DYNAMIC SERVICE RECONFIGURATION AND ENACTMENT

USING AN OPEN MATCHING ARCHITECTURE

Sander van Splunter, Frances Brazier

Intelligent Interactive Distributed Systems, VU University Amsterdam, The Netherlands

Julian Padget

Department of Computer Science, University of Bath, U.K.

Omer Rana

School of Computer Science, Cardiff University, U.K.

Keywords:

Automated adaptation, Matching, Enactment.

Abstract:

An architecture for dynamic reconﬁguration of complex services, in which the enactment is automated, and

the matching of services is not limited to a pre-determined set of matchers and repositories, is presented.

The proposed architecture consists of three, previously developed, components: the CoWS template-based

reconﬁguration service, the Knoogle MatchMaker service, and the Triana workﬂow enactment engine. This

architecture has the following innovative aspects: 1) automated adaptation of complex services, which is more

ﬂexible than existing approaches based on replacing failing instances of services within a workﬂow, 2) use of

heterogeneous components that may be both local and distributed, and 3) dynamic selection of matchers and

repositories.

1 INTRODUCTION

Web services provide uniform access to software ca-

pability with well-deﬁned interface descriptions, and

provide some limited options for semantic annotation.

Many times, however, a single web service does not

provide the precise functionality required for a spe-

ciﬁc application, therefore web services often need

to be combined to form a complex service. If one

web service fails, the continuity of the complex ser-

vice is threatened. Instance-based replacement ap-

proaches offer one possible solution, but are limited to

the availability of exact (function and interface equiv-

alent) replacement services.

Current research allows automated handling of

failing services, for example, based on late binding,

but is often limited to instance replacement. The

number of services available is large, but dispersed

over different repositories, as different communities

build and collect services. Even the design and de-

ployment approaches may vary, for example, between

Service Oriented Architecture/Grid and Peer-to-Peer

based deployments.

The approach presented in this paper focuses on au-

tomated template-based reconﬁguration of annotated

web services, using (i) a discovery infrastructure with

dynamic matcher service selection, repository selec-

tion and service policy speciﬁcation, and (ii) a work-

ﬂow enactment engine capable of handling hetero-

geneous components. This combination provides a

fully automated reconﬁguration service that can be

used without human intervention. The resulting archi-

tecture enables: (i) automated service reconﬁguration

based on constraints that havebeen previouslydeﬁned

by a user; (ii) an open matching architecture that al-

lows multiple matching mechanisms to be used, on

a set of user-deﬁned repositories—this is particularly

useful to enable application-speciﬁc matchers to co-

exist alongside generic matchers based on term syn-

tax; (iii) generality through integration with a work-

ﬂow enactment engine, so that the approach can be

combined as a component within an existing work-

ﬂow. The remainder of this paper is organised as

follows: section 2 brieﬂy describes relevant research

for the three areas addressed: web service reconﬁgu-

ration, matchmaking and workﬂow enactment. Sec-

533

van Splunter S., Brazier F., Padget J. and Rana O. (2009).

DYNAMIC SERVICE RECONFIGURATION AND ENACTMENT USING AN OPEN MATCHING ARCHITECTURE.

In Proceedings of the International Conference on Agents and Artiﬁcial Intelligence, pages 533-539

DOI: 10.5220/0001663605330539

 SciTePress

tion 3 describes the proposed architecture. Finally,

section 4 discusses the potential of the architecture,

and future work.

2 RELATED WORK

Three main areas of research are related to this pa-

per: web service reconﬁguration, matchmaking, and

workﬂow enactment. A brief description of the cur-

rent state of art is presented for each.

Web Service Reconﬁguration. The Web Service

Reusable Flexible Flows (ReFFlow) project (Karas-

toyanova and Buchmann, 2004b) focuses on extend-

ing the meta-model of web service compositions

to enable adaptability at run-time. A secondary

goal for ReFFlow is automated creation of complex

workﬂowsusing templates (Karastoyanovaand Buch-

mann, 2004a). Supported adaptations are instance

replacement, parameterised adaptation, or structural

re-design. Adaptation in ReFFlow is limited in

scope, because of the reasoning required to under-

stand the effect of a service replacement with refer-

ence to the overall workﬂow. The ASTRO framework

(Pistore et al., 2005) supports a planning approach

to automated service composition based on abstract

BPEL4WS processes, and composition requirements.

