A UNIFIED FRAMEWORK FOR APPLICATION INTEGRATION
An Ontology-Driven Service-Oriented Approach
Saïd Izza, Lucien Vincent, Patrick Burlat
Laboratoire G2I, Ecole des Mines de Saint-Etienne, 158 Cours Fauriel, Saint-Etienne, France
Keywords: Enterprise application, integration, web service, ontology, mediation.
Abstract: The crucial problem of Enterprise Application Integra
tion (EAI) is the semantic integration problem. This
latter is not correctly addressed by today's EAI solutions that focus mainly on the technical and syntactical
integration. Addressing the semantic aspect will promote EAI by providing it more consistency and robust-
ness. Some efforts are suggested to solve the semantic problem, but they are still not mature. This article
will propose an approach that combines both ontologies and web services in order to overcome some issues
related to the semantic integration problem.
1 INTRODUCTION
Over the last decade, a new technology typically
known as Enterprise Application Integration (EAI),
have emerged. In essence, EAI technologies provide
tools to interconnect multiple and heterogeneous
enterprise application systems (EAS) such as ERP
(Enterprise Resource Planning), CRM (Customer
Relationship Management), SCM (Supply Chain
Management), and legacy systems. The most diffi-
culty of these interconnections is that the integrated
systems were never designed to work together.
More recently, Web Services (WS) have
em
erged with the advent and the evolution of the
Internet and they provide a set of standards for EAI.
Even if WSs are not fully mature, they seem to be-
come the linga franca of EAI. This will notably
make integration simpler and easier through using
web protocols and standards.
Despite the whole range of available tools and
wi
despread standards adoption, the main goal of
EAI, which is the semantically correct integration of
EASs, is not yet achieved. Indeed, EAI still provides
technical and syntactical solutions but does not ad-
dress correctly the semantic problem, which consti-
tutes the real integration problem.
Semantic integration becomes very important in
o
rder to overcame semantic heterogeneities within
EAI, and which mainly concern both data and be-
havior of EASs. Although there is some related
works, which concern semantic integration, but there
has been no mature solution that deals correctly with
integration problem.
In this paper, we will focus on the semantic
p
roblem in the context of EAI. Our approach is
based on an extension of service-oriented architec-
ture (SOA), called ODSOI (Ontology-Driven Ser-
vice-Oriented Integration) and which is mainly
based on WSs and ontologies. The rest of this paper
is organized as follows. Firstly, we will present the
integration problem. Secondly, we will briefly re-
view the current state-of-the-art in EAI through pre-
senting two major kinds of solutions: traditional and
Web-Services-based EAI systems. Finally and be-
fore concluding, we will describe some aspects of
our work which attempts to provide a solution for
the integration problem.
2 THE INTEGRATION PROBLEM
Enterprise application systems (EAS) can take many
different types including batch applications, tradi-
tional applications, client/server applications, web
applications, application packages (Izza, 2004).
These systems are often materialized in enterprise
reality in form of ERP, CRM, SCM, and legacy sys-
tems.
An appropriate characterization of EASs in the
conte
xt of EAI is that EASs are HAD (heterogene-
ous, autonomous and distributed) systems (Bussler,
2003):
- Het
erogeneous systems mean that each EAS
implements its own data and process model.
- Aut
onomous systems refer to the fact that each
165
Izza S., Vincent L. and Burlat P. (2005).
A UNIFIED FRAMEWORK FOR APPLICATION INTEGRATION - An Ontology-Driven Service-Oriented Approach.
In Proceedings of the Seventh International Conference on Enterprise Information Systems, pages 165-170
DOI: 10.5220/0002521601650170
Copyright
c
SciTePress
EAS runs independently of any other EAS.
- Distributed systems mean that each EAS locally
implements its data model, which it generally do
not share with other EAS.
The consequence of these characteristics is that
EASs are generally standalone software entities,
which form what we often call islands of informa-
tion and automation. In this case, any form of inte-
gration of the EASs must happen outside of the in-
volved EASs, by using integration systems such as
EAI systems. This integration consists in intercon-
necting the interfaces of each EAS using technolo-
gies supported by the integration systems such as
queuing systems, databases or remote invocations.
