A Tool for Comparing and Selecting Workﬂow Engines

Karim Ba

ına

ENSIAS, Universit

e Mohammed V - Souissi, BP 713 Agdal - Rabat, Morocco

Keywords:

Workﬂow management systems, Workﬂow management systems evaluation, Workﬂow management systems

comparison criteria.

Abstract:

The task of selecting a workﬂow engine becomes more and more complex and risky. For this reason, organ-

isations require a broad, and a clear vision of which workﬂow engines are, and will continue to be, suitable

for changing requirements. This paper presents a workﬂow engines comparison model to analyse, compare,

and select business process management modelling and enactment engines (Workﬂow Engines or WFEs)

according to user speciﬁc requirements. After the description of the underlying model itself, we present the

implementation of this workﬂow engines comparison model through our multi-criteria workﬂow engines com-

parison and selection prototype WFESelector. The later proposes two scenarios for selecting relevant WFE :

either to express dynamically multi-criteria query upon a WFE evaluation database, or to browse the whole

WFE classiﬁcation through a reporting aggregation based dashboard. WFESelector is subsequently experi-

mented to assess criteria satisfaction on a very large number of open source workﬂow engines (as numerous

as 35).

1 INTRODUCTION

Enterprise business models are becoming more and

more complex, involving numerous interacting ap-

plications within rich business and technical con-

texts. Thus, enterprise business processes inherit this

business models growing complexity. Since those

business processes are considered in the enterprise

business core strategy, they need workﬂow engines

that handle suitably the continuously growing busi-

ness processes management complexity. Nowadays,

business process management engines and Workﬂow

Management Engines (WFE) become numerous, het-

erogeneous, and provide several functionalities which

make the task of selecting a WFE complex and risky.

For this reason, organisations require a broad, and a

clear vision of which workﬂow engines are, and will

continue to be, suitable for changing requirements.

Selection of such workﬂow engines is strategic for

nowadays organisations.

Our contribution is at one side, to present our

workﬂow engine comparison criteria model, and its

implementation within WFESelector prototype, our

multi-criteria workﬂow engines comparison and se-

lection tool, and at the other side, to experiment the

strength of our approach and prototype. The pa-

per is organised as follows : section 2 discusses re-

lated works, section 3 presents our workﬂow engines

comparison model, section 4 illustrates the approach

through the results of our workﬂow engines compar-

ison model experiment, and section 5 concludes and

gives an outlook for this work.

2 RELATED WORKS

Many other works have studied the problem of pro-

ducing a generic workﬂow engine comparison model,

but none of them has proposed a rich workﬂow com-

parison model as we propose (35 comparison sub-

criteria). (Lei and Singh, 1997) compares workﬂow

engines according to their process deﬁnition meta-

models through eight criteria : Granularity, Con-

trol Flow, Data Flow, Organisational Model, Role

330

Baïna K. (2007).

WFESelector - A Tool for Comparing and Selecting Workﬂow Engines.

In Proceedings of the Ninth International Conference on Enterpr ise Information Systems - DISI, pages 330-337

DOI: 10.5220/0002397303300337

 SciTePress

Binding, Exception Handling, Transaction Support,

and Commitment Support. (M. Rosemann and zur

Muehlen, 1998) compares workﬂow engines accord-

ing to their organizational meta model, and Process

Meta Model richness and expressivity. (Van Der

Aalst et al., 2003) tackles the comparsion problem

from the point of view of the ability of workﬂow en-

gines to support workﬂow patterns. (Yu and Buyya,

2005) bases its comparison of scientiﬁc workﬂow en-

gines on four criteria : Workﬂow Design, Workﬂow

Scheduling, Fault Tolerance, and Data Movement.

(Stoilova and Stoilov, 2006) focuses its comparison

on a subset of ﬁve sub-criteria of (McCall et al., 1977)

software quality criteria : Reliability, Usability, Efﬁ-

ciency, Maintainability, and Portability.

