Enabling Sustainable Interoperability for Enterprise Applications with

Knowledge Links

Artur Felic

1,2

, Felix Herrmann

, Christian Hogrefe

, Michael Klein

and Birgitta K¨onig-Ries

CAS Software AG, Karlsruhe, Germany

Friedrich-Schiller-Universit¨at Jena, Institut f¨ur Informatik, Jena, Germany

Keywords:

Semantic Web, Knowledge Links, Semantic Interoperability, Sustainable Interoperability, Enterprise Interop-

erability.

Abstract:

In complex and collaborative ecosystems like business networks, different partners need to share their re-

spective expert knowledge in order to be successful. Due to the diversity of business applications and highly

customized application suites used by business partners, knowledge exchange and the establishment of inter-

operability between enterprise applications is extremely difﬁcult. Different representations and meanings of

enterprise data lead to incomprehensibility between partners inside collaborative environments. This paper

presents an model-driven approach to support sustainable interoperability for enterprise applications in col-

laborative environments. Based on an event-driven architecture, Knowledge Links enable dynamic modelling

of knowledge transformations between knowledge domains. They keep background consistency between the

connected domains, thus making enterprise applications interoperable. Knowledge Links can be created and

modiﬁed at any time, which enables sustainable interoperability. Business partners are able to rely on their

enterprise applications and don’t need to switch to another system. Sensitive data stays covert due to the nature

of modular ontologies. The presented approach is exempliﬁed in the context of the OSMOSE Project and will

be evaluated by the proof-of-concept scenarios in this Project.

1 INTRODUCTION

Enterprise Applications can be described as highly

customized applications or application suites that

serve individual enterprise needs. They often com-

prise a set of expert tools for specialist capabilities.

According to Fowler (Fowler, 2003) ”Enterprise

applications are about the display, manipulation, and

storage of large amounts of often complex data and

the support or automation of business processes with

that data”. Enterprises usually compose their set of

applications by applications from different competing

vendors, which do not want to reveal sensitive data

in order to be integrated. Since the paradigm shift

from individual businesses to Complex and collabo-

rative ecosystems like business networks with many

business partners, the situation became even more

complex. Competing and globally spread enterprises,

having many different systems installed, have to

collaborate in order to create business value. This

circumstance leads to a problem from which a whole

branch of science called Enterprise Interoperability

emerged.

According to the IEEE Standard Computer Dic-

tionary (IEEE, 1991) Interoperability can be deﬁned

as the ”ability of systems or components to exchange

and use information”. Complementary, the ISO

1610 standard deﬁnes interoperability as the ”ability

to share and exchange information using common

syntax and semantics to meet an application-speciﬁc

functional relationship through the use of a common

interface”. Enterprise Interoperability concerns data,

services, processes and internal business. Thus,

different data models, various services and processes

as well as different levels of organization are aimed to

be interoperable. Many emerging Enterprise Interop-

erability Frameworks like EIF (IDABC et al., 2004),

FEI (Chen and Daclin, 2006), C4IF (Peristeras and

Tarabanis, 2006) and AIF (ATHENA Project, 2007)

aim at enabling Enterprise Interoperability. Although

the implementation of Enterprise Interoperability in

a collaborative ecosystem is desirable, the dynamics

inside the ecosystem also need to be taken into

account to be successful. The implementation is not

a one-off activity, but a continuous process that needs

to be sustainable in order to allow the entrance of

new business partners or changes inside the existing

597

Felic A., Herrmann F., Hogrefe C., Klein M. and König-Ries B..

Enabling Sustainable Interoperability for Enterprise Applications with Knowledge Links.

DOI: 10.5220/0005369605970607

In Proceedings of the 3rd International Conference on Model-Driven Engineering and Software Development (MDE4SI-2015), pages 597-607

ISBN: 978-989-758-083-3

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

ecosystem.

Sustainability and interoperability are two inher-

ently linked and inseparable aspects that need to be

addressed together (Dassisti et al., 2013). Sustainable

Interoperability can be described as ”novel strate-

gies, methods and tools to maintain and sustain the

interoperability of enterprise systems in networked

environments as they evolve with their environments”

(Jardim-Goncalves et al., 2012). Detection of new

enterprise systems, learning capabilities, adaptability,

analysis of the internal behaviour as well as com-

munication inside the network is addressed by this

research area. Interoperable collaborative ecosystems

with many different business partners running many

different enterprise applications must be seamlessly

conﬁgurable.

The Object Management Group (OMG)

(Object Management Group, 2014b) proposed

Model-Driven Architecture (MDA) (Object Man-

agement Group, 2014a) as a strategy towards

interoperability of heterogeneous systems (Singh and

Sood, 2009). The separation of computation indepen-

dent models (CIMs), platform independent models

(PIMs) and platform speciﬁc models (PSMs) enables

interoperability between enterprise applications by

allowing transformations between CIMs, PIMs and

PSMs. Long term ﬂexibility, incorporating models

from different viewpoints, technology independence,

multi platform model support and easier integration

and interoperability are some of the major beneﬁts of

MDA. Cost for the development life cycle is reduced

whereas software quality can be improved.

MDA can be extended by Semantic Web tech-

nologies, i.e. ontologies, to create unambiguous

domain vocabularies (W3C, 2006). Gruber deﬁnes

ontologies as ”an explicit speciﬁcation of a concep-

tualization” (Gruber, 1993). Knowledge sharing is

fundamental in collaborative ecosystems. Ontologies

can be used to structure knowledge and knowledge

properties in order to make knowledge understand-

able for machines and humans. There are many

ontologies deﬁned for different purposes to create

shared knowledge structures and to set up a common

knowledge base.

