Digital Ecosystems:Evolving Service-Orientated Architectures

Gerard Briscoe Philippe De WildeIntelligent Systems and Networks Group Intelligent Systems Lab

Department of Electrical and Electronic Engineering School of Mathematical and Computer ScienceImperial College London Heriot-Watt University

London, United Kingdom, SW7 2BT Edinburgh, United Kingdom, EH14 4ASe-mail: gerard.briscoe@ic.ac.uk e-mail: pdw@macs.hw.ac.uk

Abstract-A novel optimisation technique inspired by natural I. SERVICE-ORIENTATED ARCHITECTURESecosystems is presented, where the optimisation works at twolevels: a first optimisation, migration of services which are The SOA concept expresses a software architectural conceptdistributed in a decentralised peer-to-peer network, operating that incorporates modular reusable services that have clearlycontinuously in time; this process feeds a second optimisation defined and standardised interfaces.based on evolutionary computing that operates locally on sin-gle peers and is aimed at finding solutions to satisfy locally In a SOA environment, nodes on a network make resourcesrelevant constraints. Through this twofold process, the local available to other participants in the network, as independentsearch is accelerated and will yield better local optima, because services that the participants can access in a standardised way.the distributed optimisation already provides a good sampling Most definitions of SOAs identify the use of web services inof the search space by making use of computations already their implementation. Unlike traditional object-oriented archi-performed in other peers with similar constraints. We call this tecr SOAs comprise looe joned highly interpablenew distributed optimisation architecture a Digital Ecosystem, an tectures, SOAs comprise loosely joined, highly interoperableEcosystem-Orientated Architecture (EOA) created by extending a application services, and because these services interoperateService-Oriented Architecture (SOA) with Distributed Evolution- over different development technologies, the software compo-ary Computing (DEC). The Digital Ecosystem will allow services nents become very reusable [7].to recombine and evolve over time, constantly seeking to improve SOAs potentially promise to provide a huge number oftheir effectiveness for the user base. Individuals within our DigitalEcosystem will be applications (groups of services), created in services that programmers can combine, via the standardisedresponse to user requests by using evolutionary optimisation to interfaces, to create increasingly more sophisticated and dis-aggregate the services. These individuals will migrate through tributed applications.the Digital Ecosystem and adapt to find niches where they are There are multiple standards available, and still being devel-useful in fulfilling other user requests.

oped, for SOAs [6]. So, we will define how we view servicesWith the ever increasing number of services offered online in generic terms, and then we will consider an example

from Application Programming Interfacess (APIs) being made implementation. We consider each service to consist of anpublic, there is an ever increasing number of computational executable component, available remotely or locally, and aunits available to be combined to create applications. However, descriptive component. The descriptive component acting as athis is currently a task done manually by programmers. There guarantee of behaviour. The actual underlying code or serviceare several existing efforts aimed at achieving automatic only comes into play once the whole application has beenservice composition [2], [3], [4], [5], the most prevalent of assembled. A service can be a software service, e.g. for datawhich is SOAs and its associated standards and technologies encryption, or a real world service, e.g. selling books, which[6]. We propose a biologically inspired approach, expected to can be represented by a software service. This is widelymore affective than traditionally inspired approaches at greater applicable and easily adaptable definition, and we have used itscales, because it is built upon the scalable properties of natural to consider examples in the travel and manufacturing industriesecosystems. An Ecosystem-Orientated Architecture, called a [8].Digital Ecosystem, which itself is built upon a SOA and that The Semantics of Business Vocabulary and Business Rulesuses a novel form of DEC for the automatic combining of (SBVR) is a near natural language for describing businessavailable and applicable services to meet user requests for services. It features strong support for multilingualism,applications in a scalable architecture. the use of formal logic and has compliance with Meta

Object Facility (MOF) [9]. It allows business vocabulariesconstruction, and business rules definitions, enabling theirinterchange among organisations. Since it allows talking about