3 WORKFLOW ENGINES

COMPARISON MODEL

Our workﬂow engines comparison model will be

presented in three steps : ﬁrst, section 3.1 present

the chosen criteria for workﬂow engines comparison,

then section 3.2 shows how key performance indi-

cators are used to evaluate presented criteria, and ﬁ-

naly section 3.3 summarizes our approach through the

presentation of a workﬂow engines comparison meta-

model.

3.1 Workﬂow Engines Comparison

Criteria Model

WFESelector bases its Workﬂow engine comparison

upon 31 criteria and sub-criteria clustered in three

sub-classes as shown in ﬁgure 1: (1) 12 executabil-

ity criteria, (2) 10 vision criteria, and (3) 9 contex-

tual criteria. Those criteria and their sub-criteria are

derived from criteria proposed by WARIA (Workﬂow

And Reengineering International Association) (war, ),

from criteria highlighted by workﬂow research (Re-

ichert and Dadam, 1997; Bernstein et al., 1999; Rus-

sell et al., 2005; Gaaloul et al., 2005; Ba

ına et al.,

2006; Gaaloul and Godart, 2005) from software qual-

ity factors deﬁned by McCall et al. (McCall et al.,

1977), and ﬁnally from auxiliary contextual proper-

ties on Workﬂow engines. These criteria are sum-

marised in ﬁgure 1 and described below.

1. Executability criteria:

(a) API & GUI support:

i. Activity Programming (API Proposition -

WfMC, OMG, Wf-XML, SWAP, etc.-) The API

is the cornerstone of activity programming,

Figure 1: Workﬂow Engine Criteria Classiﬁcation.

with support for deadlines management, and

events mechanisms: Facilities Provided by the

Engine, API Completeness, and Events Pro-

cessing Mechanisms ;

ii. Internet Support (Web/Internet oriented

GUI) The user interface can be operated from

a simple Web browser. This covers the follow-

ing functions: worklist handler, activity exe-

cution, graphical procedure status, and start-

ing a procedure instance.

(b) Execution Properties:

i. Throughput Rates (Efﬁciency) Depends on

the efﬁciency of the engine, and its capac-

ity to run on top of a distributed cluster of

servers: Server Scalability, Distributed archi-

tecture, and Client Implementation ;

ii. Distribution (MOM support, asynchronous

messages support, triggers/notiﬁcations) sup-

port Supports co-operation, within the same

company or set of partners, of several inde-

pendent Workﬂow engines operated by differ-

ent departments (e.g. a manufacturing process

that triggers a purchase order process, and

cascades to an Extranet based supply chain

process): Import/Export, and Automatic Co-

operation Mechanism ;

iii. Dynamic Changes (Modiﬁcation of process

model at runtime) (Reichert and Dadam,

1997) Collaborative and ad-hoc procedures,

even if known in advance, must be deﬁned

on the spot (adaptive workﬂow (Bernstein

et al., 1999)). Since they might be incom-

plete, it is important to be able to change

them dynamically to factor in unforeseen sit-

uations: Change Procedure Network Deﬁ-

nition, Change Variables Deﬁnition, Change

Dispatching Rules, Late Sub-Network Deﬁni-

tion, and Change Activity Implementation ;

WFESelector - A Tool for Comparing and Selecting Workflow Engines

331

iv. Procedure Deﬁnition (Maniability and us-

ability) The best way to shorten process deﬁ-

nition time is to offer a graphical process deﬁ-

nition tool. Further assistance can be provided

if the Workﬂow engine supports additional

features like skip, undo, offered/accepted

states, suspend, delegate and reroute: Graph-

ical Deﬁnition, Assisted Deﬁnition of Rules-

/conditions, Embedded Feature, BPR and

Simulation ;

v. Ready-to-use agents (Maturity and exe-

cutability) The engine is provided together

with ready-to-use agents. No programming

is required to run deﬁned procedures. This

is mandatory for ad-hoc Workﬂow applica-

tions and includes: a worklist handler, and a

standard activity agent for managing the dia-

logue with the user: Worklist Handler, Proce-

dure Network Visualizer, and Standard Activ-

ity Agent.

i. Adaptability (Development framework

proposition) The engine minimises the nec-

essary modiﬁcation effort at requirement

evolution ;

ii. Maintainability (Code readability and docu-

mentation) The engine minimises the effort to

localise and correct bugs ;

iii. Reusability (Extensibility) The engine can be

partially or totally used with an other applica-

tion ;

iv. Reliability (Exception management and fault

tolerance) The engine does what it is sup-

posed to do, and what the user expects it to

do, and it does so without breaking anything

in the process: accuracy, error tolerance, con-

sistency and simplicity in design. (Russell

et al., 2005) details workﬂow execption as 5

types : work item failure, work item deadline

expiry, resource unavailable, external trigger,

and constraint violation).

v. Testability The engine facilitates test proce-

dures.