In highly dynamic collaborative ecosystems,

experience has shown that it is unrealistic to assume

commitment from every current and future business

partner to an integrated ontology covering all relevant

aspects of all business applications. A new business

partner entering the collaborative ecosystem or the

substitution of an enterprise application could lead

to the necessity to change the whole knowledge

base. Upper ontologies like SUMO (Niles and Pease,

2001) are purposed to structure only very general

concepts in order to support semantic interoperability.

Following the idea of modular ontologies (Ben Abbs

et al., 2012) large monolithic ontologies can be

divided into smaller domain modules in terms of

interchangeability. As the name suggests, domain

ontologies are often not comprehensible outside

the domain, but the knowledge structured inside is

of importance for other domains. Therefore, links

between those ontology modules are necessary in

order to provide different perspectives of knowledge

to the business partners that require it.

In this paper, we propose a model-driven ap-

proach using Knowledge Links (Felic et al., 2014).

Knowledge Links are used to connect the knowledge

domains and transform knowledge from one to

another. Their speciﬁcation constitute a Platform

Independent Model (PIM). Depending on the un-

derlying Platform Model (PM), i.e. the platform,

enterprise applications or database of business part-

ners involved in the Knowledge Link, the modelled

Knowledge Link is further transformed into events

and business processes by model transformations into

a Platform Speciﬁc Model (PSM).

By connecting different knowledge domains in a

black box manner, business partners of a collabora-

tive environment can rely on their knowledge struc-

tures. We present an event-driven and service ori-

ented architecture approach in which each enterprise

application constitute a separated knowledge domain

with its own ontology. Furthermore, a tool is pre-

sented with which knowledge links can be created

and maintained at any time during system runtime and

thus enable sustainable interoperability. During sys-

tem runtime, the complex event processor of the mid-

dleware is aware of keeping background consistency

by analysing events that are associated with Knowl-

edge Links and reacting according to the modelled

behaviour.

In the OSMOSE Project (OSMOSE, 2014), the

aim is to interconnect the real, digital and virtual

world for sensing liquid enterprises. In the real

world physical devices, objects and actors are present,

whereas their virtual models and hypothetical scenar-

ios are contained in the virtual world. The digital

world includes data sets about objects and their mod-

els as well as metadata and multimedia content. The

sensing liquid enterprise is composed of two enter-

prise paradigms: the sensing enterprise and the liq-

uid enterprise. The sensing enterprise is a cooperative

nervous system where virtual and physical smart ob-

jects, equipment and infrastructure form an active net-

work. In the liquid enterprise, boundaries are fuzzy

and it is hard to distinguish the inside from outside

in terms of human resources, markets, products and

MODELSWARD2015-3rdInternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment

598

processes due to the prevailing cloudiness. An event-

drivenand service-oriented middleware automatically

keeps background consistency between these worlds,

i.e. the shadow ambassadors of things in each of the

world. Each world has there own applications and

data structures, while the middleware has to be aware

of the meaning of data in order to decide which data

has to cross world borders to transform and deliver

data to the other worlds. The situation is quite similar

to business networks and enterprise applications. The

three worlds together with the middlewarecan be seen

as an collaborative environment, whereas the worlds

themselves are business partners with their installed

enterprise applications.

After this introduction, in Section 2, we will ex-

plore related work to this paper. In Section 3, the

modular and hierarchical knowledge base concept is

introduced. The proposed Event-Driven Architecture

is described in Section 4. The two latter sections

constitute the foundation of Section 5, in which the

sustainable interoperability approach is illustrated. In

Section 6 the paper is concluded and summarized and

future research directions are speciﬁed.

2 RELATED WORK

Enterprise Interoperability has received growing at-

tention in the last decade. (Lampathaki et al., 2012)

identiﬁed the lack of scientiﬁc foundations and for-

mulated several scientiﬁc areas of enterprise interop-

erability based on technological trends and enterprise

interoperability background knowledge. The main

scientiﬁc areas the present paper addresses are data,

software, knowledge and ecosystems interoperability.

Many sustainable interoperability approaches in

literature arise from the scientiﬁc ﬁelds of Semantic

Web, Model-Driven Architecture (MDA) or Event-

Driven Architecture (EDA). While one of the main

objectives of semantic web technologies is to induce a

common meaning about knowledge for human beings

and machines, MDA and EDA are concerned with

platform independence and heterogeneity of systems

besides other things. The following scientiﬁc ﬁndings

and publications give an overview about related work

in these scientiﬁc areas.

(Chen et al., 2008) analysed enterprise interoper-

ability architectures from 1985 to 2000 and states that

most effort has been put in framework approaches

rather architectures of enterprise systems. Further-

more, the authors identiﬁed the lack of an enter-

prise architecture ontology, the absence of scientiﬁc

methodologies and interoperability between existing

architectures as well as weak impact in industry. The

authors identiﬁed conceptual barriers as one of the

main category of interoperability barriers that are

”concerned with the syntactic and semantic differ-

ences of information to be exchanged”, which highly

motivates our work. Moreover, they describe the in-

teroperability of data as a category of interoperability

concerns that ”deals with ﬁnding and sharing infor-

mation from heterogeneous data sources”. According

to the categorization of interoperability approaches,

we present a federated approach, where partners do

not impose own models on the enterprise architecture.

In our approach, we allow that each partner can rely

on its own knowledge structured in an ontology, al-

though partners may share knowledge in a common

ontology. Semantic mapping is handled with Knowl-

edge Links by interconnecting the ontologies appro-

priately.