This work is supported by the European Commission under the EU project semantics, vocabulary and rules, this foundation is named asDigital Business Ecosystems (contract number 507953 [1]) 'Business Vocabulary+Rules'. The SBVR Vocabulary defined

I1-4244-0463-0/06/$20.OO ©2006 IEEE

Authorized licensed use limited to: Heriot-Watt University. Downloaded on April 29, 2009 at 05:40 from IEEE Xplore. Restrictions apply.

s an Xfor migrating

Fig.1Islan Mode ofDsrbtdEouinr0optn evcs Bi D- -servi-ce-

inoher ocablaris t crat an exene SBVRtVocabulary.

T Islandis 2a 25fw vsolaynt t S d tmo 5 2 a thereforeapplication r rt Sh e

group of servces Habitat for

54 2 v~~~~~~~~olutionsaeorepict moeBptnialspoiings imroe

solutions in the next generatio nP5 41P231 32 EX.LI m a t t n o m ptimisation n [ T

C~~rQr~.P sland isaditnc etenth4ubppuaioso ecailndnn

a probability of migration bw (response to a43 \ / ~~~~~~~~~~~~~~~ ~~~~~~~~~~~~~BizA request

/ 2wint~i~ This modelh(searchfor optimal s sfuorapplicatihe d r ao combination of t service-pooldc

Fig. 1. Island Model of Distributed Evolutionary Computing v s i

Fig. 2. Digital Ecosystem Modelis extensible since it is a vocabulary. It can also be includedin other vocabularies to create an extended SBVR Vocabulary.The following is a few vocabulary entries from the SBVR determines of ging of the different solutions relative to themetamodel: environment, and therefore their replication rate. So the better

tAt rental ~~~~~~~~~~solutions get to replicate more, potentially providing improved

GenraCaca'L business ent sltheonatral inspirednx qunealityta,athuhtopouain

etfCg0tShe name of cttat hvDEC aims to achieve parallel processing to find solutionscar rental ehasnae faster. We will focus on the Island Model which is probably the

car storage capa

most analogous to the notion of migration in nature [IO]. ThereDerrhiffi,011 numbr ofcars th1at can beatredttnccare thetuee t sue-popueafonson eaci ofiSans, an

a probability of migration between one island and another.Car rental ha car strare ac 999SOPM ~~~~~~~This model has been used successfully in the determination

of investment strategies in the commercial sector, in a productThe key point to stress is that SBVR descriptions consist known as the Galapagos toolkit [11].

Of descriptive elements defined by units and subunits, and The Island Model is visualised in Figure 1, in which therealso data elements when the model is populated. Show below are different probabilities of going from island'l' to island '2'is an SBVR vocabulary for a car rental service: as there is of going from island '2' to island 't'. This allows

car rental ~~~~~~~~~~maximum flexibility for the migration process. It also mirrorsGeneral Concept- businessentityi the naturally inspired quality that, although two populationsReferenceScheoem the name of the carrental have the same separation no matter how you look at them,

carrental h fnamne it may be easier for one population to migrate to another

solurstion Itthan vice-versa. An example of this from nature could beneriteos intoumaber ofcars that canbe yotoredecthe casrenal site the situation whereby it is easier for one species of fish to

migrate down-stream than for a different species to migrateca etlha5crs tra caa,,

upstream. However, all the islandson this approach work on

fTnes functwion whichmeis instantae withespdesciicpriob exactly the same problem, which makes it less analogousforTwhichalo sotioeentistoabefound. Tiefness tion: usto biological ecosystems in which different locations can be

The ar enta'Hrtz, ha cas strag ca.ac..t 00.environmentally different. We will take advantage of this laterin the EOA we create.

II. DISTRIBUTED EVOLUTIONARY COMPUTING

Evolutionary Computing (EC) is an optimisation technique11 H IIA csSE

capable of finding solutions in a large search space of possible The EOA of our Digital Ecosystem represents a novelsolutions. It works on the principle of using evolution to approach to DEC inspired from nature's ecosystems. An

Authorized licensed use limited to: Heriot-Watt University. Downloaded on April 29, 2009 at 05:40 from IEEE Xplore. Restrictions apply.