2. Vision criteria:

(a) Model Richness:

i. Procedure Power (Richness and process

model expressivness) The capacity of the pro-

cedure development environment to express

the real complexity of procedures in all their

smallest details: Network Structure, Variables

Deﬁnition, Exception Processing, Complexity

Management, and Veriﬁcation ;

ii. Standards compliance WfMC, OMG, IETF,

(vision: BPMI, SOA, WARIA);

iii. Data ﬂow variable containers I/O mapping,

through a database/MOM/Broker/XML Bus,

etc ;

iv. Control ﬂow sequence, split (and/or/xor), join

(and/or/xor), iteration, sub-Workﬂow, multi-

ple activity occurrences, etc ;

v. Transactional Flow After a failure, transac-

tional external transitions are ﬁred by external

entities (scheduler, human intervention, etc.)

and allow to the failed activity to interact with

the outside environment to recover a consis-

tent state. The goal is to bring the failed

process back to some semantically acceptable

state. Thus, failure inconsistent state can then

be ﬁxed and the execution resumed with the

hope that it will then complete successfully ;

vi. Dispatching & Organization Representa-

tion (Advanced group management) The ca-

pability to dispatch each individual activity to

the participant with the correct level of ac-

cess rights and appropriate role in the orga-

nization: Dispatching Rules, Organizational

model, Administration and Privacy, Substitu-

tion Rules, and Import from Directories ;

vii. Enterprise Application Integration (EAI)

(Interoperability) Evaluates the tools pro-

vided to take beneﬁts of most modern Enter-

prise Application Integration: message queu-

ing, application adapters, COM/DCOM dis-

tributed integration, CORBA and/or Web Ser-

vices support, and 2PC transaction support ;

viii. Activity Deﬁnition (Model driven or code

driven) Activity implementation (no program-

ming) can be achieved by providing embed-

ded support, libraries of activities, and forms

generation tools integrated with the process

deﬁnition tool. Additional ﬂexibility can be

provided by a library of basic actions that can

be scripted (possibly graphically) to imple-

ment activities: Activity Library, Form Gen-

eration, Action Library, Action Scripting As-

sembly, Integration Tools, and Multi-lingual

Support.

(b) Operations & statistics support:

i. Workﬂow Logs (Event log proposition) Eval-

uates the existence of WfMS logs storing

events occurring during process execution.

Activities are traceable, meaning that the sys-

ICEIS 2007 - International Conference on Enterprise Information Systems

332

tem should in one way or another keep track

of ongoing and past executions ;

ii. Reporting monitor proposition All features

that facilitate operation of the server, and col-

lection and interpretation of statistics on case

processing, including tools provided to min-

imize the cost of communications and work-

station administration: Recovery and Restart,

Statistics Recording, Operation Archiving,

Statistics Processing, and Home Work.

3. Contextual Criteria:

(a) Applicative architecture The engine is well

adapted to applications like e-government, e-

commerce, SCM (Supply Chain Management),

CRM (Custumer Relation Management), KM

(Knowledge Management), QM (Quality Man-

agement), GIS (Geographical Information Sys-

tem), etc ;

(b) Technical architecture:

i. Adopted standard WfMC, OMG, IETF, (vi-

sion: BPMI, SOA, WARIA) ;

ii. Adopted model Workﬂow patterns

(Van Der

Aalst et al., 2003), Petri Nets, Graph theory,

process algebra, PERT, Gantt, BPEL, etc ;

iii. Transactional model During the Workﬂow

execution, an activity can pass through sev-

eral stages deﬁned as activity states (aborted,

failed and completed). The transactional prop-

erties of an activity depend on the set of its

internal states transitions ;

iv. Event log model Richness of the log (in terms

of information it contains), the log follows a

standard format such as WfMC Audit Trail,

etc.