The authors of (Chungoora et al., 2013) utilize

a model driven architecture approach and ontologies

to enable knowledge sharing for manufacturing sys-

tems interoperability. The Manufacturing Core On-

tology constitutes the heart of the Platform Indepen-

dent Model (PIM), whereas other domains extend the

Manufacturing Core Ontology to specify domain spe-

ciﬁc models. To enable knowledge sharing of the

further transformed Platform Speciﬁc Model (PSM),

knowledge veriﬁcation mechanisms are speciﬁed at

PIM level. However, the approach is limited to the

speciﬁcation of constraints on PIM level and does

not allow appropriate knowledge propagation with

transformations in order to provide partners with new

knowledge build from their knowledge base.

Kadar et Al. (Kadar et al., 2013) propose a multi-

agent system to support sustainable interoperability

for networked organizations. Rule based negotiation

at business level by agents in a distributed environ-

ment enable the re-establishment of interoperability

between partners. Although the work of Kadar et Al.

is inspiring and knowledge negotiation is demand-

ing, the need to transform knowledge between differ-

ent knowledge domains and the creation of complex

knowledge relationships among different partners in

terms of sustainable interoperability is not covered.

Knowledge Links could therefore complement this

work.

(Ni and Fan, 2008) present an ontology based

approach. Domain ontologies are connected by

special collaboration ontologies, which are semi-

automatically created. These collaboration ontolo-

gies are described by OWL-S (W3C, 2004) in order

to ”enables automatic service discovery, invocation,

composition, interoperation, and execution monitor-

ing”. The OWL-S descriptions from these collab-

oration ontologies are further transformed to BPEL

EnablingSustainableInteroperabilityforEnterpriseApplicationswithKnowledgeLinks

599

and WSDL to allow automated execution of ontol-

ogy mappings. Thus, an business process execution

engine is able to react appropriately to requests ac-

cording to data changes. Although the approach is

very closely related to the knowledge link approach,

it is strongly dependent on the mechanisms behind the

semi-automatic creation process of the collaboration

ontologies. Additionally, complex links with opera-

tions and transformations cannot be created by hand.

The authors of (Agostinho et al., 2011) propose

to enhance information systems by traceability func-

tionalities to enable sustainable interoperability. Us-

ing model morphisms that allow altering and non-

altering model mapping, the relationship of models

from different knowledge domains can be formalized.

These mappings are translated in executable code in

order to exchange mapped or transformed data be-

tween business partners. The work of Agostinho et

Al. is closely related to the Knowledge Links due to

the support of morphisms between different knowl-

edge domains. Mappings are created on ontology

level, therefore Knowledge Links could complement

this work by allowing end users to model the exe-

cutable connections with given tools for end users.

By using Knowledge Links, the presented approach

could be enhanced by reusability of mappings. The

result of a mapping could further be used to deﬁne

new mappings.

The approach presented in this paper combines

ontology-driven, event-driven and model-driven ap-

proaches and is based on three pillars: A modular and

hierarchical knowledge base, an event-driven archi-

tecture for collaborative environments and a model-

driven knowledge link approach with a business pro-

cess execution engine. In the following sections, we

will take a look at these in turn.

3 KNOWLEDGE BASE

STRUCTURE

The knowledge base structure is following the work

of (Ben Abbs et al., 2012) about modular ontolo-

gies. Major advantages of the modularization ap-

proach that are of particular importance are knowl-

edge reuse across application domains, distributed

ontology engineering over different knowledge do-

mains and effective management of ontology mod-

ules. Business partners in collaborative environments

often have different applications running in their sys-

tem. Business networks in which business partners

have to comply to a centralized knowledge base and

speciﬁc enterprise applications in order to collabo-

rate are hard to establish and nearly impossible to

maintain. Furthermore, they build barriers and ob-

stacles for new business partners, which is fatal for

highly dynamic environments. Therefore, modular

ontologiesallowbusinesspartners to rely on their own

knowledge structures and maintain their own knowl-

edge base while connecting their knowledge structure

to a shared common denominator. Additionally, the

modular approach allows business partners to struc-

ture their inner knowledge base similarly.

The knowledge base structures can be categorized

in different hierarchy levels. The highest level is con-

stituted by a common ontology structuring necessary

knowledge about the business network. At the second

hierarchy level the business partners’ ontologies are

located. Each business partner structures its knowl-

edge separately. The third and lowest hierarchy level

comprise the enterprise applications’ ontologies. Fig-

ure 1 illustrates the hierarchy levels.

BP BP

Level:

Business

network

Level:

Enterprise

Application

Level:

Business

partner

Figure 1: Knowledge Base Hierarchy.

Ontologies of lower levels inherit the ontologies

of the next higher level and enrich the higher level

concepts with new speciﬁc concepts or add speciﬁc

concepts underneath higher level concepts. Accord-

ing to (Ben Abbs et al., 2012), this pattern of imports

establishes an unidirectional correspondence between

the lower level ontologies and the higher level ontol-

ogy. More precisely, the highest level ontology con-

stitutes an upper ontology for the underlying ontolo-

gies. According to (Healy et al., 2010), an upper on-

tology is a high-level, domain independent ontology

whose concepts generic and basic and ”from which

more speciﬁc ontologies can be derived”. Mid-level

ontologies between the lowest and highest level of the

ontology hierarchy ease the mapping between these

levels, whereas the domain ontologies on the lowest

level reuse the concepts from the levels above and ex-

tend them with domain speciﬁc concepts.

Upper ontologies are helpful in terms of interoper-

ability. They are often used to describe generic, plat-

form independent concepts in order to generalise the

MODELSWARD2015-3rdInternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment

600

meaning of domain speciﬁc concepts at a higher level.

There are many initiatives that aim to deﬁne standard

upper ontologies, for instance, the Suggested Upper

Merged Ontology (Niles and Pease, 2001). Domain

ontologies that are compliant with the upper ontolo-

gies are able to interoperate by shared terms and deﬁ-

nitions and mapping to these terms.

