LEVERAGING THE PUBLISH-FIND-USE PARADIGM OF SOA

Supporting Enterprise Collaboration Across Organisational Boundaries

Brahmananda Sapkota and Marten van Sinderen

Faculty of Electrical Engineering, Mathematics and Computer Science

University of Twente, Drienerlolaan 5, Enschede, The Netherlands

b.sapkota@utwente.nl, m.j.vansinderen@utwente.nl

Keywords:

Service Oriented Architecture, Enterprise Application Integration, Cross-organisational Collaboration, Busi-

ness Requirements, Business Processes, Business-IT alignment.

Abstract:

Service Oriented Architecture (SOA) has been widely recognized as an approach for ﬂexible integration of

enterprise applications across organisational boundaries based on service abstractions and Internet standards.

Enterprise collaboration and application integration is driven by business requirements, which in turn are

translated to business models and expressed as business processes. For example, business processes can be

used to represent the coordination of several published services as well as the implementation of a composite

value-added service. In this way, enterprise functions are aggregated using multi-stage business processes

fulﬁlling the speciﬁc requirements of an enterprise. In order for enterprises to stay competitive in their re-

spective businesses, such solutions must evolve in a timely and appropriate way in response to changes in

market demands and opportunities that inevitably occur. Therefore, a mechanism is required for business-IT

alignment during the complete lifecycle of SOA-based enterprise collaboration and application integration. In

this paper, we discuss issues related to cross-organisational collaboration and how service-oriented principles

and architectures can be applied to address these issues.

1 INTRODUCTION

Service Oriented Architecture (Erl, 2005) has been

widely recognized as an approach for ﬂexible integra-

tion of enterprise applications across organizational

boundaries based on service abstractions and Internet

standards. On-demand enterprise collaboration and

application integration is driven by business require-

ments, which in turn are translated to business mod-

els and expressed as business processes. For example,

business processes can be used to represent the coor-

dination of several published services as well as the

implementation of a composite value-added service.

In this way, enterprise functions are aggregated us-

ing multi-stage business processes fulﬁlling the spe-

ciﬁc requirements of an enterprise (van Sinderen and

Almeida, 2011). In order for enterprises to stay com-

petitive in their respective businesses, such solutions

must evolve in a timely and appropriate way in re-

sponse to changes in market demands and opportu-

nities that inevitably occur. Service Oriented Archi-

tecture (SOA) supports adaptation of such an evolu-

tion through the concept of service composition. The

composition of services is made possible because of

service discovery which allows to ﬁnd and select the

suitable services. The service discovery thus plays a

central role in enterprise collaboration and application

integration.

One of the fundamental requirements for the ser-

vice discovery are service descriptions and service

registries. In general, service descriptions are used

for specifying: 1) what functionalities are offered by

the service to its users (i.e., the interface deﬁnition);

2) how the service is provided (i.e., the service bind-

ing); and 3) where the service can be accessed (i.e.,

the service endpoint information). The service reg-

istries are used to announce the offered services and

thus play crucial role for a successful on-demand en-

terprise collaboration and application integration. In

a typical scenario, service providers publish their ser-

vice descriptions to a publicly accessible service reg-

istry. The service users search over these registries

and ﬁnd the information needed to use the required

services. This approach of publishing, ﬁnding and

using the services forms the so called SOA trian-

gular operational model which clearly separates the

Sapkota B. and van Sinderen M.

LEVERAGING THE PUBLISH-FIND-USE PARADIGM OF SOASupporting Enterprise Collaboration Across Organisational Boundaries.

DOI: 10.5220/0004458300340041

In Proceedings of the First International Symposium on Business Modeling and Software Design (BMSD 2011), pages 34-41

ISBN: 978-989-8425-68-3

role of service provider, service user and service reg-

istry (van Sinderen, 2009). SOA, thus, enables ﬂexi-

ble on-demand enterprise collaboration and applica-

tion integration. Despite these sound principles of

SOA, a number of practical complexities still exists,

which are preventing enterprises to fully exploit the

potential beneﬁts of SOA.