The characteristics of EASs form the main rea-
sons of the existence of the integration problem, and
the more these characteristics are extremes, the more
the integration become hard and complex. Despite
the importance of the problems described above, we
will focus, in this paper, only on the heterogeneity
problem, precisely the semantic heterogeneity prob-
lem, which is the hard problem of enterprise integra-
tion in general, and EAI in particular (Bussler,
2003).
Since EASs are HAD, a semantic mediation is
needed in order to achieve their integration. Its aim
is to resolve all the semantic conflicts that can arise
between the exchanged data, and also between in-
voked behavior interfaces. Indeed, data semantic
mediation provides mechanisms to preserve the
meaning of the data during the flow exchanges be-
tween EASs (adressing ), whereas behavior semantic
integration provides mechanisms to resolve the se-
mantic behavior interface heterogeneity when EASs
invoke each other.
Furthermore, the integration problem is more
complicated by our industrial context concerned by a
complex enterprise in the multidisciplinary microe-
lectronics area. This particular context is mainly
characterized by several and heterogeneous knowl-
edge domains that needs sophisticated semantic me-
diation in order to achieve the integration process.
3 INTEGRATION SOLUTIONS
In this section, we will describe the major existing
EAI solutions, which will be followed by some per-
tinent related works about EAI.
3.1 Today's EAI Solutions
In this paper, we will consider only two main impor-
tant solutions in the context of EAI: traditional EAI
systems and WSs. These solutions can fulfill major
integration requirements such as data synchroniza-
tion, business process execution, reconciliation of
technical and syntactic differences, fast deployment
of new applications and so on.
3.1.1 Traditional EAI Systems
Currently, EAI systems are based on a lot of tech-
nologies such as: message brokers, process brokers,
message-oriented middleware, etc. Even if EAI sys-
tems may differ from a technological point of view,
the main functionalities remain the same and we can
mainly distinguish five components, which provide
respectively transport services, connectivity ser-
vices, transformation services, distribution services
and process management services (Erasala, 2003).
The principle of EAI systems is based on using
interfaces (connectors) to integrate EASs. The inter-
faces convert all traffic to canonical formats and
protocols. These interfaces constitute the only mean
to access EASs, and they can occur in different lev-
els: user-interface level, business logic level and
data level (Linthicum, 1999).
Although EAI systems address technical and
syntactical integration, nevertheless they must ad-
dress the semantic level which is more difficult and
which can provide more added value. Today, no
traditional EAI system can provide mechanism that
correctly supports semantics. In best cases, data is
passed between EASs by-value, and in general no
shared semantic concepts are explicitly used to de-
fine semantics through different messages or to se-
mantically describe the behavior that is provided.
3.1.2 Web Services
WSs are considered as a result of convergence of
Web with distributed object technologies. They are
defined as an application providing data and services
to other applications through the Internet (Kadima,
2003). WSs promote an SOA (Service-Oriented Ar-
chitecture) that is based fundamentally on three
roles: service provider, service requestor and service
broker; and three basic operations: publish, find and
bind, and any particular EAS can play any or all
these roles (Kontogiannis, 2002).
WSs constitute the most important concretization
of the SOA model. They can be deployed inside
(EAI) or outside (B2B) the enterprise. In all cases,
WSs are published with appropriate URLs by WS
providers over the Internet or Intranet. Once pub-
lished, these WSs are accessible by WS consumers
via standards Web such as HTTP, SOAP, WSDL
and UDDI. In addition to this, WSs can be used for
integrating EASs via standards such as BPEL or
WSFL.
WSs are very promising in solving the integra-
ICEIS 2005 - DATABASES AND INFORMATION SYSTEMS INTEGRATION
166
tion problem. Today, some new integration products
based on WSs standards exist and will certainly re-
place in the near future the proprietary solutions that
are the traditional EAI systems (BIJOnline, 2004).
