i; o; wher e: i; o; i; and Graphically giv en and (see Figure 1), is represen ted the etri net sho wn in Figure 5(a). Note that in is similar to parallel op erator without comm unication, i.e., Selection. Relying on single supplier puts com- pan at its mercy reduce risk, compan should main tain relationships with ultiple suppliers. These suppliers ma e.g., harge dieren prices, prop ose dieren deliv ery dates and times, and ha dieren reliabilities. The selection construct allo ws to ho ose

the est service pro vider, using ranking crite- ria, among sev eral comp eting suppliers to outsource particular op eration. Denition 3.8 et servic ac ess oints of for ::; The servic )] is dene as )] ameS; esc; Loc; L; S; wher e: ameS is the name of the omp osite servic e, esc is the description of the omp osite servic e, Loc is the lo ation of the omp osite servic e, is the invo ation of the omp osite servic e, =1 i; o; wher e: =1 i; o; p; =1 t; u; =1 i; u; t; p; and =1 t; send req serv u; select serv Graphically giv en and for ::; )] is represen ted the etri net sho wn in

Figure 5(b). allo the disco ery of the service, assume that ser- vice pro viders register their services with particular trading comm unit (Benatallah et al. 2000, Benatal- lah et al. 2002) and that information ab out ailable services is kno wn. also assume that eac ser- vice pro vider con tains distinct parts, one part to pro cess the service request and the other part to erform the service itself. The pair of op erations send req serv and select serv represen ts prede- ned selection strategy hosen, among others, the designer of the comp osite service. The selection decision ma also

based on automated negotiation and auctions (e.g, sealed bid and op en-cry auctions) where the parties in olv ed, that is, the comp oser (auctioneer) and the service pro viders, are automated. By doing so, automated online auctions ecome fundamen tal building blo for selection decision. Renemen t. The renemen construct, in whic op- erations are replaced more detailed non empt ser- vices, is used to in tro duce additional comp onen ser- vices in to service. Renemen is the transformation of design from high lev el abstract form to lo er lev el more concrete form hence

allo wing hierarc hical mo deling. Denition 3.9 et The servic ef a; is dene as ef a; ameS; esc; Loc; L; S; wher e: ameS is the name of the ene servic e, ameS is the name of the ene servic e, esc is the description of the ene servic e, Loc is the lo ation of the ene servic (may at the same lo ation as ), is the invo ation of the ene servic e, if other ise i; o; wher e: p;
Page 6
                          

                                                                       . . . o1 . . . . . . . . . . . . . . . S1 i1 o2 i2 . . . . . . . . . . . . . . . . . . S2 S3 i3 . . . . . . p1 p2 (a) (b) i1 . . . . . . . . . . . . . . . . . . . . . . . . send_req_serv rec_req_1 rec_req_n send_resp_n send_resp_1 . . . . . . S1 Sn . . . . . . . . . select_serv . . . o3 Figure 5: Services (a) and (b) )] x; x; x; if x; if >; or

x; if >; or and t; if >; or and other ise and if if x; or and Figure sho ws an example of rened service where (p erforming either assess simple claim or assess complex claim is renemen (i.e., sp cial- ization of the generic op eration assess claim in Another in teresting use of the renemen op- erator is its com bination with the selection op- erator for est outsourcing op erations. In   assess_simple_claim convoke_customer assess_complex_claim indemnify_customer Ref(S1,assess_claim,S2) i2 assess_simple_claim assess_complex_claim o2 S2 indemnify_customer i1

assess_claim convoke_customer o1 S1 Figure 6: An Example of Renemen ef assess claim ]), for instance, assess claim is erformed the suitable service hosen among ailable assessors. The comp osite eb service generated using the ab constructs is service net that has sp ecial places, whic corresp ond to the eginning and termi- nation of the pro cessing of the comp osite eb service. This in teresting prop ert relates to the structure of the underlying etri net. Prop osition 3.1 (Preserv ation) The servic net of omp osite Web servic obtaine using the ab ove dene onstructs ontains one