An on-demand collaboration can only be achieved

if 1) the participants required for the collaboration

can be found and 2) the participants can communi-

cate with each other. In the SOA based approach,

the former requirement is supported by publishing

the descriptions of the offered services and latter re-

quirement is supported by deﬁning the messages be-

ing sent. This can be realised relatively easily by us-

ing pre-meditated message structures and function li-

braries, if the enterprises collaborate in closed envi-

ronment. If the autonomous enterprises are to col-

laborate, pre-meditated message structures or func-

tion libraries cannot be used. Therefore, semantics of

the information provided through the service descrip-

tions should be well deﬁned and the service reposi-

tory should contain valid service descriptions. While

the initiatives around Semantic Web services have

deﬁned formalisms such as WSMO (Roman et al.,

2006), OWL-S (Martin, 2004) and SAWSDL (Farrell

and Lausen, 2007), to semantically deﬁne service de-

scriptions, they cannot ensure that the published de-

scriptions correctly reﬂect the offered services at the

time of their discovery. Therefore, a mechanism is re-

quired for business-IT alignment during the complete

lifecycle of SOA-based enterprise collaboration and

application integration.

In this paper, we propose a Summary of Services

per Provider (SSP) description based approach for

service discovery and to ensure that the published de-

scriptions are valid. An SSP description provides a

means for describing the collection of services offered

by a single service provider. It speciﬁes what type of

services are offered by a particular service provider.

In the proposed approach, we use Web Service Deﬁni-

tion Language (Chinnici et al., 2007) as the language

for describing services and allow service providers

to store them in their local repository. We provide

a mechanism to generate SSP description based on

the Web Service Deﬁnition Language (WSDL) doc-

uments stored at the local service registry of the ser-

vice providers. These SSP descriptions are then pub-

lished to the public service registry. Though the ser-

vice provider still has to generate and publish these

descriptions, it can be automated. Through this sepa-

ration, we aim at reducing the registry updating bur-

den at the side of the service providers. The proposed

approach, therefore, follows the SOA triangular oper-

ational model except that SSP descriptions are pub-

lished to the service registry instead of publishing the

service descriptions.

The rest of the paper is structured as follows:

Some of the highly relevant existing works are dis-

cussed in Section 2. The service discovery challenges

are discussed in Section 3. The proposed solution is

presented in Section 4 and its use to support cross-

organisational collaboration is presented in Section 5.

Finally, the work presented in this paper is concluded

in Section 6 by highlighting possible future direc-

tions.

2 RELATED WORKS

The problem of guaranteeing correctness of the pub-

lished service descriptions and reducing the effort re-

quired for providing such guarantees is starting to at-

tract attention from the research communities. The

work presented in (K

uster and K

oning-Ries, 2007)

follows an approach similar to the one presented in

this paper. The focus of this work is to ensure that

the published service description indeed represents

the concrete service. In order to support this, an es-

timation step followed by a single execution step is

proposed. In the estimation step, additional informa-

tion than that is available in the service description

itself is gathered whereas in the execution step, the

actual invocation of service is performed.

A preference-based selection of highly conﬁg-

urable web services is presented in (Lamparter et al.,

2007). It focuses on deﬁning algorithms required for

ﬁnding optimal conﬁgurations while selecting the ser-

vices. Unlike the work presented in this paper, their

work neither considers minimisation of the extra ef-

fort required to update the service registry when ser-

vice descriptions are changed nor maximising the cor-

rectness of the published service descriptions. To

address the problem due to changes in service de-

scriptions a RSS-based mechanism to announce such

changes is proposed in (Treiber and Dustdar, 2007).

The RSS-based approach is service provider depen-

dent because the changing information should come

from them.

Treating services from the economical point of

view, (Cardoso et al., 2009) deﬁnes universal service

description language. The proposed language is de-

ﬁned to describe both the IT and non-IT services. In

the proposed language, provisions for deﬁning dy-

namic information is poorly deﬁned. The approach

presented in (Truong et al., 2010) deﬁnes mechanism

for identifying and reducing irrelevant information in

service composition and execution. This approach

LEVERAGING THE PUBLISH-FIND-USE PARADIGM OF SOA - Supporting Enterprise Collaboration Across

Organisational Boundaries

is target at increasing efﬁciency and correctness of

the composition and execution. This approach works

only after the services are discovered and does not

eliminate the possibility of discovering services with

incorrect information.

In contrast to many other traditional ap-

proaches, (Speiser and Harth, 2011) proposes a

LinkedData based approach for integrating data

providing services. Their approach is suitable for

sharing data which might change over time. In com-

parison to the work presented in this approach, their

approach requires the service providers to describe

the offered services using LinkedData principles and

does not support sharing of already existing service

descriptions which are described using WSDL. We

propose mechanisms to allow usage of WSDL while

still dealing with changing and state dependent data.

