业务流程管理BPM的认识(精)

1、Busin ess Process Man ageme nt:A SurveyWil M.P.van der Aalst1,Arthur H.M.ter Hofstede2,and Mathias Weske3IDepartme nt of Tech no logy Man ageme ntEin dhove n Uni versity of Tech no logy,The Netherla nds wmp.v.d.aalsttm.tue .nl2Ce ntre for In formatio n Tech no logy Inno vati onQuee nsla nd Uni versi

2、ty of Tech no logy,Australia .au3Hasso Platt ner In stitute for Software Systems Engin eeri ngUn iversity of Potsdam,Potsdam,Germa ny mathias.weskehpi.u ni-potsdam.deAbstract.Busi ness Process Man ageme nt(BPMi ncludes methods, tech niq ues,a ndtools to support the desig n,en act

3、me nt,ma nageme nt, and an alysis of operati onal busin ess processes .It can be con sidered as an exte nsion of classical Work?ow Manageme nt(WFMsystems and approaches.Although the practical releva nee of BPM is undisputed,a clear de?n iti on of BPM and related acronyms such as BAM,BPA,a nd STPare

4、missi ng.Moreover,a clear scie nti?c foun dati on is miss ing.lnthis paper,we try to demystify the acronyms in this doma in, describethe state-of-the-art tech no logy,a nd ague that BPM could ben e?t fromformal methods/la nguages(cf.Petri n ets,process algebras,etc.Keywords:Busi ness Process Man age

5、me nt,Work?ow Man ageme nt,Formal Methods.1ln troductio nThis volume of Sprin ger Lecture Notes in Computer Scie nee is devoted to the“ Conference on Busin ess Process Man ageme nt:O n the Applicati on of Formal Methodsto ProcessAware Information Systems ” taking place in Eindhoven (The Netherlandsi

6、n June2003.To put the con tributi ons to this conference into perspective,we discuss theideas,technology,and foundations hidden be-hind acronyms likeWFM,BPM,BAM,BPA,STP,etc.The goal of this paper is to provide an overview of the scienti?c and practical issues in the con text of bus in ess process ma

7、n ageme nt systems.Thisway we hope to trigger researchers and practiti oners to address the challe nges in this domain. The de?n iti on of a bus in ess process man ageme nt system used throughout this paperis:a gen eric software system that is drive n by explicit process desig ns to enact and man ag

8、eoperati onal bus in ess processes.The system should be process-aware and gen eric in thesense that it is possible to modify the processes it supports.The process desig ns are ofte ngraphical and the focus is on structured processes that n eed to han dle many cases.W.M.P.van der Aalst et al.(Eds.:BP

9、M2003 丄 NCS2678,pp. 2,2003.c Spri nger-Verlag Berlin Heidelberg20032W.M.P.va n der Aalst,A.H.M.ter Hofstede,a nd M.WeskeTo show the releva nee of bus in ess process man ageme ntsystems,it is in ter-est ing to put them in a historical perspective.。 。 nsider Figure 1,whichshows some of the ongoing tre

10、nds in information systems 3.This ?gure shows that today sinformation systems consist of a number of layers.The center is formed by the operat ingsystem,i.e.,the software that makes the hardware work.The sec ond layer con sists of generic applicati ons that can be used in a wide range of en terprise

11、s.Moreover,these applications are typically used within multiple departme nts within the same en terprise.Examples ofsuch gen eric applicatio ns are a database man ageme nt system,a text editor,a nd aspreadsheet program.The third layer con sists of doma in speci?c applicati on s.Theseapplicati ons a

12、re only used with in speci?c types of en terprises and departme nts.Examplesare decisi on support systems for vehicle routi ng,call cen ter software,a nd huma n resourceman ageme nt software.The fourth layer con sists of tailor-made applica-i on s.Theseapplicati ons are developed for speci?c orga ni

13、 zati ons.operati ngsystemgen ericapplicati onsdoma inspecificapplicatio ns tailor-madeapplicati onsTrends in 1. From program ming toassembli ng.2. From data orie ntati on toprocess orie ntatio n.3. From desig n to redesig nand orga nic growth.Fig.1.Tre nds releva nt for bus in ess process man ageme

14、 nt 3.In the sixties the second and third layer were missingnformation systems were built ontop of a small operati ng system with limited fun ctio nality.Si nce no gen eric nor doma inspeci?c software was available,these systems mai nly con sisted of tail-made applicati on s.Since the n,the sec ond