input plac and one output plac Pro of. Immediate consequence of Denition 2.2 and of the denitions of the comp osition constructs dened previously In what follo ws, giv an example of comp osite service mo deled as etri net to illustrate some of the constructs dened ab e.
Page 7
Example 3.1 Figure sho ws eb service comp osed of three ba- sic services, represen ting an Online Computer Store and and represen ting resp ectiv ely the Sony Monitors and the Intel Processors Up on reception of an order rec ord PC for computer from customer, starts, in parallel,

the out- sourced services to order monitor and to order pro cessor erforming the op erations send ord mon and send ord pr resp ectiv ely The set of comm unication elemen ts are send ord mon rec ord mon ), send del mon rec del mon and send ord pr rec ord pr ), send del pr rec del pr Once the requested items are receiv ed, OCS erforms the assemble PC op eration. Note that, for the sak of simplicit and clarit not all the op erations of the scenario (e.g, deliv ery and billing) are represen ted and lab els are used instead of names for the transitions.           

     rec_ord_pr rec_ord_mon send_ord_mon send_del_pr rec_del_mon rec_del_pr send_del_mon SM IP COS assemble_PC send_ord_pr rec_ord_PC OCS || IP Figure 7: Service Let us assume no that, instead of basic ser- vice Sony Monitors use comp osite service Monitors The monitor pro vider can dynam- ically selected among the ailable services (e.g., whic are mem ers of Monitors (see Figure 8). Note that send ord mon send req serv ), rec del mon 3.3 Algebraic Prop erties Our seman tics can used to pro algebraic prop- erties of the constructs. As general observ ation, it ma the case that designer

pro duces com- plex eb service com bining set of existing eb services using the algebra op erators. The algebraic prop erties could then used to transform and op- timize comp osed eb services based on comp onen eb services op erational metrics suc as cost and duration, although not in estigated in this pap er. description of the algebraic prop erties of the op- erators in tro duced ab is giv en in able (where =1 stands for )]). The alge- braic op erators satisfy some usual prop erties, suc as comm utativit asso ciativit and reexivit (1) (2) (3) (4) (5) (6) (7) (8) " (9) (10)

(11) (12) (13) (14) (15) 8f :::; :::; 2f ;:::;i =1 (16) If then =1 =1 ;i (17) ef a; if (18) (from (8) and (4)) (19) (from (8) and (6)) (20) (from (8), (2), (3), and (6)) (21) able 1: Desired Prop erties of the Service Algebra Simple prop erties can easily deriv ed, for exam- ple (2), (3), (9), and (12). range of other equations can deriv ed, eac an easy case analysis. An in teresting observ ation is that our seman tics equates the services and (prop ert (8)). This means that arbitrary se- quence can expressed using sequence and alter- nativ e. Ho ev er, it is still useful to include the arbi-

trary sequence op erator in the algebra since the ob vi- ous implemen tation will inecien for the service ). eb Service Analysis comp osite eb service is system that consists of sev eral conceptually autonomous but co op erating units. It is dicult to sp ecify ho this system should eha and ensure that it eha es as required the sp ecication. The reason is that, giv en the same in- put and initial state, the system ma pro duce sev eral dieren outputs. eb services in teract with eac other for conduct- ing sp ecic task although they are created and de- plo

ed indep enden service pro viders. The pieces of ork within the sp ecic task are tigh tly coupled. or example, an Online Computer Store ma need services pro vided Monitor and Processor eb services. The business logic of those eb ser- vices are indep enden of eac other but, for partic- ular computer request, the pieces of ork, that is, the monitor ordering and the pro cessor ordering, are related. Since eb services whic con tain errors ma lead to angry customers and loss of go dwill, it is th us imp ortan to analyze eb services efore they are put in to op eration. The goal is to pro

vide mec h- anisms to supp ort correct eb service comp osition. The prop erties to eried ma general, suc as absence of deadlo ks and liv elo ks, or application- sp ecic, suc as if customer eep reserving and canceling the reserv ation indenitely in an Online Ticket Sales eb service, sophisticated system migh den serving suc customer after certain
Page 8
                                rec_ord_pr send_ord_mon send_del_pr rec_del_mon . . . rec_req_1 rec_req_n