A crawl based approach for collecting, annotat-

ing and classifying public Web services has been pro-

posed in (AbuJarour et al., 2010) attempting to in-

crease the role of service registry in service-oriented

architecture by providing correct information. In their

approach, publicly available web service descriptions

are crawled, annotation information is gathered, Web

services are annotated and classiﬁed based on this in-

formation. Through such classiﬁcation, authors aim

at supporting better discovery of services. In this di-

rection, the work presented in (Obrst et al., 2010)

aims at enabling rich discovery of Web services by

projecting weak semantics from structural speciﬁca-

tions. These approaches, however, cannot guarantee

that the information required for service discovery is

gathered.

Combination of document classiﬁcation and on-

tology alignment schemes is proposed in (CRASSO

et al., 2010) to semantically enrich Web services.

Though this scheme helps in efﬁciently discovering

required services, it does not tackle the problem of

outdated service descriptions. The work presented in

(da Silva et al., 2011) speciﬁes mechanisms for run-

time discovery, selection and composition of seman-

tic services. The proposed approach supports seman-

tic descriptions of the services but lacks support for

dealing with outdated service descriptions.

3 SERVICE DISCOVERY

CHALLENGES

In an open environment, it is difﬁcult to support on-

demand collaboration if the published service de-

scriptions are either outdated or provide ambiguous or

incorrect information. This difﬁculty escalates when

service descriptions contain limited information, ei-

ther because the service providers are unwilling to

share all the information or because the information

is state dependent (Treiber and Dustdar, 2007; K

uster

and K

oning-Ries, 2007), i.e., the information may

change as the service is invoked. This will result in

a poor discovery results.

The correctness problem arises due to the fact

that service registries are passive. If the function-

alities of the offered services are changed, service

providers should take the initiative to update the cor-

responding descriptions published in the service reg-

istry. In practice, the published descriptions are rarely

updated. Instead of publishing the service descrip-

tions to the service registries, they are published on

the Web (Michlmayr et al., 2007). Such a practice,

violates the original SOA model (i.e., the triangular

operational model) and consequently undermines the

role of service registries in SOA (AbuJarour et al.,

2010). This shift in practice is due to the lack of efﬁ-

cient and elegant support to update the service registry

whenever a service description is updated. The latter

essentially requires an additional effort on part of the

service providers. It becomes more problematic be-

cause the existing service registries emerge and disap-

pear (Sabou and Pan, 2007) and cannot be fully relied

upon for service discovery. The results of the investi-

gation of Web services on the Web published in (Al-

Masri and Mahmoud, 2008) reveals that only around

63% of the discovered Web services are in fact ac-

tive. This lack of reliable service repositories makes

on-demand collaboration between autonomous enter-

prises difﬁcult, if not impossible.

Besides these technical difﬁculties, there are other

reasons why public service registries have not been

successful so far. One of the reasons is that a consid-

erable amount of the published service descriptions

are unusable (Treiber and Dustdar, 2007). Some of

the information (e.g., the information that depends

on the change in state) which might be important

for discovery purposes cannot be included in the ser-

vice descriptions in a simple way. In addition, some

providers may not be willing to disclose informa-

tion related to nun-functional properties and the qual-

ity of service parameters because of the fear of bar-

gain or competition from other providers (K

uster and

oning-Ries, 2007). This contributes to the retrieval

of imprecise service descriptions leading to false pos-

itives and consequently reducing the usability and the

reliability of service registries.

The above mentioned problems could be resolved

if service providers are allowed to store their service

description locally and provide them with a tool that

extracts summary information from their repository,

builds an SSP description and publishes it to the ser-

BMSD 2011 - First International Symposium on Business Modeling and Software Design

vice registry. The beneﬁts of this approach are: 1)

service providers do not have to publish all the infor-

mation. It will also reduce the extra effort required for

maintaining and updating the service registry. Service

providers do not need to update service registry every

time the service description is updated, updates are

done locally. 2) service users can use these SSP de-

scriptions to ﬁnd the potential providers, and ﬁnally

obtain the up-to-date information. Figure 1 shows the

overall architecture of this approach.

Service Providers

Registeries

Service Registry

Local Repository

Repository

Scanner

SSP Description

Generator

Publisher

Figure 1: Overall architecture.