The framework supports initial conﬁguration, recon-

ﬁguration, monitoring, and automated instrumenta-

tion of the services. Similar to ReFFlow, the reason-

ing on the effect of a replacement of a service in rela-

tion to the properties of the complex workﬂow is lim-

ited. Cardoso and Sheth (Cardoso and Sheth, 2002)

address the integration of new web services into ex-

isting workﬂows. Web services in this project are

blackbox components with associated Quality of Ser-

vice (QoS) parameters speciﬁed in DAML-S (the pre-

cursor to OWL-S). Component composition is based

on workﬂow integration, and supported by abstract

Service Templates, a self-deﬁned local repository and

discovery service. This approach requires human in-

tervention for adaptation for the reﬁnement of abstract

Service Templates. Neither the matching architec-

ture or the enactment is open. The MoSCoE frame-

work (Pathak et al., 2006) is a goal-based, design time

approach to composition, supporting both sequential

and parallel composition. Services are modelled as

Symbolic Transition Systems. MoSCoE starts with an

abstract description of the desired service, provided

by a human user, and supports the user in reﬁning the

description to a realisable BPEL workﬂow. The semi-

automated nature limits its use for dynamic reconﬁg-

uration.

Matchmaking. Matchmaking is used to discover dy-

namically suitable web services given service prop-

erties (which may also include QoS properties). In

the literature generic brokerage mechanisms are de-

scribed that use syntactic or semantic techniques for

this purpose, or a combination of both. Some of

the earliest systems, enabled by the development of

KIF (Knowledge Interchange Format) (Genesereth

and Fikes, 1992) and KQML (Knowledge Query

and Manipulation Language) (Finin et al., 1994), are

SHADE (Kuokka and Harada, 1996) operating over

logic-based and structured-text languages and the

complementary COINS (Kuokka and Harada, 1996)

that operates over free-text using well-known term-

ﬁrst index-ﬁrst information retrieval techniques. Sub-

sequent developments such as InfoSleuth (Nodine

et al., 1999) apply reasoning technology to the adver-

tised syntax and semantics of a service description,

while the RETSINA system (Sycara et al., 2003) has

its own specialised language inﬂuenced by DAML-S

and uses a belief-weighted associative network rep-

resentation of the relationships between ontological

concepts as a central element of the matching process.

While technically sophisticated, a particular problem

with the latter is how to make the initial assignment

of weights without biasing the system inappropri-

ately. A distinguishing feature of all of these sys-

tems is their monolithic architecture, in sharp contrast

to GRAPPA (Veit, 2003) (Generic Request Architec-

ture for Passive Provider Agents) which allows for the

use of multiple matchmaking mechanisms. GRAPPA

utilises a conventional multi-attribute clustering tech-

nology to reduce attribute vectors to a single value.

Workﬂow Enactment. Workﬂow enactment in-

volves managing the coordinated execution of web

services. A variety of representation schemes en-

code the connectivity between services ranging from

web service standard speciﬁcations such as WS-

BPEL (Jordan and Evdemon, 2007) to specialist lan-

guages that are application speciﬁc, such as Condor

DAGMan (Tannenbaum et al., 2002). Once such a

connectivity graph has been described (often referred

to as an “abstract” workﬂow), it is now necessary to

manage its execution – either on a single machine or

on distributed set of machines – often undertaken ei-

ther by a single scheduler (often managing assign-

ment of services to time slices on a single resource)

or via some distributed group of schedulers. Hence,

workﬂow scheduling is a kind of global task schedul-

ing as it focuses on mapping and managing the ex-

ecution of interdependent tasks on shared resources

that are not directly under its control. The workﬂow

scheduler/enactor needs to coordinate with diverse lo-

cal management systems taking account of their local

ICAART 2009 - International Conference on Agents and Artificial Intelligence

534

Figure 1: Architecture for dynamic service composition and enactment.

conﬁguration and policies (Yu and Buyya, 2005).

Our approach uses an open matchmaking architec-

ture, enabling application speciﬁc matching to be

used alongside general purpose matching. Integration

of this with the templates and workﬂow, render the

approach more generalisable and adaptive.