send_req_serv send_resp_1 send_resp_n . . . . . . . . . . . . rec_ord_mon_n rec_ord_mon_1 send_del_mon_1 send_del_mon_n send_ord_pr rec_del_pr select_serv . . . . . . p1 q1 M1 Mn COS IP assemble_PC rec_ord_PC OCS || IP [M1(p1,q1):Mn(pn,qn)] Figure 8: Service )] um er of reserv ation or cancellation requests. The correctness of services is vital to companies. eb service whic con tains errors (e.g., deadlo k) ma lead to angry customers and loss of go dwill. It is th us imp ortan to analyze service efore it is put in to op eration. The correct termination, is one of the main prop erties prop er

eb service should satisfy The erication of the correctness prop ert consists in hec king whether the underlying etri net is liv and ounded (Murata 1989). Beha vioral equiv alences are useful in erication as they la the conceptual basis for deciding that the eha vior of eb services can consid- ered to \the same". They can also used as to ol for reducing erication eort replacing the etri net of service smaller (in size), but \equiv alen t" one. The bisimulation equiv alence (P ark 1981, Milner 1989) et een eb services is relation et een their ev olutions suc that

for eac ev olution of one of the services there is corre- sp onding ev olution of the other service suc that the ev olutions are observ ationally \equiv alen t" and lead to services whic are again bisimilar. This harac- terization of the eha vior of eb services using the notion of bisim ulation helps service designer optimize comp osite services e.g., hanging their comp onen eb services with \equiv alen t" ones. Another moti- ation is customization of services. enhance com- etitiv eness service pro vider ma mo dify his/her service for customer’s con enience (e.g., to conform to his/her wn

business mo del) and this customized service ust conform to the original one. Related ork Sev eral approac hes in estigated the issue of sharing eb services. Ho ev er, none of them oers gen- eral theoretical framew ork for the comp osition of eb services. In this section, briey erview the ap- proac hes that are closely related to our ork. orko managemen systems (Georgak op oulos, Hornic Sheth 1995, WfMC 1999) are based on the premise that the success of an en terprise re- quires the managemen of business pro cesses in their en tiret Indeed, an increasing um er of

organi- zations ha already automated their in ternal pro- cess managemen using orko tec hnology and en- jo ed substan tial enets in doing so. Ho ev er, B2B e-commerce requires the exible supp ort of cross- en terprises relationships. Curren researc eorts in the orko area promise to deliv er next gener- ation orko systems that has the abilit to eas- ily thread together cross-organizational business pro- cesses, supp orting the in tegration of div erse users, applications, and systems. The purp ose of cross- organization orko ws is to

automate business pro- cesses that in terconnect and manage comm unication among disparate systems. Early pro jects in this direc- tion fo cus mostly on the in tegration of small um ers of tigh tly coupled business pro cesses. Recen emerg- ing orko pro jects suc as CMI (Georgak op oulos, Sc uster, Cic ho ki Bak er 1999), eFlo (Casati, Il- nic ki, Jin Shan 2000), and CrossFlo (Ludwig Honer 1999, Honer, Ludwig, Gulcu Grefen 2000) fo cus on lo osely coupled pro cesses. Ho ev er, they lac formal mo del for sp ecifying and erifying eb services. They also do not address the

seman tics of eb service comp osition op erators. Other approac hes mo deling business pro cesses us- ing etri nets are in the orko setting. orko nets, class of etri nets similar to our service net, ha een in tro duced in (Aalst 1997, Aalst 1998) for the represen tation and erication of orko pro- cesses. In (Adam, Alturi Huang 1998), the authors dev elop ed etri net-based approac that uses sev- eral structural prop erties for iden tifying inconsisten dep endency sp ecication in orko w, testing for
Page 9
its safe termination, and hec