i. Incubator/Initiative OBJECT WEB,

APACHE, etc ;

ii. License APACHE, MIT, IBM, LGPL, GPL,

etc.

iii. Relations and Networking (Historical rela-

tions with other workﬂow engines) Links with

other open source or commercial workﬂow

engines ;

iv. Web Site accessibility and richness ab-

sent/not reliable/not referenced, up to date, re-

liable/rich, etc.

Workﬂow patterns : www.workﬂowpatterns.com

3.2 Workﬂow Engines Comparison Key

Performance Indicators Model

We can divide the presented criteria in two categories:

those one may evaluate through a mark and those

which are contextual and less subject to an effective

mark. For example, what should mean a mark af-

fected to Licencing and Afﬁliation? We consider that

contextual criteria have not to be marked while exe-

cution properties and vision criteria have to be. Each

high level criteria is seen as a hierarchical marked

criterion that aggregates the set of all its submarked

criterion marks trhough a criterion formula (in the

same vain of multidimensional database roll-up ag-

gregation operation). In the following, aggregation

formulas of each hierarchical marked criterion given

at section 3.1 will be presented and explained.

The criterion mark of a hierarchical marked cri-

terion is computed as the uniform average of its sub

marked criterion (which means that each atomic sub

marked criterion has the same weigth - e.g. each

Executability marked sub criterion has the same

weigth

with 12 the number of Executability

marked sub criteria, and each Vision marked sub cri-

terion has the same weigth

with 10 the number of

Vision marked sub criteria). For formulas simplicity,

each marked criterion name will represent, by lan-

guage abuse, its evaluation given ﬂoating mark (e.g.

Executability(w ∈ W FE) denotes Executability cri-

terion mark of the WFE o. For simple visualisation

tuning, h

will denote a 2D (Executabilty × Vision)

space homothety transformation scale, and (−t

, −t

)

will denote a 2D space origin translation.

1. Executability criterion mark is the uniform aver-

age of 2-sub-criteria of API & GUI support, 5-

sub-criteria of Execution Properties, and 5-sub-

criteria of Software Quality.

Executability(w ∈ W FE) =

((

2 ∗API & GU I support +5 ∗Execution Properties +5 ∗So f tware Quality

)(w) ∗ h

)−

(a) API & GUI support(w ∈ WF E) =

(

Activity Programming + Internet Support

)(w)

(b) Execution Properties(w ∈ W FE) =

(

T hroughput Rates + Distribution + Dynamic Changes + Procedure De f inition

)

+ (

Ready−to−use agent s

)(w)

(

Adaptability + Maintainability + Reusability + Rel iability + Testability

)(w)

2. Vision criterion mark is the uniform average of 8-

sub-criteria of Model Richness, and 2-sub-criteria

of Operations & statistics support.

Vision(w ∈ W FE) =

((

8 ∗Model Richness + 2 ∗Oper ations & statistics su pport

)(w) ∗ h

) − t

(a) Mod el Richness(w ∈ W FE) =

(

Procedure Power + Standard s compl iance + Data f low + Control f low

)

WFESelector - A Tool for Comparing and Selecting Workflow Engines

333

+ (

Transactional F low + Dispatching & Organization Re present ation

)

+ (

Enter prise Application Integration + Activity De f inition

)(w)

(b) Operations & statistics support(w ∈ W FE) =

(

Work f low Logs + Reporting monitor proposition

)(w)

3.3 Workﬂow Engines Comparison

Metamodel

Now that we have presented the criteria and their eval-

uation we can present our workﬂow engines com-

parison metamodel which is composed of 6 meta-

concepts, as shown in ﬁgure 2):

Figure 2: WFESelector reporting Metamodel.