The service's semantic description acts as a guarantee of,'~/ , \\ \ 0its functionality, and is the inheritable component from one

generation to the next in the evolutionary optimisation. In anevolving population, thefitness of a combination of services(application) will be based on comparing the descriptivecomponents of the services with the complex description of

H\ & oX X, \\V/>ED \ X ll the user request, complex in the sense that more than oneH /,, \ ,/' " >,/' service is required to fulfil the request.

~---- - ' - The service is the base unit for the evolutionary mechanism,- - ~ ~ ~ which means that the evolutionary process optimises the

Fig. 3. Habitat Network combination of services to a user request. It does not changethe services themselves. The aggregated structure of servicesis a set. Therefore, the service-pool of a habitat is the set of

SOA, and we have enhanced this node to operate equivalently service(-set)s available at a habitat, and it will vary betweento a biological habitat, being a more sophisticated version the set of services available at the habitat and the power setof an island from the Island Model of DEC. We call this of the available services.enhanced node a habitat, with a network of habitats creating B Bythe Digital Ecosystem (Figure 2). Therefore, the organisms B olving Populationsof the Digital Ecosystem are the services of the SOA offered An evolving population, similar to the evolutionary aspectson the network. The habitats are connected through a peer-to- of an island from the Island Model of DEC, will be created topeer decentralised network. The connections between peers are use EC and evolve the fittest (optimal) solution(s) to a requestdynamically adapted on the basis of observed migration paths defined by the user. The population is seeded with the contentsof the services used between users within the habitat network. of the service-pool from the habitat in which the evolvingFollowing the idea of Hebbian learning, the habitats which population is instantiated. This allows the evolutionary op-exchange individuals more often obtain stronger connections timisation to be bootstrapped by previous solutions storedand habitats which do not exchange individuals will become in the service-pool. The evolving population of services willless strongly connected. This way, a network topology is then search the available service combination space throughdiscovered with time, which reflects the structure of the evolution to find the optimal solution(s) to a user request. Thebusiness-sectors within the user base. The resulting network fitness of individual service-sets within a population will bewill resemble the connectivity of the businesses within the determined by a selection pressure applied as a fitness functionuser base, typically a small-world network in the case of instantiated from the user request, and works primarily onSmall and Medium sized Enterprisess (SMEs) [13], [14], comparing the semantic descriptions of the services with the[15]. We have focused on SMSs since they make up 99% of semantic description in the user request. Mutations can occurenterprises in most EU Member States [16]. Such a network by switching services in and out of the service-set structure.has many strongly connected clusters called sub-networks Recombination (Crossover) can occur by combining elements(quasi complete graphs) and a few connections between these of two service-sets into a new service-set. The optimal solutionclusters. Graphs with this topology have a very high clustering (service-set) found can then be migrated through the intercon-coefficient and small characteristic path lengths [17], [18], as nected habitats, recombining with other services to meet moreshown in Figure 3. user requests in other Evolving Populations.

In the Digital Ecosystem, local and global optimisations C Habitatsoperate concurrently in finding solutions to satisfy differentoptimisation problems. The global optimisation here is not a The habitat is the key component for supporting the ecosys-super-peer based central control mechanism, but a completely tem concept within the EOA. The Digital Ecosystem willdistributed one which avoids a single point of failure. This consist of interconnected habitats just as in a biologicalallows for an optimised dynamic orchestration of services ecosystem. Each user has a dedicated habitat node, whichand their combinations, as well as a high degree of learning includes a service-pool of service(-set)s that are of potentialcapabilities for the services and for the infrastructure itself. use to the user, and which acts as a gene pool for the evolvingA. Services populations instantiated within the habitat. New solutions

evolved from the service-pool in response to user requestsAn organism in the Digital Ecosystem is actually a pointer are also stored within the service-pool, as the combination

to the software service it represents, which leads to the of services has value as well as the services themselves. Theservice's semantic description and executable component, habitat also provides a place for service migration to occur, aswhether they are local or remote. So, from a logical point the habitats are interconnected with one another. The union ofof view the individuals are equivalent to the services they rep- the habitats creates the EOA of the Digital Ecosystem.resent, and we will therefore refer to the services represented The novelty of our approach comes from the evolvinginstead of the pointers that represent them. populations being created in response to similar requests. So