king for the feasibilit of its execution for giv en starting time when tem- oral constrain ts are presen t. An implemen tation is pro vided for conducting the ab analyses. Ho w- ev er, the approac is restricted to acyclic orko ws. Nev ertheless, these previous approac hes treat ev ery orko pro cess as separate en tit They do not consider business pro cess comp osition. eb service in tegration requires more complex functionalit than SO AP WSDL, and UDDI can pro- vide. The functionalit includes transactions, ork- o w, negotiation, managemen t, and securit There are sev

eral eorts that aim at pro viding suc func- tionalit for example, the recen tly released Busi- ness Pro cess Execution Language for eb Services (BPEL4WS) whic represen ts the merging of IBM’s eb Services Flo Language (WSFL) (Leymann 2001) and Microsof ’s XLANG (Thatte 2001), is o- sitioned to ecome the basis of standard for eb service comp osition. These languages are based on SO AP+WSDL+UDDI basic stac k, are complex pro- cedural languages, and ery hard to implemen and deplo There are also some prop osals suc as AML- whic is part of ARP Agen Markup Lan- guage pro ject that aims at

realizing the Seman tic eb concept. Ho ev er, AML-S is complex pro cedural language for eb service comp osition. Conclusions In this pap er, prop osed etri net-based algebra for comp osing eb services. The formal seman tics of the comp osition op erators is expressed in terms of etri nets pro viding direct mapping from eac op erator to etri net construction. Th us, an ser- vice expressed using the algebra constructs can translated in to etri net represen tation. By means of set of algebra prop erties, are able to transform and optimize eb service expressions guaran teeing the same seman tics

of initial expressions. In addi- tion, the use of formal mo del allo ws the erication of prop erties and the detection of inconsistencies oth within and et een services. There are other issues in B2B e-commerce whic h, eliev e, could successfully addressed extend- ing the framew ork presen ted in this pap er, e.g., to in- clude managemen of time and resources. exp ect that these problems can dealt with using suitable high-lev el etri net, suc as timed and coloured etri nets (Jensen 1997). Another issue is that strong link with pro cess algebras, suc as CCS (Milner 1989) and CP

(Bergstra Klop 1985), should allo one to imp ort pro cess algebra sp ecic erication tec hniques (e.g., axiomatizations of eha vioral equiv alences) in to the etri net based framew ork. Finally comp osing eb services that satisfy requiremen ts suc as trans- actional pro cess, negotiation, exception handling, and securit through externally visible in teraction or con- ersation is hallenging issue. Ac kno wledgmen ts The authors ould lik to thank Asso ciate Professor Arth ur ter Hofstede from Queensland Univ ersit of ec hnology for his constructiv criticisms on an ear- lier ersion

of this pap er. References Aalst, W. v. d. (1997), erication of orko Nets, in Azema G. Balb o, eds, ‘Pro ceedings of ttp://www.ibm.com/dev elop erw orks/library/ws-bp el/. ttp://www.daml.org/services/. the Application and Theory of etri Nets’97’, oulouse, rance. Aalst, W. v. d. (1998), ‘The Application of etri Nets to orko Managemen t’, The Journal of Cir- cuits, Systems and Computers (1), 21{66. Aalst, W. v. d. (1999), ‘In terorganizational ork- o ws: An Approac based on Message Sequence Charts and etri Nets’, Systems nalysis Mo d- el ling Simulation 34 (3),

335{367. Aalst, W. v. d., Hofstede, A. t., Kiepuszewski, B. Barros, A. (2002), orko atterns, ec hnical Rep ort FIT-TR-2002-02, Queensland Univ ersit of ec hnology Brisbane, Australia. Adam, N., Alturi, V. Huang, W.-K. (1998), ‘Mo d- eling and Analysing of orko ws Using etri Nets’, Journal of Intel ligent Information Sys- tems 10 (2), 131{158. Benatallah, B., Dumas, M., Sheng, Q. Ngu, A. (2002), Declarativ Comp osition and eer-to- eer Pro visioning of Dynamic eb Services, in ‘Pro ceedings of the 18th In ternational Confer- ence on Data Engineering (ICDE’02)’, IEEE Computer So