15、and third layer have developed and the ongoing trend is that the fourcircles are increasing in size,i.e.,they are moving to the outside while absorbing newfunctionality.Today sBusin ess Process Man ageme nt:A Survey3 operat ing system? er much more fun ctionality.Database man ageme nt systems that r

16、eside in the sec ond layer er fun ctio nalitywhich used to be in tailor-made applicati on s.As a result of this tren d,the emphasis shiftedfrom program ming to assembli ng of complex software systems.The challe nge no Ion ger isthe cod ing of in dividual modules but orchestrat ing and gluing togethe

17、r pieces of softwarefrom each of the four layers.Ano ther trend is the shift from data to processes.The seve nties and eighties weredomin ated by data-drive n approaches.The focus of in formati on tech no l-ogy was onstoring and retrieving information and as a result data modeling was the starting p

18、oint forbuild ing an in formati on system.The modeli ng of bus in ess processes was ofte n n eglectedand processes had to adapt to in formati on tech-no logy.Ma nageme nt trends such as bus iness process ree ngin eeri ng illustrate the in creased emphasis on processes.As aresult,system engin eers ar

19、e resort ing to a more process drive n approach.The last trend we would like to men ti on is the shift from carefully pla nned desig ns toredesig n and orga nic growth.Due to the omn iprese nce of the In ter- net and its sta ndards,informatio n systems cha nge on-th?y.As a result,fewer systems are b

20、uilt from scratch .Inmany cases existi ng applicati ons are partly used in the new system.Althoughcomponent-based software development still has its problems,the goal is clear and it is easyto see that software developme nt has become more dyn amic.The trends show n in Figure1provide a historical co

21、n text for bus in ess process manageme nt systems.Bus in ess process man ageme nt systems are either separateapplications residing in the second layer or are integrated components in the do-mainspeci?c applicati on s,i.e.,the third layer.Notable examples of bus in ess proess man agement systems resi

22、di ng in th sec ond layer are work?ow man ageme ntsystems6,19,22,23,24such as Staware,MQSeries,a nd COSA,a nd case ha n-dli ngsystems such as FLOWer.Note that lead ing en terprise resource pla nning systemspopulating the third layer also c? er a work?ow management module. The work?ow enginesof SAP,B

23、aa n,PeopleSoft,Oracle,a nd JD Edwards can be con sidered as in tegrated bus iness process man ageme nt systems.The idea to isolate the man ageme nt of bus in essprocesses in a separate comp onent is con sis-te nt with the three tren dsrided.Bus in essprocess man ageme nt systems can be used to avoi

24、d hard-codi ng the work processes intotailor-made applicati ons and thus support the shift from program ming to assembling.Moreover,process orie ntati on ,redesig n,and orga nic growth are supported.Forexample,today s work?ow management systems can be used to integrate existing applicati ons andsupp

25、ort process cha nge by merely cha nging the work?ow diagra m.I solat ing the man ageme nt of bus iness processes in a separate comp onent is also con siste nt with rece nt developme nts in the domai n ofweb services:Web services composi-tio n Ian guages such as BPEL4WS,BPML,WSCI,XLANG,a ndWSFL can b

26、e used to glue services de?ned using WSDL together.An interesting starting point from a scienti?c perspective is the early wok on o? ce in formatio nsystems .In the seve nties,people like Skip Ellis13,A natol Holt17,a nd Michael Zisman28alreadyworked on so-called o? ce informa-4W.M.P.va n der Aalst,

27、A.H.M.ter Hofstede,a nd M.Wesketion systems,which were drive n by explicit process models .It is in terest ing to see that the three pioneers in this area in depe nden tly used Petri-net varia nts to mode? oe procedures.Duri ng the seve ntiesand eighties there was great opti-mism about the applicabi

28、lity of o? ce informationsystems.Unfortunately,few applications succeeded.As a result of these experie nces,both the applicati onof this tech no logy and research almost stopped for a decade.Co nseque ntly,hardly any adva nces weremade in the eighties .In the nin eties,there aga in was a huge in ter

29、est in ese systems.The nu mber ofwork?ow man ageme nt systems devebped in the past decade and the many papers on work?ow tech nology illustrate the revival of o? ce in formatio n systems.Today work?ow man ageme nt systems are readilyavailable22.However,their applicati on is still limited to speci?c

30、in dustries such as banking and in surance.As was in dicated by Skip Ellis it is importa nt to lear n from these ups and dow ns14.The failures inthe eighties can be expla ined by both tech ni cal and con ceptual problems .In the eighties, n etworks wereslow or not prese nt at all,there were no suita