4 SERVICE DISCOVERY

SOLUTIONS

We describe a mechanism to extract type information

from the WSDL ﬁles stored in the service provider’s

local repository and a mechanism to publish these in-

formation as SSP description to the service repository.

4.1 Generating Type Information

One of the purposes of using type information, in

the proposed approach, is to provide indication of

what type of services are offered by the service

providers. This kind of information serves the pur-

pose of guiding service requests towards the poten-

tial service providers. The type information based ap-

proaches are expected to help in narrowing down the

search space and allowing applications to deal with

the ever increasing number of Web services. We ex-

tract these information from service descriptions en-

coded in WSDL, which is a commonly used service

description language. A WSDL document is struc-

tured into four elements describing Service, Bindings,

Interface and Types deﬁnitions. A Service deﬁnition

speciﬁcation speciﬁes a collection of endpoints (i.e.,

URLs). The Bindings deﬁnition typically speciﬁes

what data formats and communication protocols to

use when invoking the service. The operations, mes-

sage exchange patterns and mechanisms for fault han-

dling are speciﬁed in the Interface deﬁnition. The

data types used in messages and faults are speciﬁed

in the Types deﬁnition. Figure 2 shows some of these

parts pictorially where as the Listing 1 shows the type

and service deﬁnition parts of a WSDL document.

Description

Interface

Types

Element Declaration

Type Definition

Operation

Fault

Binding

Service

Input

Output

Interface

Endpoint

Figure 2: Pictorial representation of WSDL structure.

The data types speciﬁed in the Types deﬁnition es-

sentially model the domain knowledge and thus are

useful for extracting the necessary information for

generating SSP description from a given WSDL doc-

ument.

<wsdl:description

targetNamespace=”http://org.example.com/services/AvailabilityService/”

xmlns=”http://org.example.com/services/AvailabilityService/”

xmlns:wsdl=”http://www.w3.org/ns/wsdl”

xmlns:xsd=”http://www.w3.org/2001/XMLSchema”>

<wsdl:types>

<xsd:schema targetNamespace=”http://org.example.com/resources/

AvailabilityService”>

<xsd:element name=”ServiceRequest”>

<xsd:complexType>

<xsd:sequence>

<xsd:element name=”product” type=”xsd:string”/>

<xsd:element name=”date” type=”xsd:string”/>

<xsd:element name=”quantity” type=”xsd:ﬂoat”/>

</xsd:sequence>

</xsd:complexType>

</xsd:element>

<xsd:element name=”ServiceResponse” type=”response”/> <xsd:element

name=”ServiceResponse” restriction=”xsd:boolean”/>

<xsd:simpleType name=”response”>

<xsd:restriction base=”xsd:boolean”/>

</xsd:simpleType>

</xsd:schema>

</wsdl:types>

<wsdl:interface name=”AvailabilityServiceInterface”>

<wsdl:operation name=”RequestOperation” pattern=”http://www.w3.org/ns/

wsdl/in−out”>

<wsdl:input element=”ServiceRequest”/>

<wsdl:output element=”ServiceResponse”/>

</wsdl:operation>

</wsdl:interface>

</wsdl:description>

Listing 1: Fragment of a WSDL document.

Given a WSDL document, we extract the Types

deﬁnitions and represent them as RDF (Klyne and

Carroll, 2004) triples. Representing those deﬁnitions

in RDF has several advantages. RDF provides an

abstract model for describing resources with proper-

ties, scoping them to a particular application domain

through RDF Schema (Hayes, 2004) and deﬁning re-

lationships between these resources. In addition, stan-

dard RDF query language SPARQL (Prud’hommeaux

and Seaborne, 2007) can be used to provide ﬂexi-

ble means to allow users to express their requests.

The RDF triples are generated based on the XML

LEVERAGING THE PUBLISH-FIND-USE PARADIGM OF SOA - Supporting Enterprise Collaboration Across

Organisational Boundaries

Schema to RDF Schema mappings approaches pre-

sented in (Thuy et al., 2008) and (CRASSO et al.,

2010). The resulting RDF triples are shown in List-

ing 2.