Even if WSs are promising, they do not correctly
address the semantic aspect that is currently some-
what supported by UDDI registries with the help of
some standard taxonomies such as NAICS,
UN/SPSC and ISO 3166 (Dogac, 2004). In addition,
WSs do not provide neither data nor behavior me-
diation (Cordoso, 2002) (Fensel, 2002) (Bussler,
2004). These drawbacks are due mainly to the lack
of service ontology and mediation support in current
WSs. This lack penalizes the efficiency of current
WS integration in the context of EAI.
3.2 Related works
Recently, the importance of WSs has been recog-
nized and widely accepted both by industry and aca-
demic research. This section will review some im-
portant related works about enterprise integration,
mainly those that concerns WS-based integration
and ontology-based integration.
In the context of data integration, there are many
general works which use ontology-based approaches
such as COIN project (Goh, 1994), OBSERVER
project (Mena, 1996), INFOSLEUTH project
(Woelk, 1994), BUSTER project (Stuckenschmidt,
2000) and so on. These works are ontology based
but they are not concerned about the mediation in
the context of SOA.
In addition to the listed related works above,
there are some other works that are addressing the
WS viewpoint such as Active XML from GEMO
project (Abiteboul, 2002) and SODIA from IBHIS
project (Turner, 2004). Active XML extends XML
language by allowing embedding of calls to WSs.
SODIA is an implementation of Federated Database
System in the context of WSs. These works do not
support any mediation services.
In the context of application and process integra-
tion, some important initiatives and works exist
around the semantic web service concept (Dogac,
2004) (McIlraith, 2001) that aim to bridge the cur-
rent WS gap such as OWL-S (W3C, 2004), BPEL
(EBPML, 2004), WSMF (Fensel, 2002], SWSI
(SWSI, 2004), METEOR-S (METEOR-S, 2004).
OWL-S provides an ontology markup language in
order to semantically describe capabilities and pro-
prieties of WSs. BPEL is a standard providing a lan-
guage to define business processes that can be used
in application integration. WSMF and SWSI are
initiatives that provide frameworks in order to sup-
port the concept of semantic web service.
METEOR-S is an effort, which provide semantic
web services through the extension of WS standards
(WSDL, UDDI). But, most of these efforts do not
provide mature concepts for mediation, particularly
in the context of EAI.
4 ODSOI APPROACH
This section will succinctly describe the important
characteristics of our approach called ODSOI that
aims to extend the state-of-the-art in EAI in order to
address the semantic problem.
4.1 Global Architecture
First of all, ODSOI approach is a solution to the in-
formation system integration problem. This means
that our approach addresses the heterogeneity prob-
lem by providing a mediation-based solution using
ontology concept. Indeed, our approach is based on
service-oriented since it uses WSs for integrating
EASs. The architecture integration that we suggest is
called ODSOA (ODSO Architecture). This latter
extends SOA with a semantic layer that aims to en-
hance service mediation in the context of EAI.
The ODSOA concept provides a unified frame-
work in order to integrate EASs. In this framework,
three main types of services (Fundamental-Services)
are defined: Data-Services, Functional-Services and
Business-Services. This different types can respec-
tively address data, application and process integra-
tion.
Figure 1: Global View of ODSOA Architecture.
Data-Services (DS) are services that expose data
sources as a service. Functional-Services (FS) are
services that expose application systems, fundamen-
tally functional systems (software that can perform
enterprise functions such as administrative and tech-
nical ones). Business-Services (BS) are defined as
the combination of the above services in order to
expose business processes. Our service typology can
be seen as an extension of the one proposed by
(Turner, 2004) which distinguishes two concepts:
A UNIFIED FRAMEWORK FOR APPLICATION INTEGRATION - An Ontology-Driven Service-Oriented Approach
167
SaaS (Software-as-a-Service) and DaaS (Data-as-a-
Service).
Figure 1, which is a particular SOA, recapitu-
lates these important types of services. Indeed, there
are of course some other important technical ser-
vices that are mainly Brokering-Services, Descrip-
tion-Services, Mediation-Services, Publication-
Services, Discovery-Services and Execution-
Services. Some of them will be described below.