31、ble graphical in ter-faces,a nd proper developme nt softwarewas miss in g.However,there were also more fun dame ntal problems:a uni ?ed way of modeli ngprocesses was miss ing and the systems were too rigid to be used by people in the workplace.Most of thetech ni cal problems have bee n resolved byno

32、 w.However,the more con ceptual problems remai n. Good sta ndards for bus in ess process modelingare still missing and even today s work?ow managemfoitesystemsunn ecessary con stra ints on the process logic(e.g.,processes are made more seque ntial.2Bus in ess Process Man ageme nt Demysti?edMany peop

33、le con sider Busin ess Process Man ageme nt(BPMto be the“next step ” afterthe work?ow wave of the nin eties.Therefore,we use work?ow term- no logy to de?ne BPM.The Work?owMa nageme nt Coalitio n( WfMCde? nes work?ow as:“ The automatio nof a bus in ess process,i n whole or part,dur- ing which docume

34、nts,i nformatio n or tasks are passed fromone participa nt to ano ther for action, accord ing to a set of procedural rules. ” 22.A Work?owManagement System(WFMSis de?ned as:“ A system thatde?n es,creates and man ages the executio n of work?ows through the use of software,r unning on one ormore work?

35、ow engin es,which is able to in terpret the process de?n iti on, in teract with work?ow participants an d,where required,i nv oke the use of IT tools and applications. ” 22.Note that both de?nitionsemphasize the focus on en actme nt,i.e.,the use of software to support the executio n of operati onalp

36、ro-cesses .In the last couple of years,ma ny researchers and practiti oners started to realize that thetraditional focus on enactment is too restrictive.As a result new terms like BPM have been coined.Thereexist many de?nitions of BPM but in most cases it clearly includes Work?ow Man ageme nt(WFM.We

37、de?ne BPM as follows:Support ing bus in ess processes using methods,tech niq ues,a nd software todesig n,en act,c on trol,a nd an alyze operati onal processes involving huma ns, orga ni zati on s,applications,docume nts and other sources of in form atio n.Note that this de?n iti on restricts BPM to

38、operati onalprocesses,i.e.,processes at theBusin ess Process Man ageme nt:A Survey 5strategic level or processes that cannot be made explicit are excluded.Note thatprocess aware ” ,i.e.,without information about theoperational processes at hand little support isprocess desig n systemcon figurati onp

39、rocessen actme nt diag no sisWorkflowMan ageme nt Bus in ess Process Man ageme ntFig.2.The BPM lifecyle to compare Work?ow Man ageme nt and Busin ess ProcessMan ageme nt.Figure 2shows the relatio nship betwee n WFM and BPM usi ng the BPM lifecyle.TheBPM lifecyle describes the various phases in suppo

40、rt of operati onal bus in ess processesn the desig nphase,the processes are (redesejh In the con? gurati on phase,desig ns are impleme nted by con? guri nga process aware in formatio n system (e.g.,a WFMS.After con? gurati on ,the en actme nt phase startswhere the operati onal bus in ess processes a

41、re executed using the system con? gured .In the diagnosphase,the operati onal processes are an alyzed to ide ntify problems and to ?nd things that can beimproved.The focus of tradi tional work?ow man ageme nt (systemsis on the lower half of the BPMlifecyle.As a result there is little support for the

42、 diag no sis phase.Moreover,support in the desig n phase islimited to providing an editor and analysis and real design support are missing.lt is remarkable that fewWFM systems support simu la-on, veri?cati on,and validatio n of process desig ns.lt is also remarkable thatfew systems support the colle

43、cti on and in terpretati on of real-time data.Note that most WFM systems logdata on cases and tasks executed.However, no tools to support any form of diag no sis are?)ered by thetraditi onal systems.Curre ntly,ma ny work?ow ven dorsare positi oning their systems as BPM systems.Gartner expects the BP

44、M market to grow and also ide nti?es Busin ess Process Analysis (BPAas an importantsystems support ing BPM n eed to bepossible.aspect 16.It is expected that the BPA market will con ti nue to grow.Note that BPA covers aspects neglectedy traditi onal work?ow products (e.g.,diagnosis,simulation,etc.Bus

45、iness Activity Monitoring (BAMisone of the emerging areas in BPA.The goal of BAM tools is to use data logged by the information system todiag nose the operati onal processes.A n example is the ARIS Process Performa nee Ma nager (PPMof IDSScheer 18.ARIS PPM extracts in formation from audit trails (i.