3 ARCHITECTURE

The architecture (Figure 1) combines (i) the CoWS

template-based web service reconﬁguration, (ii) the

Knoogle matchmaker, and (iii) the Triana enactment

engine. These three components are linked together

through a description of services expressed in OWL-

S. The reconﬁguration process is started after the de-

tection of service failure in the complex service. It

is activated with a description of the initial complex

service, and a pointer to the service that failed. Based

on the requirements posed to the failed service and to

the initial complex service, the CoWS reconﬁguration

uses the Knoogle matchmaker to identify a replace-

ment service, which is subsequently enacted by the

Triana workﬂow engine. After a brief introduction of

OWL-S in section 3.1, each of the three components

in the architecture are described in sections 3.2, 3.3,

and 3.4.

3.1 OWL-S

OWL-S (Martin et al., 2004) is an ontology for anno-

tating web services, based on the Web mark-up lan-

guage OWL (McGuinness and Van Harmelen, 2004).

OWL-S describes a main service using three different

sub-ontologies: the Service Proﬁle, Service Model,

and Service Grounding ontology. The Service Pro-

ﬁle deﬁnes what the service does, in terms of service

type, input, output, pre- and post-conditions. The Ser-

vice Model deﬁnes the internal workings of a service.

A Composite Process in a Service Model deﬁnes a

process composition of sub-processes to which refer-

ences are deﬁned, together with the control ﬂow spec-

iﬁed with pre-deﬁned control constructs. The Service

Grounding links the abstract description of the service

to actual implementation details, such as message ex-

change formats and network protocols so that auto-

matic invocation of the service is possible.

3.2 CoWS Web Service Reconﬁguration

CoWS is an automated template-based approach to

reconﬁguration of complex web services, in which

the scope of the adaptation is adjustable. In addition

CoWS enables reasoning on the effect of the replace-

ment of a service in relation to the properties of the

complex service (van Splunter et al., 2008).

A template describes a single-levelcomposition of

web services. This is deﬁned as one or more slots, a

control structure, and an associated template descrip-

tion. A slot is a placeholder for a web service or tem-

plate, and deﬁnes a set of requirements for that com-

ponent. The control structure deﬁnes the conditions

for activation of the slots. The template description

describes the functionality, behaviour and structure of

the template, and dependencies between the slots.

retrieve data

create display

slot

store data

filter data

slot

data

display

template

InterfaceinOWL-SGrounding

DescriptionofServiceinOWL-SProfile

ProcessdependenciesusingOWL-Process

SpecificationofInterfaceandrelevant

propertiesofabstractwebservices

Figure 2: Template description in OWL-S.

DYNAMIC SERVICE RECONFIGURATION AND ENACTMENT USING AN OPEN MATCHING ARCHITECTURE

535

A simple extension (Richards et al., 2004; Sabou

et al., 2003) has been deﬁned to specify templates in

OWL-S, as illustrated in Figure 2. Every template has

an OWL-S Proﬁle, describing what this combination

of slots would do, if used and reﬁned. The OWL-S

Service Model is deﬁned as a CompositeProcess with

slots deﬁning abstract sub-processes. Each slot is de-

ﬁned as a subclass of Process, and requirements for

each slot are included in this speciﬁcation. A tem-

plate description in itself has no Grounding Model.

The Grounding is added when a complete conﬁgura-

tion is created, based on the services and templates

inserted in the template.

In our architecture, the CoWS service is activated

given an initial complex service and an identiﬁer of a

failed service. CoWS then determines which slot was

ﬁlled with this service. Based on the slot deﬁnition,

functional and non-functional requirements are deter-

mined for a replacement service that needs to be cre-

ated. A conﬁguration process is started, which makes

use of Knoogle, whilst providing a selection policy,

the matcher and the repositories from which web ser-

vices are to be retrieved. The CoWS process ends

when an adapted template-based web service conﬁg-

uration is created which: (i) satisﬁes all requirements

in the slot of the failing service; (ii) has no open slot

in the resulting conﬁguration, and (iii) violates none

of the slot requirements in the resulting conﬁguration.

As shown in ﬁgure 1, CoWS supports policy determi-

nation, component selection and reﬁnement, and web

service/template retrieval.

The policy determination process chooses a pol-

icy specifying the strategy used for candidate selec-

tion. Examples of these policies are:

• Favour service above templates. If both services

and templates fulﬁl the requirements, choose a

service (the objective being to reduce the expected

complexity of the conﬁguration);

• Fewest number of slots is preferred. If a choice

between templates is needed, choose the template

with the fewest number of slots (the objective be-

ing to reduce the complexity of the conﬁguration);

• Most reﬁnement candidates. If a choice between

templates is needed, choose the template with the

highest number of reﬁnement candidates (the ob-

jective being to increase the likelihood of suc-

cess).