A cross section of the integration bus (also called
ODESB – Ontology-Driven Enterprise Service Bus)
(figure 2) shows many concentric existing standard
layers such as Transport layer, Exchange layer, Reg-
istry layer and Transversal layer.
Figure 2: Cross Section of the ODESB Bus.
In addition to these standard and existing layers, we
suggest to adopt in a similar way as semantic web
services, another layer, called Semantic-Layer,
which includes two sub-layers that are Domain-
Layer and Integration-Layer. The Domain-Layer
aims to describe and publish the three fundamental
services described above using specific descriptions
such as DSD (Data Service Description) for DSs,
FSD (Functional Service Description) for FSs, and
BSD (Business Service Description) for BSs. All
these descriptions exploit some specific ontologies
and are the specialization of OWL-S (Web Ontology
Language-Services). Concerning the Integration-
Layer, it provides some technical services (men-
tioned above) in order to semantically discovery,
mediate and execute fundamental services that are
described and published by the layer above (domain
layer). In the next section, some important technical
services of the Semantic-Layer will be developed.
4.2 Semantic Layer Services
Semantic-Layer services are the main services that
address the semantic problem.
They are divided into Domain-Layer-Services
and Integration-Layer-Services. The most important
technical service of each layer (which are Descrip-
tion-Services and Mediation-Services) will be de-
scribed below.
4.2.1 Description Services
The principle of ODSOA is based on the use of
some knowledge registries that store some formal
ontologies, which are exploited by Description-
Services in order to define the semantic description
of services. According to Gruber, an ontology is
defined as an explicit and formal specification of a
conceptualization
(Gruber, 1993), and for our pur-
pose, we have defined three major types of ontolo-
gies: information or data-based ontologies, behavior
or functional–based ontologies and process or busi-
ness-based ontologies.
Data-based ontologies are the most basic ones.
They provide semantic description of the data. The-
ses ontologies are required in all cases, no matter if
we leverage functional-based or business-based on-
tologies.
Functional-based ontologies define semantic de-
scription around functions that are provided by the
multiple EASs (and then services) and that can be
remotely invoked. These ontologies are generally
required in order to provide a better reuse of func-
tionalities.
Business-based ontologies define semantic de-
scription around coordinating business processes.
These ontologies are generally required in order to
integrate both business processes and applications.
Furthermore, Description-Services are based on
the context of a service (Service-Context), which is
defined by a set of ontologies, related to the con-
cerned service and used for the annotation process.
This Service-Context is also called local ontology,
which means that there are several ontology levels.
For our purpose, three ontology levels have been
identified: local level, domain level and global level.
In essence, local ontologies concern services,
whereas domain ontologies concern the generaliza-
tion of local ones that belong to the same domain
(Production, Metrology, Packaging, etc.) and they
can serve in aligning the involved local ontologies.
At last, global ontology is considered as generaliza-
tion of domain ontologies, it is the root of the ontol-
ogy hierarchy, and they can serve both in aligning
domain ontologies and also in B2B integration that
constitutes a natural prospect of our present work.
Our ontology architecture is somewhat an exten-
ICEIS 2005 - DATABASES AND INFORMATION SYSTEMS INTEGRATION
168
sion of the hybrid ontology approach mentioned in
the case of information integration in (Wache,
2001). This extension is motivated by the fact that
none of the approaches proposed by (Wache, 2001)
(single ontology, multi-independent-ontologies and
hybrid-ontology approach) are appropriate to fully
capture and correctly structure semantics in our case.
This ontology clustering, which is firstly used in
a general fashion in (Visser, 1999), is a very impor-
tant concept in order to master the ontology evolu-
tion. We call this structuring Ontology Urbanization
and it is related to the concepts of city-planning
(zones, areas and islands) applied to EASs. It takes
an important role in our integration approach and it
will be more developed in future work.