In order to describe resources like docu-

ments, websites or parts of them in a computer-

understandable manner, resources of enterprise appli-

cations need to be semantically annotated. In (Oren

et al., 2006), annotations of various domains are anal-

ysed and a uniﬁed formal model for semantic annota-

tions is presented. Using such an annotation model al-

lows enterprise application resources to be structured

and processable by our presented event driven knowl-

edge link approach.

In our approach, we use Apache Jena (Foundation,

2014) as storage and interface for ontologies. Jena is

a semantic web framework consisting of three mod-

ules. An OWL (W3C, 2012) and inference API al-

lows interaction with ontologies and reasoning over

data. The built-in triple store enables persistence and

allows querying over SPARQL (W3C, 2008) and http.

An RDF (W3C, 2014) API allows users to create, read

and manipulate triples and handles serialization.

The knowledge base structure of the OSMOSE

Platform follows the schema described above. Like

depicted in Figure 2, the OSMOSE Upper-Ontology

constitutes a common upper ontology for the whole

platform at the highest hierarchical level. It contains

generic concepts about events, entities, services and

processes as well as platform speciﬁc concepts like

’OSMOSE World’ or ’OSMOSE Process’. This com-

mon knowledge base is imported by each of the three

worlds allowing the worlds to extend these concepts.

The Intra-World Ontologies are ontologies speciﬁc

for each of the world, acting as a mid-level ontology.

Inner world applications use own ontologies by im-

porting the Intra-World Ontology and adding speciﬁc

concepts. Thus, they deﬁne the lowest level of on-

tologies, the domain ontologies. By semantic annota-

tions, the inner-world applications can map their data

to their ontologies.

To give a brief example, let us assume that the

OSMOSE Upper-Ontology deﬁnes the concept of a

generic entity. Inside the real world, the concept ’hu-

man actor’ extends the generic ’entity’ concept. For

a physical machine having its own ontology, ’human

actor’ is extended by concepts for administrators and

users.

Middleware

(Membranes)

Digital

World

Ontology

Real

World

Ontology

Virtual

World

Ontology

OSMOSE Upper-Ontology

OSMOSE Platform Model

Entity Ontology

Event Ontology

Process Ontology

Service Ontology

Common Ontology

Human

Actor

Entity

Figure 2: OSMOSE Knowledge Base Hierarchy.

4 EVENT-DRIVEN

ARCHITECTURE APPROACH

An Event-Driven Architecture is a style of Service-

Oriented Architecture, where the occurrence of an

event can trigger the invocation of services, which

in turn can generate new events. (Michelson, 2006)

Event-driven systems are designed to process events

as they occur, allowing the system to observe, react

dynamically to and issue personalized data depend-

ing on the recipient and situation.(Etzion and Niblett,

2010). Event-driven architectures are deployed in

many areas and have proofed as a competitive advan-

tage for a lot of organizations. Thus, they are becom-

ing increasingly popular. (Grabs and Lu, 2012)

Events can be seen as kind of messages that are

generated by resources in domains and recognised by

information systems. Events can occur and be con-

verted from anywhere in an environment (Stojanovic

et al., 2011), indicating hat an occurrence has been

observed in a system (Grabs and Lu, 2012). Luckham

deﬁnes events as objects that are records of activities

in a system and can have a relation to other events

with a timestamps as properties (Luckham, 2002).

According to McGovern, ”an event is a change in the

state of a resource or request for processing” (McGov-

ern et al., 2006).

Complex Event Processing (CEP) is a specialized

ﬁeld of event processing and part of event-driven ar-

chitectures dealing with complex events. Complex

events are combinations of multiple events consist-

ing of all the activities that the aggregated events

signify (Luckham, 2002) (Robins, 2010). Complex

Event Processing deals with the identiﬁcation and

handling of interdependent events across organiza-

tions, analysing their impact and taking subsequent

action in real time into account. It is an emerging

technology which allows to ﬁnd real time relation-

ships between different events using elements such

as timing, causality, and membership in a stream of

data in order to extract relevant information (Perro-

EnablingSustainableInteroperabilityforEnterpriseApplicationswithKnowledgeLinks

601

chon et al., 2001). Luckham states that ”Complex

Event Processing (CEP) is a deﬁned as a set of tools

and techniques for analysing and controlling the com-

plex series of interrelated events that drive modern

distributed information systems” (Luckham, 2008).

According to (Vernadat, 2007), loose coupling

among services and applications is the key to building

interoperable enterprise systems. Message-Oriented

Middleware (MOM) or their extension, the Enterprise

Service Bus (ESB), enable loose coupling using mes-

sage queuing techniques. They implement the Event-

Driven Architecture pattern and provide neutral mes-

sage formats with which applications are able to ex-

change messages independently and asynchronously

using simple transport protocols. Furthermore, it en-

ables communication between different business units

with heterogeneous platforms and environments.