• WFE meta-concept: represents a workﬂow engine

entity with its WF E ID, and WFE NAME ;

• CRITERION meta-concept: represents an ab-

stract workﬂow engine criterion entity with

its CRITERION ID, CRITERION NAME, CRI-

TERION DESCRIPTION, and given evaluation

CRITERION COMMENT;

• CONTEXTUAL CRITERION meta-concept:

represents a concrete workﬂow engine textual

criterion entity that describes the evaluation result

of WFE according to a contextual property ;

• HIERARCHICAL CONTEXTUAL CRITERION

meta-concept: represents a hierarchical concrete

workﬂow engine contextual criterion entity that is

composed of many CONTEXTUAL CRITERION

entities ;

• MARKED CRITERION meta-concept: repre-

sents a concrete workﬂow engine criterion en-

tity with its given evaluation ﬂoating CRITE-

RION MARK ;

• HIERARCHICAL MARKED CRITERION

meta-concept: represents a hierarchical con-

crete workﬂow engine marked criterion

entity that is composed of many MAR-

KED CRITERION entities. This hierarchy

aggregates the set of all its SUB MAR-

KED CRITERION.CRITERION MARK through

CRITERION FORMULA (in the same vain of

multidimensional database roll-up aggregation

operation).

4 WORKFLOW ENGINES

COMPARISON MODEL

EXPERIMENT

4.1 Workﬂow Engines Comparison

Model Implementation:

WFESelector

Figure 3 presents the applicative architecture of WFE-

Selector of four layers.

Figure 3: Applicative architecture of WFESelector.

The evaluation database layer manages a rela-

tional database that stores for each studied workﬂow

engine its evaluation information (i.e. the set of con-

textual and marked criteria). Those WFE evaluation

marks are based upon studying their related research

ICEIS 2007 - International Conference on Enterprise Information Systems

334

papers, marketing white papers, and slides, setting up

and testing technically all engines according to a com-

plete case study process (in our case, the case study

was based on ISO 9002 preventive, and corrective ac-

tions circulation process). Evaluation of a WFE is

achieved by giving subjective marks: for each WFE,

and CRITERION a ﬂoating evaluation MARK is given

from 1 to 5 (1 is the worst mark, and 5 is the best

mark). The Data Mart layer manages a multidimen-

tional hypercube that stores for each studied workﬂow

engine its aggregated marks and key performance in-

dicators on the basis of the evaluation database in-

formation. The Querying engine layer provides an

API to query relevantly either the evaluation database

and the data mart. The Reporting Dashboard layer

proposes synthetic visual reports on workﬂow engines

key performane indicators.

WFESelector user, wanting to select a WFE ac-

cording to speciﬁc requirements, has, in fact, two pos-

sible scenarios : (S1) to select some WFE among

those present in WFESelector evaluation database in

order to classify them function of their marked critera

and sub-criteria (through a reporting dashboard), or

(S2) directely to express a multi-criteria query based

on the hierarchical criteria model and to obtain a list

of classiﬁed WFE (through a querying engine).

As a ﬁrst validation step, we have experimented

our WFESelector prototype on a set of open source

workﬂow engines. Nowadays, WFESelector gath-

ers a database of 35 WFE (experimentation sample).

Without loosing in generality, we present an execu-

tion experiment of WFESelector execution scenario

(S1) in which the user selects 9 signiﬁcant WFE (anal-

ysis sample) among WFESelector evaluated WFE.

The following ﬁgures 4 and 5, and table classify

graphically, and describe textually those queried 9

WFE in detail.

WFESelector scenario execution experiment re-

sults informs that generally, priorities taken into ac-

count by open source workﬂow engines software ed-

itors, are nowadays: (1) API & GUI support is the

most common functionality; then (2) the model rich-

ness, then (3) the execution properties; then (4) soft-

ware quality factors; and ﬁnally (5) Operations and

statitics.

Moreover, WFESelector scenario execution ex-

periment shows that Executability KPI is not gener-

ally dependent of Vision KPI. Four WFE classes are

distinguished: (1) class of WFE whose Executabil-

ity and Vision are correlated (SHARK and jBPM); (2)

class of WFE whose Executability is more important

than Vision (Jfolder and OPEN WFE); WFE whose

Vision is more important than Executability (Bigbross

Bossa, Lenya, and Runa); and (4) (OSWORKFLOW

Figure 4: WFESelector Pivot Table, and 2D results exam-

ples of reporting dashboard GUI.