<rdfs:Class rdf:ID=”http://org.example.com/resources/AvailabilityService#

ServiceRequest”/>

<rdfs:Class rdf:ID=”http://org.example.com/resources/AvailabilityService#

ServiceResponse”/>

<rdf:Property rdf:about=”http://org.example.com/resources/AvailabilityService#

response”>

<rdfs:domain rdf:resource=”http://org.example.com/resources/

AvailabilityService#ServiceResponse”/>

<rdfs:range rdf:resource=”http://www.w3.org/2001/XMLSchema#boolean”/>

</rdf:Property>

<rdf:Property rdf:ID=”http://org.example.com/resources/AvailabilityService#

product”>

<rdfs:domain rdf:resource=”http://org.example.com/resources/

AvailabilityService#ServiceRequest”/>

<rdfs:range rdf:resource=”http://www.w3.org/2001/XMLSchema#string”/>

</rdf:Property>

<rdf:Property rdf:ID=”http://org.example.com/resources/AvailabilityService#date”

<rdfs:domain rdf:resource=”http://org.example.com/resources/

AvailabilityService#ServiceRequest”/>

<rdfs:range rdf:resource=”http://www.w3.org/2001/XMLSchema#string”/>

</rdf:Property>

<rdf:Property rdf:ID=”http://org.example.com/resources/AvailabilityService#

quantity”>

<rdfs:domain rdf:resource=”http://org.example.com/resources/

AvailabilityService#ServiceRequest”/>

<rdfs:range rdf:resource=”http://www.w3.org/2001/XMLSchema#int”/>

</rdf:Property>

Listing 2: RDF representation of extracted information.

In Listing 2, each triples are encoded in the form

<s, p, o> where s, p, and o are called the sub-

ject, predicate and the object respectively. In a triple,

a predicate is called the property of the triple and de-

notes the relationship between the subject and the ob-

ject that it connects. The subject always appears at the

left whereas the object appears at the right side of the

predicate in the triple.

4.2 SSP Description

An SSP description models the service provider as a

collection of services and enumerates all the offered

services. In particular, the SSP description speciﬁes

the type of services that are offered by a particular ser-

vice provider. We use the information extracted from

the locally stored WSDL ﬁles to describe domain spe-

ciﬁc concepts and to avoid ambiguities between ser-

vices from different application domains. Using the

SSP description, functional properties of the offered

services are described collectively. These descrip-

tions provide information sufﬁcient enough to ﬁlter

out the completely irrelevant service providers.

We deﬁne SSP description as <n, e, Q>, where n

is the URL of the service provider, e is the SPARQL

endpoint of the local repository and Q is the collec-

tion of (<p(s

, o

), f>) pairs, where s

and o

repre-

sents the type of the subject and object connected by

the predicate p whereas f represents the total num-

ber of occurrences of the predicate p together with

the s

and o

. The frequency of occurrences of sub-

ject and object type combination of each predicate is

measured to indicate the number of services from a

particular domain. We included the subject and ob-

ject types in the SSP description because they repre-

sent the domain and range of a predicate and hence is

useful to unambiguously select the required informa-

tion. The structure of the SSP description is deﬁned

as RDF graph as shown in Listing 3.

<?xml version=‘‘1.0”?>

<rdf:RDF xmlns:ex=‘‘http://www.example.org/ontology#”

xmlns:ont=‘‘http://www.example.org/ns/ms−ontology#”

xmlns:xsd=‘‘http://www.w3.org/2001/XMLSchema#”

xmlns:rdf=‘‘http://www.w3.org/1999/02/22−rdf−syntax−ns#”>

<ont:Service rdf:about=‘‘http://www.example.org/provider/xyz”>

<ont:endPoint rdf:resource=‘‘http://www.example.org/services/abc”/>

<ex:name>

<rdf:Description>

<ex:subjectType rdf:resource=‘‘http://www.example.org/ontology#Hotel”/>

<ex:objectType rdf:resource=‘‘http://www.w3.org/2001/XMLSchema#string

”/>

<ex:count>435</ex:count>

</rdf:Description>

</ex:name>

<ex:product>

<rdf:Description>

<ex:subjectType rdf:resource=‘‘http://www.example.org/ontology#

Automobile”/>

<ex:objectType rdf:resource=‘‘http://www.example.org/ontology#Wheel”/>

<ex:count>1481</ex:count>

</rdf:Description>

</ex:product>

</ont:Service>

</rdf:RDF>

Listing 3: An example SSP description.