ciet California, USA, pp. 297{308. Benatallah, B., Medjahed, B., Bouguetta a, A., El- magarmid, A. Beard, J. (2000), Comp osing and Main taining eb-based Virtual En terprises, in ‘Pro ceedings of the orkshop on ec hnologies for E-Services (in Co op eration with VLDB’00)’, Cairo, Egypt. Bergstra, J. Klop, J. (1985), ‘Algebra of Comm uni- cating Pro cesses with Abstraction’, TCS 37 77{ 121. Casati, F., Ilnic ki, S., Jin, L. Shan, M.-C. (2000), An Op en, Flexible, and Congurable System for E-Service Comp osition, ec hnical Rep ort HPL- 2000-41, HP Labs. Georgak op oulos, D., Hornic k, M.

Sheth, A. (1995), ‘An Ov erview of orko Managemen t: rom Pro cess Mo deling to orko Automation In- frastructure’, Distribute and Par al lel Datab ases (2). Georgak op oulos, D., Sc uster, H., Cic ho ki, A. Bak er, D. (1999), ‘Managing Pro cess and Service usion in Virtual En terprises’, Information Sys- tems, Sp cial Issue on Information Systems Sup- ort for Ele ctr onic Commer 24 (6), 429{456. Honer, Y., Ludwig, H., Gulcu, C. Grefen, (2000), Arc hitecture for Cross-Organizational Business Pro cesses, Researc rep ort, IBM, Zuric h, Switzerland. Jensen, K. (1997), Colour

Petri Nets: Basic Con- epts, nalysis Metho ds and Pr actic al Use. ol. 1, Basic Conc epts EA TCS Monographs on The- oretical Computer Science, Springer-V erlag. Leymann, F. (2001), Web Servic es Flow anguage (WSFL) version 1.0 IBM Soft are Group. ttp://ibm.com/soft are/solutions/w ebservices /p df/WSFL.p df/. Ludwig, H. Honer, Y. (1999), Con tract-based Cross-Organisational orko ws The Cross- Flo Pro ject, in D. Georgak op oulos, W. Prinz A. olf, eds, ‘Pro ceedings of the In ternational Join Conference on ork Activities Co ordina- tion and Collab oration (W CC’99)’, San ran-

sisco, USA.
Page 10
Milner, R. (1989), Communic ation and Concurr ency Pren tice Hall In t. Murata, T. (1989), etri Nets: Prop erties, Analysis and Applications, in ‘Pro ceedings of the IEEE’, ol. 77(4), pp. 541{580. ark, D. (1981), Concurrency and Automata on In- nite Sequences, in Deussen, ed., ‘Pro ceedings of the 5th GI Conference’, LNCS 104, Springer erlag. eterson, J. (1981), Petri Net The ory and the Mo del- ing of Systems Pren tice Hall, Englew Clis. etri, C. (1962), Komm unik ation mit Automaten, PhD thesis, Univ ersit of Bonn, German (In German). Reisig, W.

(1985), Petri Nets: Intr duction EA TCS Monographs on Theoretical Computer Science ol.4, Springer-V erlag, Berlin, German Sc uster, H., Georgak op oulos, D., Cic ho ki, A. Bak er, D. (2000), Mo deling and Comp os- ing Service-based and Reference Pro cess-based Multi-en terprise Pro cesses, in ‘Pro ceedings of the 12th Conference on Adv anced Information Systems Engineering (CAiSE’00)’, Sto kholm, Sw eden. Thatte, S. (2001), XLANG: Web Servic es for Busi- ness Pr ess Design Microsoft Corp oration. ttp://www.gotdotnet.com/team/xml wssp ecs /xlang-c/default.h tm/. WfMC (1999), Workow

Management Co alition, erminolo gy and Glossary Do cumen Num er WFMC-TC-1011. ttp://www.wfmc.org/standards/do cs.h tm/.