4.2.2 Mediation Services
Mediation-Services are generally invoked by Bro-
kering-Services (technical services that aim to pro-
vide global mechanism to integration process) in
order to perform matching or resolution of semantic
heterogeneity between Fundamental-Services. They
exploit the description provided by the Description-
Services described above.
Since we use an hybrid ontology approach, this
requires the integration (mediation) of ontologies
which are performed by Ontology-Mediation-
Services (OMS) and that are based on ontology
mapping (Kalfoglou, 2003). This latter is the process
whereby two or more ontologies are semantically
related at conceptual level. According to the seman-
tic relations defined in the mappings (e.g. equiva-
lence, subsumption), source ontology instances can
then be transformed (or matched with) into target
ones (Noy, 2004).
In addition to OMS and according to the above
different fundamental types of services, we can
mainly distinguish three other types of mediation
services: Data Mediation Service (DMS), Functional
Mediation Service (FMS), Business Mediation Ser-
vice (BMS). These mediation services aim to medi-
ate respectively between DSs, FSs, BSs and they are
based on OMS that match and mediate between dif-
ferent ontologies. To be performed, Mediation-
Services can exploit two particular utility services
that are Inference-Service and Matching-Service.
These particular services can be respectively
supported by academic or commercial inference
engine and matching tool. For the initial prototype
that is ongoing, we decide to use Racer engine
(Racer, 2004) and OLA (OWL Lite Alignment)
matcher (OLA, 2004) that seems be appropriate to
our approach.
4.3 The initial Prototype
The initial prototype (also called ODSODI - Ontol-
ogy-Driven Service-Oriented Data Integration),
which is ongoing, aims to provide a first implemen-
tation of some functionalities of our architecture. We
have restricted this first prototype to data integra-
tion. Further versions of the prototype will address
application and process integration.
The underlying architecture of this first proto-
type is based around a fusion of WS concepts with
the concepts of data mediation, especially the media-
tors concepts like those defined by (Wiederhold,
1992).
Figure 3: Principles of the initial prototype (ODSODI).
Figure 3 illustrates the principle of the ODSODI
prototype. As shown, this prototype implements a
local-centric approach (aka local-as-view approach)
(Calvanese, 2001). In this approach, the query is
done over the global ontology and the Mediation-
Services access the data sets by a series of mappings:
from global to domain (which are done by global
mediation service), and then from domain to local
(which are done by local mediation services). This
choice is appropriate in the context of EAI in general
and in the context of our microelectronics society in
particular. It is motivated by the fact that users and
EASs are autonomous and have a limited knowledge
about DSs.
5 CONCLUSION
The semantic integration of enterprise application
systems is a hard problem that can concern data,
applications and processes. This problem needs on-
tology-based semantic mediation and, in our opin-
ion, is best resolved in the context of service-
oriented architectures.
A UNIFIED FRAMEWORK FOR APPLICATION INTEGRATION - An Ontology-Driven Service-Oriented Approach
169
This paper has proposed a unified approach for
Enterprise Application Integration that exploits both
ontology mediation and Web services. This ap-
proach called ODSOI (Ontology-Driven Service-
Oriented Integration) aims to extend the current web
services stack technology by a semantic layer offer-
ing some specific services that can mainly define the
service semantics and also perform semantic media-
tion in the context of EAI. Typologies of services
and also of ontologies have been suggested, and the
initial prototype is described. This latter is of course
limited, and its extensions, which may increase the
field of use and the usefulness of our approach, will
no doubt constitute important prospects in the future.
REFERENCES
Abiteboul S., Benjelloum O., Milo T., 2002. Web Services
and Data Integration. INRIA report.
Amjad U., 2004. The Emerging Role of the Web for En-
terprise, Proceedings of the IEEE, 92(9), pp. 1420-
1438.
BIJOnline, 2004. www.bijonline.com.
Bussler C., 2003. The Role of Semantic Web Technology
in EAI. In the Bulletin of the IEEE Computer Society
Technical Committee on Data Engineering.
Bussler C., 2004. Semantic Web Services. ICWE'04.
Munich.