Our approach utilizes the WSO2 Enterprise Ser-

vice Bus (WSO2, 2014) together with the RabbitMQ

Message Broker (Pivotal Software, 2014) in order to

take advantage of concepts explained before. The

WSO2 Enterprise Service Bus is characterized by

WSO2 as the highest performance, lowest footprint,

and most interoperable SOA and integrated middle-

ware today. High connectivity and compliance to

standards are additional key beneﬁts. Additionally,

it provides features for routing, mediation and trans-

formation of messages as well as monitoring, secu-

rity and service management features. The high-

performance and lightweight ESB is advertised as

highly available, scalable and stable. ESBs support

distributed environments. Each business partner in

a business network may have its own ESB running

or even multiple ESBs that are connected to a main

ESB of the business partner. The ESBs of each busi-

ness partner are in turn connected to the business

networks’ ESB. RabbitMQ is capable of propagating

messages sent between business partners. RabbitMQ

is described as robust, highly available, easy to use

and platform independent message broker. It supports

a variety of protocols and allows ﬂexible routing and

clustering. In order to utilize Complex Event Process-

ing, Esper (EsperTech, 2014) is used. Esper supports

management of Data Windows for ﬁne-grained event

expiry, event series analysis, and differentevent repre-

sentation types. Additionally, it is based on SQL stan-

dards, scalable and offers full API control. Addition-

ally, it provides its own Domain Speciﬁc Language

(DSL) called Event Processing Language (EPL). Ad-

ditionally, we’ve added access to the knowledge base

to combine the power of semantic web technologies

with complex event processing. Following the work

of (Schaaf et al., 2012), events and their relations to

each other can be enriched by semantics in order to

utilize knowledge inference about simple and com-

plex events as well as everything that is related to this

event, for instance the event publisher. Figure 3 illus-

trates the components of the middleware. Events that

are sent have to proceed the middleware, i.e. the event

channel. During this process, events and entities re-

lated to them are analysed by the Context Manager

and the complex event processor, i.e. the Esper CEP

Engine. The Context Manager encompass the Con-

tect Store and the Reasoner component, implemented

by Apache Jena. Additionally, the knowledge link en-

gine, which will be explained in more detail, handles

activation of modelled knowledge links.

Middleware

Event Channel

Rabbit MQ

WSO2 Enterprise Service Bus

Context Manager

Esper

Event Stream Connectors

Esper CEP Container

Historical Data

Context Store

Reasoner

Knowledge Link Engine

Figure 3: Middleware components.

WSO2 ESB, RabbitMQ and Esper are used to-

gether in the OSMOSE Project in order to allow com-

munication and analysis about events between the

real, digital and virtual world. Autonomous compo-

nents deploy services in a WSO2 ESB. Thereby, they

can react to events, call services or negotiate with

other autonomous components. The communication

is handled by RabbitMQ and analysed by the Esper

CEP Engine.

5 KNOWLEDGE LINKS

In the previous two main sections, we deﬁned the

foundation of our approach for sustainable interop-

erability. We have deﬁned a knowledge base con-

cept with which structuring of knowledge from dif-

ferent knowledge domains of business partners is en-

abled. Thus, business partners can rely on their data

structures that are connected to upper ontologies in

terms of compliance to the business networks’ generic

structures. The event drivenarchitecture approach en-

ables business partners to connect their enterprise ap-

plications to an ESB in their environment, while those

environments of all business partners can be con-

nected together to a business network environment by

utilizing ESB and message broker technologies.

MODELSWARD2015-3rdInternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment

602

In this section, the application of knowledge links

is explained as a linkage of knowledge concepts be-

tween knowledge domains that runs on the event

driven architecture. Due to the fact that knowledge

links are used to link structured knowledge between

enterprise applications and business partners, they are

suitable to address interoperability concerns. Further-

more, they can be deﬁned and maintained during run-

time of the business network. Thus, interoperability

can be sustained whenever business partners enter the

environment or change enterprise applications.

Knowledge links are connections between knowl-

edge concepts of different knowledge domains in a

black box manner. Inside the black box, operators

and transformations are applied to convert knowledge

from one or more source domains appropriately to the

target domain. They constitute a morphism of how

knowledge in one domain can be constructed by us-

ing knowledge from other domains.

Knowledge Links are intended to be modelled by

end users or domain experts. Modelling should be as

easy as possible to reduce the effort to share knowl-

edge to a minimum. Therefore, we’vechosen a graph-

ical way. In order to model knowledge links graphi-

cally, we implemented the Knowledge Link Conﬁg-

urator. This graphical modelling tool allows users

to import knowledge structures from different knowl-

edge domains and create knowledge links between

them. Creating a new knowledge link or loading an

existing one opens a graphical editor that allows drag-

ging and droppingof imported knowledgeconcepts as

well as operators from a palette. By drawing connec-

tions between dropped knowledge concepts and op-

erators, knowledge links can be deﬁned. Currently,

we have deﬁned a set of numeric (Sum, Subtraction,

Average...), text (Concat...) and boolean (AND, OR,

XOR...) operators. After the user ﬁnished with mod-

elling, knowledge links can be stored and uploaded to

the server. The uploaded knowledge links are further

translated and pushed to the Knowledge Link Engine

in the Context Store of the middleware.

In Figure 3, modelling of knowledge links with

the Knowledge Link Conﬁgurator is illustrated. On

the top right corner of the screenshot, the different

knowledge base structures are imported. On the lower

right side the knowledgelink deﬁnitions can be found.

Creating a new knowledge link deﬁnition will cre-

ate a rule sheet on which the graphical modelling is

done (squared area in the middle). The operators can

be found on the left side. In the knowledge link on

the screenshot, the set point of a machine part is sub-

tracted from it’s measurement in order to calculate the

deviation of the machine part. The measurement and

set point is taken from the real world machine and

machine part, more precisely, from the real world on-

tology. The deviation is a digital world concept and

thus described in the digital world ontology. The de-

scription of the knowledge link can be interpreted as

follows:

1. First two nodes, ’measurement’ and ’set point’:

Retrieve ’measurement’ and ’set point’.

2. Subtraction node, ’SUB’:

Subtract the set point from the measurement.

3. Equals node, ’==’:

Store the result ...