These selection policies are linked to the policies

given to Knoogle. The component selection and re-

ﬁnement process is responsible for the overall con-

ﬁguration process, in which a replacement service is

created for the failing service. The replacement ser-

vice is either a single web service, or a complex ser-

vice composed using templates. Initially, a slot is fo-

cused on for reﬁnement, and the requirements associ-

ated with the slot are retrieved. These requirements

and the selection policies are used to invoke web ser-

vice/template retrieval. If a match is found, then the

conﬁguration process refocuses, until no open slots

are left. If no match is found, the process backtracks,

as a result of which it may mark a larger part of

the initial conﬁguration as failing. If no alternative

services and/or templates are available at the high-

est node of the initial conﬁguration, the conﬁguration

process fails. Web service/template retrieval pre-

pares the requests to Knoogle to discover web service

and template descriptions that match OWL-S based

Proﬁle descriptions of the services/templates that will

fulﬁl slot requirements. The request includes the se-

lected policy.

3.3 Knoogle Matchmaker

Knoogle offers a brokerage framework that can be

used to deploy partially or fully-conﬁgured brokers

that query multiple service repositories, employ mul-

tiple matching services and apply pre-deﬁned or be-

spoke selection policies. The matchmaking and bro-

kerage process is characterised in terms of three es-

sential actions: 1. where to ﬁnd descriptions of en-

tities to match against; 2. how to match the query

against a description; and 3. how to choose between

the matched descriptions. To achieve this, Knoogle

receives four parameters from the CoWS service (see

Figure 1):

1. A query specifying the requirement in XML (by

default). However, the actual structure will be

application-speciﬁc, since it must be compatible

with whatever the matching service(s) expect. In

this case it will be an OWL-S expression.

2. A list of URLs of UDDI-compliant repositories.

These are the repositories to search when process-

ing a query. This application currently works with

two instances of Grimoires (Moreau, 2008) repos-

itories, one containing semantic descriptions of

services and one containing templates.

3. A list of URLs of match services. These are the

services used to compare a query against a candi-

date service or template description from a repos-

itory. A match service must return a number in

the range [0,1] to indicate the degree of match.

This data is stored in the Knoogle broker as an

RDF triple relating the match service, the degree

of match and the candidate service. Currently, we

utilise one similarity service (matcher), namely

OWLSM (Jaeger et al., 2005).

ICAART 2009 - International Conference on Agents and Artificial Intelligence

536

Algorithm 1: The core Knoogle algorithm.

input : a query q

input : a set of repository URLs R = {r

. . . r

}

input : a set of matcher URLs M = {m

. . . m

}

input : a selection policy sp

output: a service URL or set of service URLs

var : TS, a set of hm

, [0, 1], s

i triples

foreach r in R do1

foreach s in r.services() do2

foreach m in M do3

TS = TS.insert(triple(m, m(q, s), s))4

end5

end6

end7

4. A selection policy, speciﬁed as a query over the

RDF triples generated by the match services. The

query language is triple-store dependent. The cur-

rent version of Knoogle uses the Sesame(Openrdf,

www) triple store and so queries may be speci-

ﬁed in any of the three languages supported by

Sesame (RDQL, RQL and SeRQL). The poli-

cies used here determine either the most appropri-

ate service/template, or a ranked list of services

and/or templates, depending on the CoWS query.

The matching process of Knoogle is as follows: the

broker iterates overall the services/templates in all the

repositories, computing a match for each one against

the query using each one of the matchers, and then

applies the selection policy to determine the “best”

candidate(s). This process is formalised in Algo-

rithm 1. We emphasise that the algorithm captures

the requirements for the matching process, in that

it shows each matcher being invoked (serially) on

each query/service or query/template pairing. The

current proof-of-concept implementation does pre-

cisely this, but clearly for large numbers of heavily-

populated registries better technology, taking advan-

tage of caching and parallelism, is needed to enable

scalability.

3.4 Triana Enactment

After the CoWS reconﬁguration creates an adapted

complex service, the resulting service conﬁguration

needs to be enacted, for example as implemented

by the combination of CoWS and the SMDS work-

ﬂow manager in (van Veelen et al., 2008). However,

SMDS is limited to strictly controlled environments.