The number of instances of objects and subjects in

a repository is typically far larger than the number of

distinct predicates. Inclusion of the objects in the SSP

description is therefore likely to increase its size as

close to as the size of the repository. In order to avoid

such problems, the SSP description includes only the

subject and object types and not their instances. This

allows to select the relevant service provider based

on this type information. For example, if there are

a large number of printing service providers and only

few of them are providing book printing services it is

much more efﬁcient to send a book printing service

requests only to those that provide the book printing

services. This type of SSP description will be helpful

in selecting the service providers in situations where

ﬁne grained information is needed for answering the

service requests.

4.3 Finding Potential Service Providers

In order to ﬁnd the required service we follow two

step discovery mechanism. In the ﬁrst step, the dis-

covery request is sent to the service registry from

which a list of potential service providers are iden-

tiﬁed. In the second step, direct communication with

the potential service providers is established to ﬁnd

the most appropriate service provider. The goal of

the ﬁrst step is to reduce the search space whereas the

BMSD 2011 - First International Symposium on Business Modeling and Software Design

second step is intended to select the service provider

based on up-to-date information.

The major design goals of the proposed approach

is simplicity and extensibility. The SSP description is

essentially the collection of summary of the WSDL

ﬁles stored in the local repository of the providers

which simpliﬁes the process of service description up-

dates. The service providers need to update only the

local repository. If the new service providers arise,

they can simply publish the SSP description to the ser-

vice registry.

We assume that the service providers employ the

mechanisms discussed above for extracting the infor-

mation needed for generating SSP descriptions. First,

the service requester initiates the lookup over SSP de-

scriptions and obtains a list of endpoints of the po-

tential partners. These endpoints are then queried to

obtain further information for selecting the potential

service providers.

4.4 Maintaining Freshness

The SSP descriptions are published to the service reg-

istry and hence can still pose the same problems as

with publishing the service descriptions if the do-

main information is changed. In order to ensure that

the SSP descriptions are still up-to-date, the service

repository requires a mechanism to reverse look up

the service providers local repository and synchro-

nize the information that is being provided through

the SSP descriptions.

5 ENTERPRISE

COLLABORATION

Let us now return to our original goal of facilitating

enterprise collaboration using SOA. We previously

concluded that the SOA architectural triangle with its

’publish-ﬁnd-use’ paradigm is in principle very con-

venient to enterprises to utilize distributed capabilities

that may be under the control of different ownership

domains. The convenience stems from the loose cou-

pling of services - supporting ﬂexible composition -

and the external-oriented representation of services -

allowing interactions between users and providers ir-

respective of their internal implementation. However,

we also concluded that the ’publish-ﬁnd’ part of the

triangle has practical limitations, which so far has pre-

vented the successful uptake of public-registry/open-

discovery based enterprise collaboration. The main

limitations are: (a) it is hard to ﬁnd and compose ser-

vices based on current service descriptions, since the

descriptions lack unambiguous and precise semantics;

(b) it is expensive and laborious to maintain service

descriptions, as the corresponding services continu-

ously evolve and therefore the descriptions require

frequent and manually managed updates; (c) trust is a

hindrance for publishing service descriptions and us-

ing services discovered with public registries.

We propose to leverage the publish-ﬁnd-use

paradigm by using public descriptions that are au-

tomatically generated and semantically enhanced, as

described in Section 4. In the following, we ﬁrst show

which interactions are necessary for enterprise collab-

oration using our approach, and subsequently discuss

the potential beneﬁts of our approach.

Service User

Public

Service

Registry

Service

Provider

Local

Service

Registry

Mediator

Find

Find Provider

Publish SSP Description

Store Service Description

Use

Find Service

Figure 3: Proposed service discovery approach.

Figure 3 illustrates the basic interactions:

1. A business organisation acting as a service

provider can publish relevant information on all

its services using a single keyword-based descrip-

tion (SSP) at a service broker that maintains a

public registry. If services offered by the business

organization evolve, the SSP is re-generated or in-

crementally updated, depending on the nature of

the change. The updated SSP can be pushed to

the service broker, in similar to the original pub-

lication, where it is used to replace the old SSP.

Alternatively, the service broker periodically asks

the service provider for updates.

2. A business organisation looking for partners that

can offer certain services can contact the service

broker and ﬁnd SSPs based on keyword match-

ing. Although keyword matching can already be

reasonable efﬁcient (Obrst et al., 2010), we can

further improve the recall and precision of service

discovery by exploiting the RDF-based semantics

of SSPs.