4. Last node, ’deviation’:

...as ’deviation’ in the ontology

The translation of knowledge links is done in two

steps. First, the knowledge concepts of the source do-

main are identiﬁed that participate in the knowledge

link. For all of those knowledge concepts, a listener in

the CEP Engine is registered that listens to changes of

the data behind that knowledge concept. It is assumed

that for all changes to data that are relevant for the

knowledge base semantic annotations are used and a

change of data results in an event that is sent to the

middleware having the knowledge concept as subject.

Thus, when knowledge concepts of the source do-

main are recognized in events that represent changes

of the data behind, the knowledge link will be exe-

cuted in order to keep background consistency. After

registering the listeners, an event automata is created

that handles requests on middleware level. When re-

questing knowledge that is deﬁned by a knowledge

link, events for gathering knowledge from each do-

main and for calculation need to be executed.

For the latter case, we use business processes to

perform actions modelled with palette operators. Dur-

ing the second step of translation, the modelled be-

haviour is translated into process-ﬁles that are associ-

ated with the knowledge link. If a listener recognises

an event that is relevant for a knowledge link, the data

from knowledge bases that refers to the knowledge

link is requested by events and passed to the pro-

cess. The process is built from predeﬁned building

blocks that correspond to the palette operators. For

instance, the ’Subtraction’ palette operator has a pre-

deﬁned building block for business processes. When

used, the two values that should be subtracted are

loaded by event requests from the appropriate knowl-

edge bases and passed to the business process for cal-

culation. After the business process has ﬁnished, the

result is passed back to the event automata in order

to proceed with another business process or to push

the data in the appropriate knowledge base by using

events. The translated process ﬁle knows about the

execution order and the input parameters that are nec-

EnablingSustainableInteroperabilityforEnterpriseApplicationswithKnowledgeLinks

603

Figure 4: Knowledge Link Conﬁgurator.

essary. Because of the ﬂexibility of the predeﬁned

building blocks for processes, sequences of them can

be reused when similar parts of knowledge links are

modelled.

5.1 Model-Driven Architecture

Approach

Knowledge Links follow the Model-Driven Architec-

ture (MDA) (Object Management Group, 2014a) ap-

proach by separating business and application logic

from platform-speciﬁc software of dynamic business

networks. According to (Truyen, 2006), MDA distin-

guishes between three different MDA Models. The

Computation Independent Model (CIM) is also called

business model and describes system expectations,

i.e. the output of the system, independently from

implementation. It bridges the gap between domain

experts and information technologists. The Platform

Independent Model (PIM) abstracts out technical de-

tails by deﬁning appropriate services and exhibiting a

sufﬁcient degree of independence to enable mapping

to platforms. Contrary, the Platform Speciﬁc Model

(PSM) combines speciﬁcations of the PIM with plat-

form speciﬁc details.

From their demand to their speciﬁcation, mod-

elling and translation, knowledge links undergo sev-

eral model transformations between MDA Models.

Figure 5 illustrates the different MDA Models and

the transformations that are explained below. Vertical

arrows represent CIM to PIM or PIM to PSM trans-

formations respectively. Horizontal arrows represent

CIM to CIM or rather PIM to PIM transformations.

In some cases, one-to-one transformations may not

be suitable. Instead, many-to-one, one-to-many or

even many-to-many transformations will be applied.

This is illustrated by multiple consecutively arranged

boxes as input or output for the arrows in Figure 5.

5.1.1 Computation Independent Model (CIM)

Models of the CIM are business requirements that

are expressed in natural language. They describe the

knowledge demand that business partners have. Fur-

ther, they can be transformed or consolidated to other,

higher-level CIM models to express the knowledge

demand inside the business network. These descrip-

tions are computation independent.

For instance, a quality manager of a business part-

ner inside a business network would like to know

about the deviation of a measurement from a product

component, which is manufactured by another busi-

ness partner, from its set point. The access to the

knowledge base could be restricted or the value that

is demanded is not comprehensible. Therefore, the

quality manager could express his or her demand in

natural language. Other business partners could in

MODELSWARD2015-3rdInternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment

604

turn read the description and provide a solution for

the demand and further rewrite the description.

Up to now, there is no appropriate automated

methodology described to transform business require-

ments to PIM models. Thus, the transformation from

CIM to PIM has to be done manually, i.e. with the

Knowledge Link Conﬁgurator by knowledge link cre-

ation.

5.1.2 Platform Independent Model (PIM)

Knowledge links themselves are models of the PIM.

Their description is platform independent and can

be graphically modelled with the Knowledge Link

Conﬁgurator. Furthermore, they are translated into

other PIM models, the event automata and business

processes (PIM to PIM transformation), after storing

them on the platform. The transformation result is in-

dependent from enterprise applications dthat are im-

plemented by business partners. Business processes

and event descriptions are also platform independent

and could be used by various engines or middleware

components in different integrated systems.

By way of example, the quality manager in the

scenario described above uses the Knowledge Link

Conﬁgurator and creates a knowledge link between

the concept of ’deviation’ in the quality managers’

knowledge domain and the ’measurement’ and ’set

point’ concepts of a business partners’ knowledge do-

main. After uploading this knowledge link, the trans-

lation algorithm generates an event automata that lis-

tens to changes of ’measurement’ and ’set point’ val-

ues and triggers business processes that are generated

from building blocks if changes to these values hap-

pen or knowledge about the ’deviation’ is requested.

5.1.3 Platform Speciﬁc Model (PSM)

During knowledge link translation, the PIM models

are transformed to PSM models taking semantic an-

notations into account. Semantic annotations map the

knowledge concepts of ontologies to real data like

documents in different formats. Thus, knowledge

gathering depends on the underlying platform and en-

terprise application that business partners are using.