On the other hand, Triana

(Majithia et al., 2004) is

http://www.trianacode.org/

a problem-solving environment designed for both the

creation and execution of workﬂow graphs. Execu-

tion of these graphs can be done both locally and via a

distributed enactment process (using Triana servers).

Enabling enactment of reconﬁgured services in Tri-

ana allows integration of heterogeneous components

and both local and distributed management of the ex-

ecution of the services. To enable the execution of

complex template-based services in Triana, a wrapper

creates workﬂow graphs from the template-based web

service conﬁgurations submitted to Triana. The wrap-

per utilises the control structures of templates, which

are expressed in the ServiceModel of the OWL-S de-

scription of the templates. Currently a simple wrapper

is used, supporting sequential activation.

4 DISCUSSION

Combining a reconﬁguration service with a discovery

system and a workﬂowenactmentengine is described.

The architecture is designed to be ﬂexible: supporting

workﬂow adaptation, extensible from a matcher per-

spective, and generalisable through its use in a work-

ﬂow. One of the challenges for future research is to

increase robustness and scalability of the reconﬁgu-

ration and matching process (for instance, the ability

to pre-select repositories to search rather than search

all user deﬁned repositories). One option is to ex-

tend the conﬁguration and matching environment to

support parallel processing, another is to extend the

system to support a more extended set of policies in-

cluding heuristic-based policies. Interaction between

policies within the three different systems has also yet

to be explored.

Integrating three systems in the manner described

makes the approach more complex to use, as it is gen-

erally required for a user to specify: (i) service de-

scription; (ii) selection policy; (iii) choice of suitable

matchers; (iv) list of repositories to search. However,

some of these can be pre-conﬁgured—such as (ii),

(iii) and (iv). Currently, approaches to automated ser-

vice composition (as outlined in section 2) are limited,

either due to their lack of adaptability to speciﬁc ap-

plication domains, or their overall complexity of use.

By allowing different levels of system conﬁguration,

and enabling some of these to be pre-speciﬁed (not by

an end user, but an application/system administrator),

we believe it is possible to balance complexity of use

with generality of the overall approach.

DYNAMIC SERVICE RECONFIGURATION AND ENACTMENT USING AN OPEN MATCHING ARCHITECTURE

537

ACKNOWLEDGEMENTS

The research reported here is part of the Interac-

tive Collaborative Information Systems (ICIS) project

(http://www.icis.decis.nl), supported by the Dutch

Ministry of Economic Affairs, grant nr: BSIK03024.

The authors are also grateful for the support offered

for this research by NLnet (http://www.nlnet.nl), and

through the Open Middleware Infrastructure Institute

(http://www.omii.ac.uk) managed program (project

Knoogle). The authors also wish to thank Pieter van

Langen of VU University Amsterdam for his exten-

sive input and Neil Chapman of Cardiff for his contri-

bution to implementation of workﬂow enactment sys-

tems.

REFERENCES

Cardoso, J. and Sheth, A. (2002). Semantic e-workﬂow

composition. Technical report, LSDIS Lab, Computer

Science Department, University of Georgia.

Finin, T., Fritzson, R., McKay, D., and McEntire, R. (1994).

Kqml as an agent communication language. In Pro-

ceedings of 3rd International Conference on Informa-

tion and Knowledge Management, pp. 456-463.

Genesereth, M. and Fikes, R. (1992). Knowl-

edge interchange format, version 3.0 reference

manual. Technical report, Computer Sci-

ence Department, Stanford University. Avail-

able from

http://www-ksl.stanford.edu/

knowledge-sharing/papers/kif.ps

Jaeger, M. C., Rojec-Goldmann, G., M¨uhl, G., Liebetruth,

C., and Geihs, K. (2005). Ranked matching for ser-

vice descriptions using owl-s. In Kommunikation in

verteilten Systemen (KiVS 2005), Informatik Aktuell,

Kaiserslautern, Germany. Springer.

Jordan, D. and Evdemon, J. (2007). Web services business

process execution web services business process ex-

ecution language version 2.0. WS-BPEL. Available

at http://docs.oasis-open.org/wsbpel/2.0/OS/wsbpel-

v2.0-OS.html.

Karastoyanova, D. and Buchmann, A. (2004a). Au-

tomating the development of web service com-

positions using templates. In Proceedings of

Gesch¨aftsprozessorientierte Architekturen Workshop

at ’Informatik 2004’, Ulm, Germany.