Calvanese D., De Giacomo G., Lenzerini M., 2001. On-
tology of integration and integration of ontologies. In
Proceedings of the 2001 International Description
Logics Workshop (DL2001), pp. 10-19.
Cordoso J., Bussler C., Sheth A., Fensel D., 2002. Seman-
tic Web Services and Processes. Tutorial at Federated
Conferences.
Dogac A., 2004. Semantic Web Services, ICWE'04, Mu-
nich.
EBPML, 2004. www.ebpml.Org.
Erasala N., Yen D. C., Rajkumar T. M., 2003. Enterprise
Application Integration in the electronic commerce
world. Computer Standards & Interfaces 25, pp. 69-
82. Elsevier.
Fensel D., Bussler C., 2002. The Web Service Modeling
Framework WSMF. In Electronic Commerce Re-
search and Applications, 1(2). Elsevier.
Goh C. H., 1997. Representing and Reasoning about Se-
mantic Conflicts in Heterogeneous Information
Sources. Phd Thesis, MIT. USA.
Gruber R. T., 1993. Towards Principles for the Design of
Ontologies Used for Knowledge Sharing. KSL, Stand-
ford University.
Izza S. et al., 2004. Intégration d'applications: Etat de l'art
et Perspectives. JIE 2' 04 - Algérie.
Kadima H., Monfort V., 2003. Les Web Services. Dunod.
Kalfoglou Y., Schorlemmer M., 2003. Ontology Mapping.
The Knowledge Engineering Review, Vol. 18(1), pp. 1-
31. Cambridge University Press.
Kontogiannis K., Smith D., O'Brien L., 2002. On the Role
of Services in EAI. In the Proceedings of the 10th In-
ternational Workshop on Software Technology and
Engineering.
Linthicum D. S., 2004. Next Generation Application Inte-
gration. Addison-Wesley.
Lithicum D. S., 1999. Enterprise Application Integration,
Addison-Wesley.
McIlraith S., Son T., Zeng H., 2001. Semantic Web Ser-
vices, IEEE Intelligent Systems, pp. 46-53.
Mena E., Kashyap V., Sheth A., Illarramendi A., 1996.
OBSERVER.
In Proceedings 1st IFCIS International
Conference on Cooperative Information Systems,
Brussels.
METEOR-S, 2004. http://lsdis.cs.uga.edu/Projects/
METEOR-S.
Noy N. F., 2004. Semantic Integration: A Survey Of On-
tology- Based Approaches. Standford Medical Infor-
matics. Standford.
OLA, 2004. www.iro.umontreal.ca/~owlola/alignment.
html.
OWL-S, 2004. www.daml.org/services/owl-s/1.0/.
Racer, 2004. http://www.racer-systems.info.
Stuckenschmidt H., Wache H., Vogele T., Visser U, 2000.
Enabling technologies for interoperability. Workshop
on the 14th ISCSEP, pp. 35-46. Germany.
SWSI. 2004. Semantic Web Services Initiative,
www.swsi.org.
Turner M., Zhu F., Kotsiopoulos I., Russel M., Bennett K.,
Brereton P., Keane J., Rigby M., 2004. Using Web
Service Technologies to Create an Information Broker.
26th International Conference on Software Engineer-
ing (ICSE'04). IEEE Computer Society. pp 552-561.
Visser P., Tamma V., 1999. An experience with ontology
clustering for information integration. In Proceedings
of the IJCAI-99 Workshop on Intelligent Information
Integration, Sweden.
W3C, 2004. www.w3.org.
Wache H. et al.., 2001. Ontology-Based Integration of
Information. In Proceedings of IJCAI-01 Workshop:
Ontologies and Information Sharing (2001) pp.108-
117.
Wiederhold G., 1992. Mediators in the Architecture of
Future Information Systems, IEEE Computer Society,
25(3), pp. 38-49.
Woelk D., Tomlinson C., 1994. The InfoSleuth project. In
Second World Wide Web Conference ’94: Mosaic and
the Web.
ICEIS 2005 - DATABASES AND INFORMATION SYSTEMS INTEGRATION
170