Additionally, code fragments like event listeners and

event triggers are generated in order to control the

knowledge and event ﬂow of the knowledge link at

the middleware level. Business processes are trans-

lated and stored into speciﬁc process ﬁles that can be

executed by the business process execution engine.

In case of the quality manager that created and up-

loaded the ’deviation’ knowledge link to the middle-

ware, the business processes of the knowledge link is

provided with input from the semantically annotated

ﬁles that are maintained by the other business part-

ners’ enterprise applications. the event listeners and

triggers are created according to the middleware tech-

nology that is implemented.

Business Partner

Requirements

Business Partner

Requirements

Business Partner

Requirements

Business Partner

Requirements

Business Partner

Requirements

Business Partner

Requirements

Business Partner

Requirements

Business Partner

Requirements

Business Partner

Requirements

Business Partner

Requirements

Business Partner

Requirements

Business

Network

Requirement

Knowledge Link

Event Automata

Business

Processes

Platform Specific

Events/Listeners

Process Files

Real Data

CIM

PIM

PSM

Figure 5: Model Transformations of Knowledge Links.

6 SUMMARY AND CONCLUSION

We combined semantic web technologies, event-

drivenand model-drivenarchitecture approaches with

knowledge links to propose a solution for sustainable

interoperability in collaborative environments. Our

approach is based on three pillars: (1) a hierarchi-

cal and modular knowledge base concept utilizing

upper and modular ontologies, (2) an event-driven

and message-message oriented middleware based on

enterprise service busses that can be connected and

(3) knowledge links, a model-driven approach with

which domain ontologies can be linked in order to

achieve enterprise interoperability. Sustainable inter-

operability is enabled due to the possibility to create

and maintain knowledge links during runtime allow-

ing to react to changes in collaborative environments

by end users that are affected by these changes as they

appear. Middleware technologies and a tool to create

knowledge links have been presented. Consistency

between knowledge bases of business partners is war-

ranted automatically in the background. Obstacles are

reduced due to the ability of business partners to de-

ﬁne their own knowledge structures.

We exempliﬁedour approach with use cases of the

OSMOSE Project and will further evaluate our so-

lution in this Project. The software components of

the architecture presented in this paper are currently

prototypical and will be evaluated after integration

with the proof-of-concept scenarios of the OSMOSE

Project.

EnablingSustainableInteroperabilityforEnterpriseApplicationswithKnowledgeLinks

605

ACKNOWLEDGEMENTS

The research leading to these results has re-

ceived funding from the European Union Seventh

Framework Programme (FP7/2007-2013) under grant

agreement n

◦

610905, the OSMOSE project.

REFERENCES

Agostinho, C., Sarraipa, J., Goncalves, D., and Jardim-

Goncalves, R. (2011). Tuple-based semantic and

structural mapping for a sustainable interoperability.

In Camarinha-Matos, L. M., editor, Technological In-

novation for Sustainability, volume 349 of IFIP Ad-

vances in Information and Communication Technol-

ogy, pages 45–56. Springer Berlin Heidelberg, Berlin,

Heidelberg.

ATHENA Project (2007). Da8.2: Guidelines and

best practices for applying the athena inter-

operability framework to support sme partici-

pation in digital ecosystems. http://www.asd-

ssg.org/html/ATHENA/Deliverables/Deliverables,

Accessed 17.11.2014.

Ben Abbs, S., Scheuermann, A., Meilender, T., and

D’Aquin, M. (2012). Characterizing modular ontolo-

gies. In 7th International Conference on Formal On-

tologies in Information Systems - FOIS 2012, pages

13–25, Graz, Austria.

Chen, D. and Daclin, N. (2006). Framework for enter-

prise interoperability. Interoperability for Enterprise

Software and Applications: Proceedings of the Work-

shops and the Doctorial Symposium of the Second

IFAC/IFIP I-ESA International Conference: EI2N,

WSI, IS-TSPQ 2006, pages 77–88.

Chen, D., Doumeingts, G., and Vernadat, F. (2008). Archi-

tectures for enterprise integration and interoperabil-

ity: Past, present and future. Computers in Industry,

59(7):647–659.

Chungoora, N., Young, R. I., Gunendran, G., Palmer, C.,

Usman, Z., Anjum, N. A., Cutting-Decelle, A.-F.,

Harding, J. A., and Case, K. (2013). A model-driven

ontology approach for manufacturing system interop-

erability and knowledge sharing. Computers in Indus-

try, 64(4):392–401.

Dassisti, M., Jardim-Goncalves, R., Molina, A., Noran, O.,

Panetto, H., and Zdravkovi´c, M. M. (2013). Sustain-

ability and interoperability: Two facets of the same

gold medal. In On the Move to Meaningful Inter-

net Systems: OTM 2013 Workshops, volume 8186 of

Lecture Notes in Computer Science, pages 250–261.

Springer Berlin Heidelberg, Berlin, Heidelberg.

EsperTech (2014). Esper - complex event processing.

http://esper.codehaus.org/, Accessed 30.11.2014.

Etzion, O. and Niblett, P. (2010). Event Processing in Ac-

tion. Manning Publications Co, Greenwich, CT, USA,

1st edition.

Felic, A., K¨onig-Ries, B., and Klein, M. (2014). Process-

oriented semantic knowledge management in product

lifecycle management. Procedia CIRP, 25(0):361 –

368. 8th International Conference on Digital Enter-

prise Technology - DET 2014 Disruptive Innovation

in Manufacturing Engineering towards the 4th Indus-

trial Revolution.

Foundation, A. S. (2014). Apache jena - home.

https://jena.apache.org/, Accessed 30.11.2014.

Fowler, M. (2003). Patterns of enterprise application

architecture. The Addison-Wesley signature series.