Figure 5: WFESelector Radar Graph results examples of

reporting dashboard GUI.

and XFlow) seams to be an isolated class from the

the other 8 WFE. The user can then conclude from

his/her ﬁrst WFESelector scenario use that SHARK

and jBPM are in the top level distinction, and that

OSWORKFLOW and XFlow are apparently atypical

within the 9 WFE that the user selected as analysis

sample.

WFESelector - A Tool for Comparing and Selecting Workflow Engines

335

5 CONCLUSION

This paper has presented our workﬂow engines com-

parison model for analysing, classifying, compar-

ing, and selecting workﬂow engines according to

speciﬁc requirements. The model has been imple-

mented within WFESelector prototype. Our work-

ﬂow engines comparison model has been experi-

mented upon a set of 35 open source workﬂow en-

gines, and has revealed preliminary criteria based

classiﬁcations among open source workﬂow engines.

Our ongoing work involves the integration of a

thin time dimension into the WFESelector Data Mart

in order to track the evolution of WFE. Another ac-

tual outlook is related to the integration of Data min-

ing within WFESelector reporting process. For ex-

ample, WFESelector should help in automatic selec-

tion of evaluation weigths for roll-up operation in-

stead of taking uniform average. We also aim to in-

tegrate WFESelector within a broader Workﬂow en-

gines portfolio management platform.

In summary we believe that, once the research and

development work on the aspects described above has

been completed, this approach will result in a com-

prehensive platform that can substantially reduce (i)

WFE presenting and understanding effort both for

editors and customers, and (ii) WFE selection effort

and therefore foster the widespread adoption of either

open source or commercial workﬂow technology.

ACKNOWLEDGEMENTS

The author thanks all 2006 and 2007 ENSIAS engi-

neers that participated with their experimental devel-

opments, and evaluations in WFESelector Project.

REFERENCES

Apache Lenya. http://lenya.apache.org.

Bigbross Bossa. http://www.bigbross.com/bossa.

Enhydra Shark. http://www.enhydra.org/workﬂow/shark.

jBPM. http://www.jbpm.org.

JFolder. http://www.jfolder.com.

OPEN WFE. http://web.openwfe.org/display/openwfe/Home.

OpenSymphony OSWorkﬂow.

http://www.opensymphony.com/osworkﬂow.

Runa. http://wf.runa.ru.

Waria workﬂow comparative study.

http://www.waria.com/books/study-2004.htm.

XFlow. http://xﬂow.sourceforge.net.

ına, K., Benali, K., and Godart, C. (2006). DISCOBOLE:

A service Architecture for Interconnecting Workﬂow

Processes. Computers in Industry - Special issue

on Collaborative Environments for Concurrent Engi-

neering, Elsevier Science Publisher, pages 768–777.

Bernstein, A., Dellarocas, C., and Klein, M. (1999). To-

wards adaptive workﬂow systems: Cscw-98 work-

shop report. SIGGROUP Bull., 20(2):54–56.

Gaaloul, W., Ba

ına, K., and Godart, C. (2005). Towards

Mining Structural Workﬂow Patterns. In Andersen,

K. V., Debenham, J. K., and Wagner, R., editors, 16th

International Conference on Database and Expert

Systems Applications (DEXA’05), volume 3588, pages

24–33, Copenhagen, Denmark. Springer-Verlag.

Gaaloul, W. and Godart, C. (2005). Mining workﬂow re-

covery from event based logs. In van der Aalst, W.

M. P., Benatallah, B., Casati, F., and Curbera, F., ed-

itors, Business Process Management, volume 3649,

pages 169–185.

Lei, K. and Singh, M. (1997). A comparison of workﬂow

metamodels.

M. Rosemann, M. and zur Muehlen, M. (1998). Evaluation

of Workﬂow Management Systems - a Meta Model

Approach. Australian Journal of Information Systems,

6(1):103–116.

McCall, J., Richards, P., and Walters, G. (1977). Factors of

software quality. NTIS, 3.