46、e.,information logged during the execution ofcasesand displays this in formatio n in a graphical way (e.g.,?ow times,bottle necks,utilizatio n,etc.BAMalso in cludes processminin g,i.e.,extracti ng process models from logs10.BAM creates a nu mber of scie nti?c and practicalchalle nges(e.g.,which proc

47、esses can be discovered and how much data is n eeded to provide useful informatio n.Whe n it comes to redesig ning operati onal processes two trends can be ide n- ti?ed:Straight ThroughProcessi ng(STPa nd Case Han dli ng(CH.STP refers to the complete automatio n of a bus in essprocess,i.e.,ha ndli n

48、g cases without huma n invo Iveme nt.STP is ofte n only possible if the process isredesig ned.More-over,STP is ofte n only possible for a selected set of cases.The latter means that casesare split into two groups:(1cases that can be han dled automatically(i n Dutch these cases are called“ Gladde gev

49、allen ” and(2cases that remunertvolvement.By separating both groups it is oftenpossible to reduce?ow time and cut costs.While STP strives for more automation,CH addresses theproblem that many processes are much too variable or too complex to capture in a process diagram4.l nCH the no rmal route of a

50、 case is modeled but at the same time other routes are allowed if not explicitlyexcluded.One way to do this is to make work?ows datadrive n rather tha n process-drive n and allow forauthorizations to skip or undo activities.Also the focus is on the case as a whole rather tha n on in dividualwork-ite

51、ms distributed over work-lists.To summarize:BPM exte nds the traditi onal WFM approach by support for the diag no sisphase(cf.BPA and BAM softwarea nd allowi ng for new ways to support operati onal processes(cf.CH andSTP .In the rema in der,we focus on the scie nti?c foun dati ons and curre nt tech

52、no logy.3On the In terplay betwee n Busin ess Process Man ageme nt and Formal MethodsBusin ess Process Models should have a formal foun datio n. Well-k nown reas ons (seee.g.1include:1formal models do not leave any scope for ambiguity,and 2formal models in crease the potential for an alysis(see also

53、 e.g.26.It is desirable that a Bus in ess Process Model can be un derstood by the various stakeholders invoIved in an as straightforward manner as possible.This could e.g.be achieved through the use of graphicalreprese ntati on s.At the same time, these stakeholders should assig n the same meaning t

54、o such amodel,there should not be any scope for alter native in terpretati on s.Busi ness Process Models can bequite complex and the use of a formal la nguage for their speci?cati on is the only sure way to guara nteethat alter native in terpretati ons are ruled out.After consen sus among the stakeh

55、olders has bee nreached,a bus in ess process model can be deployed and if a formal la nguage was used,its behavior canbe expla ined in terms of the formal sema ntics of that speci?catio n Ianguage.As remarked in21,the lack ofa formal semantics has resulted in ?ierent in terpretati ons by ven dors of

56、 eve n basic on trol?ow constructs,de ?n iti ons in n aturalIan guage such as provided by the Work?ow Man ageme nt Coalitio n are not precise eno ugh.As always,it is preferable to identify any problems in software before it is actually deployed .In the caseof Bus in ess Process Models this is especi

57、ally im-porta nt as they may invo Ive core bus in ess an d/orcomplex bus in ess tran sact ions.To reduce the risk of costly correcti on s,a thorough an alysis of a Busin ess Pro-cess Model can beben e?cial.A nalysis of Busin ess Process Models canlso be used to in vestigate ways of improvi ngprocess

58、es(e.g.reduc ing their cost.Formal la nguages may have associated an alysis tech niq ues whichcan be used for inv es-tigat ing properties of speci?cati on s.These tech niq ues can the n be relied upon toprovidesight into the behavior and characteristics of a Bus in ess Process Model speci?ed in such

59、 a Ianguage.In 1three reas ons are stated argu ing the ben e?ts of the use of Petri n ets for the speci?cati on ofwork?ows.The reas ons brought forward are the fact tha? etri n ets are formal,have associated an alysistech niq ues,a nd are state-based rather tha n eve nt- based.The developme nt of Wo

60、?a n( see e.g.25demon strates that work?ows speci?ed as work?ow n ets2,a subclass of Petri n ets,ca n be an alyzed in orderto determinehether they are e.g.deadlock free .In the con text of UML activity diagrams,tool support forveri?cati on is discussed in 15.Through the no ti on of place,Petri n ets