3. If an SSP fulﬁls the search criteria, the service

provider is contacted via the endpoint that is part

LEVERAGING THE PUBLISH-FIND-USE PARADIGM OF SOA - Supporting Enterprise Collaboration Across

Organisational Boundaries

of the SSP. Then the local repository of the service

provider is used to ﬁnd available services. Both

step (2) and step (3) may be performed via an in-

termediary, making the two-step service discov-

ery transparent to the requesting business organi-

zation. For example, the service broker may take

the intermediary role. Possibly, the requested ser-

vices are not available as single services or are not

available from a single service provider. In that

case, the services have to be offered as a bundle or

must be composed. Again, an intermediary may

automate or support this process (da Silva et al.,

2011).

4. Once the (composed) services are found, the re-

questing business organization can start using

them, effectively entering collaborations with one

or more partner business organizations that are in-

volved in the offering of these services.

Comparing this with the enterprise collaboration

through public UDDI registries, we observe the fol-

lowing beneﬁts:

• The SSP is based on extracting keywords from

WSDL type deﬁnitions, and represents these key-

words and their relationships with RDF. In this

way, the semantic properties of keywords can

be captured (CRASSO et al., 2010; Thuy et al.,

2008). This addresses the limitation (a) men-

tioned above.

• Since the extraction is automatic, the burden for

service providers to update descriptions is dramat-

ically lowered. Moreover, if the service broker is

able to poll for updates, the problem of ’disap-

pearing’ business organizations and ’ghost’ ser-

vices can be tackled. If a business organization

no longer supports its previously published ser-

vices, e.g. because it no longer exists, a poll for

updates by the service broker gets no reaction and

the service broker can decide to remove the SSP

from its registry. This addresses the limitation (b)

mentioned above.

• The two-step service discovery approach has the

advantage that it ﬁrst determines the services

providers that offer potentially relevant services,

and then limits the search for services to those of

the selected service providers. Although we still

have to conﬁrm this with experiments, we believe

that this approach has a better scalability than one-

level semantic search. Furthermore, by favoring

services from the same or a few providers, it is

more likely that these services are deﬁned and im-

plemented in a consistent way, making search and

composition easier and more efﬁcient (Forestiero

et al., 2010).

• In order to address limitation (c) mentioned

above, the local registry of a service provider

may be enhanced in two ways. First, the ser-

vice provider may monitor who wants to access

the local registry, and expose information on its

services depending on some trust classiﬁcation

scheme (e.g., based on previous collaborations).

Secondly, the service provider may provide ad-

ditional information through the local registry,

which facilitates the (non-) selection of services.

For example, non-functional properties based on

resource availability or historical data may be

published, including information on trust, security

or privacy aspects.

• The implications for existing standards, most no-

tably UDDI, is minimal. Most of the interactions

described above can be supported with UDDI as

is.

6 CONCLUSIONS

Our research demonstrates how SSP descriptions al-

lows us to reduce maintainability cost at the service

providers side while still providing the relevant infor-

mation required for service discovery. This type of

approach has the ability to guarantee better service

results due to the separation of abridged service de-

scriptions and the actual detailed descriptions. Cur-

rent approaches either do not provide adequate sup-

port for publishing accurate information of the offered

services or the offered solutions are too restrictive in

terms of cost and time required for maintaining the

published descriptions. This is mainly because the

service descriptions are valid only at the time they are

created and subject to frequent change depending on

changes in market trends.

ACKNOWLEDGEMENTS

This material is based upon works jointly sup-

ported by the IOP GenCom U-Care project

(http://ucare.ewi.utwente.nl) sponsored by the

Dutch Ministry of Economic Affairs under contract

IGC0816 and by the DySCoTec project sponsored

by the Centre for Telematics and Information

Technology (CTIT), University of Twente, The

Netherlands.

BMSD 2011 - First International Symposium on Business Modeling and Software Design

REFERENCES

AbuJarour, M., Naumann, F., and Craculeac, M. (2010).

Collecting, Annotating, and Classifying Public Web

Services. In Proc. of International Conference on On

the Move to Meaningful Internet Systems, pages 256–

272.

Al-Masri, E. and Mahmoud, Q. H. (2008). Investigating

Web Services on the World Wide Web. In Proc. of the

World Wide Web Conference, pages 759–804.

Cardoso, J., Winkler, M., and Voigt, K. (2009). A Service

Description Language for the Internet of Services. In

Proc. of the International Symposium on Services Sci-

ence.