Reichert, M. and Dadam, P. (1997). A Framework for

Dynamic Changes in Workﬂow Management Sys-

tems. In Wagner, R., editor, 8th International Work-

shop on Database and Expert Systems Applications

(DEXA’97), pages 42–48, Toulouse, France. IEEE

Computer Society Press.

Russell, N., van der Aalst, W. M. P., ter Hofstede, A.

H. M., and Edmond, D. (2005). Workﬂow resource

patterns: Identiﬁcation, representation and tool sup-

port. In CAiSE, pages 216–232.

Stoilova, K. and Stoilov, T. (2006). Comparison of work-

ﬂow software products. In International Conference

on Computer Systems and Technologies, CompSys-

Tech’2006.

Van Der Aalst, W. M. P., Ter Hofstede, A. H. M., Kie-

puszewski, B., and Barros, A. P. (2003). Workﬂow

patterns. Distrib. Parallel Databases, 14(1):5–51.

Yu, J. and Buyya, R. (2005). A taxonomy of scientiﬁc work-

ﬂow systems for grid computing. SIGMOD Record,

34(3):44–49.

ICEIS 2007 - International Conference on Enterprise Information Systems

336

WFE NAME License Pros & Cons

Bigbross

Bossa (Big, )

GPL Light weigth easy to use Petri Nets based WFE, manages user roles over activities. Log ﬁle contains

information about persistent progression status that is reloaded at each transition

Enhydra

Shark (Enh, )

(Object Web) LGPL has an XPDL based process deﬁnition language, has a Wf-XML based process interface, is a model driven

engine, is a mature modelling, enactement and administration engine, has a reliable role management.

Possesses good process instance tracking, and visualisation. Has many graphical modelers (e.g. Jawe,

Together), wrappers (e.g. Swish), is integrated by OFBiz, and can be deployed within Pentaho. Has No

transactional properties handled in the process model

jBPM (JBP, ) (Object Web) LGPL Eclipse plugin, has a rich API, enables process application extensibily Model driven engine, describes

rigorously graphically business processes and origanisation roles (Graph Oriented Programming). Enables

process deﬁnition to be modiﬁed at runtime, with logging process deﬁniton versions

Jfolder (Jfo, ) LPDL has a rich API, Web Base engine, JBOSS based, has no procedure programming environment, is a true

model driven engine. No event logs, no reporting tools

Lenya (Apa, ) (Apache)Apache

Software License

is a CMS speciﬁc WFE, is XML based, integrates Cocoon framework, and is supported by Apache Foun-

dation with an Apache License. Lenya dataﬂow enables XML attachment to be created, modiﬁed, and

removed. Lenya is web oriented, attached documents status can be veriﬁed, at each instant. Lenya users

can be notiﬁed by e-mail for an eventual document validation. Lenya has a scheduler providing a log of

achieved tasks on documents with their relative begining and ending dates

OPEN WFE

(OPE, )

BSD OMG Compliant with a majority of Wf patterns (Van Der Aalst et al., 2003). stores its work data either

in XML ﬁles or in relational databases. Workﬂow Participant roles are handled as accounts within LDAP

repository. Manages suitably workitems dispatching between participants

OSWorkﬂow

(OSW, )

OpenSymphony

Software License

OMG compliant, State Machine based Model, several interface mechanisms (SOAP, EJB, Java classes).

Events Journal can be extracted, but no operation can be processed over those events. The Web version is

somehow complex to interact with. Automatic Process centered, and coded driven process deﬁnition and

management

Runa (Run, ) is cross-platform end user workﬂow environment for JBPM, is web oriented, has an interesting user group

(swimlane) based task assignment

XFlow (XFl, ) (Apache)Apache

Software License

based on JBoss, may be used through a Web Service directely via JMS, integrates Log4j. Is extensible and

eases application integration. Provides ﬂexibility in: designing process workﬂow, reconﬁguring processes,

and deﬁning routing rules. Provides visibility into bottlenecks in a process workﬂow. Eases gathering

statistics on the performance of participants in a process workﬂow. Has no procedure programming envi-

ronnement, and is mainly coded driven

WFESelector - A Tool for Comparing and Selecting Workflow Engines

337