Karastoyanova, D. and Buchmann, A. (2004b). ReF-

Flow: A model and generic approach to ﬂexibility of

web service compositions. In Proceedings of Inter-

national Conference on Information Integration and

Web-based Applications and Service, pages 27–29,

Jakarta, Indonesia.

Kuokka, D. and Harada, L. (1996). Matchmaking for infor-

mation integration. Journal of Intelligent Information

Systems, 6(2–3):261–279.

Majithia, S., Shields, M. S., Taylor, I. J., and Wang, I.

(2004). Triana: A graphical web service composition

and execution toolkit. In Proceedings of the IEEE In-

ternational Conference on Web Services (ICWS’04),

pages 514–524. IEEE Computer Society.

Martin, D., Paolucci, M., McIlraith, S., Burstein, M., Mc-

Dermott, D., McGuinness, D., Parsia, B., Payne, T.,

Sabou, M., Solanki, M., Srinivasan, N., and Sycara,

K. (2004). Bringing semantics to web services: The

OWL-S approach. In Proceedings of the First Interna-

tional Workshop on Semantic Web Services and Web

Process Composition (SWSWPC 2004), San Diego,

Califonia, USA.

McGuinness, D. and Van Harmelen, F. (2004). Owl web

ontology language overview, world wide web consor-

tium (w3c) candidate recommendation. february 10,

2004.

Moreau, L. (2008). Grid RegIstry with Metadata Oriented

Interface: Robustness, Efﬁciency, Security”. Avail-

able at http://twiki.grimoires.org/bin/view/Grimoires/.

Nodine, M., Bohrer, W., and Ngu, A. (1999). Semantic

brokering over dynamic heterogenous data sources in

infosleuth. In Proceedings of the 15th International

Conference on Data Engineering, pp. 358-365.

Openrdf (www). Sesame: Rdf schema querying and stor-

age. Available via

http://www.openrdf.org

. Last

accessed: February 2007.

Pathak, J., Basu, S., Lutz, R., and Honavar, V. (2006). Par-

allel web service composition in : A choreography-

based approach parallel web service composition in

MoSCoE: A choreography-based approach. In 4th

European Conference on Web Services (ECOWS ’06),

pages 3–12, Zurich, Switzerland.

Pistore, M., Marconi, A., Bertoli, P., and Traverso, P.

(2005). Automated composition of web services by

planning at the knowledge level. In Kaelbling, L. P.

and Safﬁotti, A., editors, IJCAI, pages 1252–1259.

Professional Book Center.

Richards, D., van Splunter, S., Brazier, F., and Sabou, M.

(2004). Composing Web Services using an Agent Fac-

tory, volume 13 of Multiagent Systems, Artiﬁcial So-

cieties, and Simulated Organizations, pages 229–252.

Springer Verlag.

Sabou, M., Richards, D., and Splunter, S. v. (2003). An

experience report on using daml-s. In Proceedings of

WWW 2003 Workshop on E-Services and the Semantic

Web (ESSW’03), Budapest, Hungary.

Sycara, K., Paolucci, M., Van Velsen, M., and Giampapa.,

J. (2003). The retsina mas infrastructure. Autonomous

Agents and Multi-Agent Systems, 7(1–2):29–48.

Tannenbaum, T., Wright, D., Miller, K., and Livny, M.

(2002). Condor: a distributed job scheduler. pages

307–350.

van Splunter, S., van Langen, P., and Brazier, F. (2008).

Decentralized performance-aware reconﬁguration of

complex service conﬁgurations. In Proceedings of the

Performance for Peer-to-Peer Systems (P4P2P) work-

shop. University of Warwick.

van Veelen, J., van Splunter, S., Wijngaards, N., and Bra-

zier, F. (2008). Reconﬁguration management of cri-

sis management services. In The 15th conference

of the International Emergency Management Society

(TIEMS 2008).

ICAART 2009 - International Conference on Agents and Artificial Intelligence

538

Veit, D. (2003). Matchmaking in Electronic Markets, vol-

ume 2882 of LNCS. Springer. Hot Topics.

Yu, J. and Buyya, R. (2005). A taxonomy of workﬂow man-

agement systems for grid computing. Journal of Grid

Computing, 3(3):171–200.

DYNAMIC SERVICE RECONFIGURATION AND ENACTMENT USING AN OPEN MATCHING ARCHITECTURE

539