Addison-Wesley, Boston.

Grabs, T. and Lu, M. (2012). Measuring performance of

complex event processing systems. In Topics in Per-

formance Evaluation, Measurement and Characteri-

zation, pages 83–96. Springer.

Gruber, T. R. (1993). A translation approach to portable

ontology speciﬁcations. Knowledge Acquisition,

5(2):199–220.

Healy, M., Poli, R., and Kameas, A. (2010). Theory and

Applications of Ontology: Computer Applications.

Springer.

IDABC, Echipa, and D. G. Industry (2004). European inter-

operability framework for pan-european egovernment

services: Draft document as basis for eif 2.0. Euro-

pean Communities.

IEEE (1991). Ieee standard computer dictionary: A compi-

lation of ieee standard computer glossaries. IEEE Std

610, pages 1–217.

Jardim-Goncalves, R., Popplewell, K., and Grilo, A. (2012).

Sustainable interoperability: The future of internet

based industrial enterprises. Computers in Industry,

63(8):731–738.

Kadar, M., Muntean, M., Cretan, A., and Jardim-

Goncalves, R. (2013). A multi-agent based negoti-

ation system for re-establishing enterprise interoper-

ability in collaborative networked environments. In

2013 UKSim 15th International Conference on Com-

puter Modelling and Simulation (UKSim), pages 190–

195.

Lampathaki, F., Koussouris, S., Agostinho, C., Jardim-

Goncalves, R., Charalabidis, Y., and Psarras, J.

(2012). Infusing scientiﬁc foundations into enterprise

interoperability. Computers in Industry, 63(8):858–

866.

Luckham, D. (2002). The power of events, volume 204.

Addison-Wesley Reading.

Luckham, D. (2008). The power of events: An introduction

to complex event processing in distributed enterprise

systems. Springer.

McGovern, J., Sims, O., Jain, A., and Little, M. (2006).

Event-driven architecture. Enterprise Service Ori-

ented Architectures: Concepts, Challenges, Recom-

mendations, pages 317–355.

Michelson, B. M. (2006). Event-driven architecture

overview. Patricia Seybold Group, 2.

Ni, Y. and Fan, Y. (2008). Ontology based cross-domain

enterprises integration and interoperability. In IEEE

Congress on Services Part II (SERVICES-2), pages

133–140.

MODELSWARD2015-3rdInternationalConferenceonModel-DrivenEngineeringandSoftwareDevelopment

606

Niles, I. and Pease, A. (2001). Towards a standard upper

ontology. In Guarino, N., Smith, B., and Welty, C.,

editors, the international conference, pages 2–9.

Object Management Group (2014a). Model driven archi-

tecture (mda). http://www.omg.org/mda, Accessed

17.11.2014.

Object Management Group (2014b). Object management

group. http://www.omg.org/, Accessed 17.11.2014.

Oren, E., M¨oller, K. H., Scerri, S., Handschuh,

S., and Sintek, M. (2006). What are se-

mantic annotations? http://www.siegfried-

handschuh.net/pub/2006/whatissemannot2006.pdf,

Accessed 29.11.2014.

OSMOSE (2014). Osmose project - homepage.

http://www.osmose-project.eu/, Accessed 29.11.2014.

Peristeras, V. and Tarabanis, K. A. (2006). The connection,

communication, consolidation, collaboration interop-

erability framework (c4if) for information systems in-

teroperability. Ibis, 1(1):61–72.

Perrochon, L., Kasriel, S., and Luckham, D. C. (2001).

Managing event processing networks. Stanford Uni-

versity.

Pivotal Software, I. (2014). Rabbitmq - messaging that

just works. http://www.rabbitmq.com/, Accessed

30.11.2014.

Robins, D. (2010). Complex event processing. In Second

International Workshop on Education Technology and

Computer Science. Wuhan.

Schaaf, M., Grivas, S. G., Ackermann, D., Diekmann, A.,

Koschel, A., and Astrova, I. (2012). Semantic com-

plex event processing. Recent Researches in Applied

Information Science, pages 38–43.

Singh, Y. and Sood, M. (2009). Model driven architecture:

A perspective. In 2009 IEEE International Advance

Computing Conference (IACC 2009), pages 1644–

1652.

Stojanovic, N., Stojanovic, L., Anicic, D., Ma, J., Sen,

S., and St¨uhmer, R. (2011). Semantic complex

event reasoning—beyond complex event processing.

Springer.

Truyen, F. (2006). The fast guide to model driven architec-

ture. Cephas Consulting Corp.

Vernadat, F. B. (2007). Interoperable enterprise systems:

Principles, concepts, and methods. Annual Reviews in

Control, 31(1):137–145.

W3C (2004). Owl-s: Semantic markup for web ser-

vices. http://www.w3.org/Submission/OWL-S/, Ac-

cessed 13.12.2014.

W3C (2006). Ontology driven architectures

and potential uses of the semantic web

in systems and software engineering.

http://www.w3.org/2001/sw/BestPractices/SE/ODA/,

Accessed 17.11.2014.

W3C (2008). Sparql query language for rdf.

http://www.w3.org/TR/rdf-sparql-query/, Accessed

3.12.2014.

W3C (2012). Owl 2 web ontology lan-

guage document overview (second edition).

http://www.w3.org/TR/owl2-overview/, Accessed

3.12.2014.

W3C (2014). Rdf - semantic web standards.

http://www.w3.org/RDF/, Accessed 3.12.2014.

WSO2 (2014). Wso2 enterprise service bus.

http://wso2.com/products/enterprise-service-bus/,

Accessed 30.11.2014.

EnablingSustainableInteroperabilityforEnterpriseApplicationswithKnowledgeLinks

607