Chinnici, R., Moreau, J.-J., Ryman, A., and Weerawarana,

S. (2007). Web Services Description Language

(WSDL) Version 2.0.

CRASSO, M., ZUNINO, A., and CAMPO, M. (2010).

Combining Document Classiﬁcation and Ontology

Alignment for Semantically Enriching Web Services.

New Generation Computing, 28:371–403.

da Silva, E. G., Pires, L. F., and van Sinderen, M. (2011).

Towards runtime discovery, selection and composi-

tion of semantic services. Computer Communications,

34(2):159–168.

Erl, T. (2005). Service-Oriented Architecture Concepts,

Technology, and Design. Prentice Hall Professional

Technical Reference.

Farrell, J. and Lausen, H. (2007). Semantic Annotations for

WSDL and XML Schema.

Forestiero, A., Mastroianni, C., Papuzzo, G., and Spez-

zano, G. (2010). A Proximity-Based Self-Organizing

Framework for Service Composition and Discovery.

In Proc. of the 10th IEEE/ACM International Confer-

ence on Cluster, Cloud and Grid Computing, pages

428–437.

Hayes, P., editor (2004). RDF Semantics. W3C Recom-

mendation.

Klyne, G. and Carroll, J. J., editors (2004). Resource De-

scription Framework: Concepts and Abstract Syntax.

W3C Recommendation.

uster, U. and K

oning-Ries, B. (2007). Supporting Dynam-

ics in Service Descriptions - The Key to Automatic

Service Usage. In Proc. of the 5th International Con-

ference on Service-Oriented Computing, pages 220–

232.

Lamparter, S., Ankolekar, A., and Grimm, S. (2007).

Preference-based Selection of Highly Conﬁgurable

Web Services. In Proc. of the 16th International Con-

ference on World Wide Web, pages 1013–1022.

Martin, D., editor (2004). OWL-S: Semantic Markup for

Web Services. W3C Member Submission.

Michlmayr, A., Rosenberg, F., Platzer, C., Treiber, M., and

Dustdar, S. (2007). Towards Recovering the Broken

SOA Triangle: A Software Engineering Perspective.

In Proc. of the 2nd International Workshop on Service

Oriented Software Engineering, pages 22–28.

Obrst, L., McCandless, D., and Bankston, M. (2010). En-

abling Rich Discovery of Web Services by Project-

ing Weak Semantics from Structural Speciﬁcations.

In Proc. of Semantic Technology for Intelligence, De-

fense, and Security.

Prud’hommeaux, E. and Seaborne, A., editors (2007).

SPARQL Query Language for RDF. W3C Candidate

Recommendation.

Roman, D., Lausen, H., and Keller, U., editors (2006). Web

Service Modeling Ontology (WSMO). WSMO Work-

ing Group.

Sabou, M. and Pan, J. (2007). Towards semantically en-

hanced Web service repositories. Web Semantics: Sci-

ence, Services and Agents on the World Wide Web,

5(2):142–150.

Speiser, S. and Harth, A. (2011). Integrating Linked Data

and Services with LIDS. In Proc. of the 8th Extended

Semantic Web Conference.

Thuy, P. T. T., Lee, Y.-K., Lee, S., and Jeong, B.-S. (2008).

Exploiting XML Schema for Interpreting XML Doc-

uments as RDF. In Proc. of the 2008 IEEE Interna-

tional Conference on Services Computing, pages 555–

558.

Treiber, M. and Dustdar, S. (2007). Active Web Service

Registries. IEEE Internet Computing, 11(5):66–71.

Truong, H.-L., Comerio, M., Maurino, A., Dustdar, S.,

Paoli, F. D., and Panziera, L. (2010). On Identifying

and Reducing Irrelevant Information in Service Com-

position and Execution. In Proc. of the International

Conference on Web Information Systems Engineering,

pages 52–66.

van Sinderen, M. (2009). From Service-Oriented Architec-

ture to Service-Oriented Enterprise. In Proc. of the

Third International Workshop on Enterprise Systems,

pages 3–16.

van Sinderen, M. and Almeida, J. P. A. (2011). Empow-

ering Enterprises through Next-Generation Enterprise

Computing. Enterprise Information Systems, 5(1):1–

LEVERAGING THE PUBLISH-FIND-USE PARADIGM OF SOA - Supporting Enterprise Collaboration Across

Organisational Boundaries