Authorized licensed use limited to: Heriot-Watt University. Downloaded on April 29, 2009 at 05:40 from IEEE Xplore. Restrictions apply.

connections of the DEC Island Model shown in Figure 1). Theatmigrating connection probabilities affect the migration of services and

serviceare updated by the success or failure of the migration.

tat Hab t The migration of a service is triggered by it being deployedita __ 3 Xinto its Home habitat or when it is used by a user at any

X / "// g / habitat. When migration occurs, depending on the probabilitiesservice-set II associated with the connections; an exact copy of the service

Evolving Habitat is made at the connected habitat. The copy of the serviceis identical, then its migration history is updated which willdifferentiate it from the original.

E.m The success of the migration, the migration feedback, ismLtttion vo v ng E]ElC determined by the usage of services at the habitats to which

replieation/ FitnessX Population they migrate. When a solution (service-set) is found to a

d\ath \// service-pooluser request, then the individual service migration histories

- - rvicepoo can be used to determine where the used service has comefrom and update the connection probabilities. The challengeis in managing the feedback to the connection probabilities for

Fig. 4. Distributed Evolutionary Computing in the Digital Ecosystem migrating service-sets, because for a service-set the value iswithin the combination. The benefit of where the combinationor subsets of the combination were created needs to be passed

whereas in the Island Model (Figure 1) there are multiple on to the connection probabilities. Specifically, the mechanismevolving populations in response to one request, in the Digital needs to know where used service-sets are created. So that,Ecosystem there will be multiple evolving populations in new connections can be created or existing connections canresponse to similar requests. This is shown in Figure 4 where be reenforced, to where the service-sets where created.the shades of the evolving populations indicates similarity in The escape range is the number of escape migrationsthe requests being managed. For example, the four habitats in possible upon the risk of death. If a service migrates to athe top left of Figure 4 may all be travel agents, the three with habitat and remains unused after several user requests, thenevolving populations may be looking for package holidays it will have the opportunity to migrate (moved not copied)within the same continent. A great holiday package (service- randomly to a another connected habitat. After this happensset) found and used in one habitat, will then be migrated to several times the service will be deleted (die). The escapethe other connected habitats where it will be integrated into range is a parameter that will be better determined by futureany currently active evolving populations via the local service- work and practical experience. Ideally the escape range will bepool. So this will potentially help to optimise the search of adjusted to the size of the habitat cluster it exists within. This issimilar package holidays at the habitats of the other travel not obvious, because the view of a cluster very much dependsagents. This will also work in a time-shifted manner, because on the point of view of the habitat. Creating a definition thatthe great holiday package (solution) will be stored in the is dynamically responsive to the state of the system is theservice-pools of the habitats to which it is migrated. So it preferred solution. This would effectively create a dynamicwill potentially be available to optimise similar requests placed time-to-live for the services, but even so, services that are usedlater. will live longer than those that are not.

Our approach of utilising similar requests does require thatthe Digital Ecosystem has sufficient critical mass in the user E. Habitat Clusteringbase, for there to be similarity in the requests of users within The connectivity in the habitats will be parallel to thethe community. Assuming there is hundreds of habitats there existing communities that are within the user base, but thewill be potentially three or more times the number of Evolving infrastructure also allows for the spontaneous formation ofPopulations at any one time. There will then be thousands of new communities via clustering in the habitat network. Asservices, and a potentially huge number of possible service-set shown in Figure 5, many successful service migrations willcombinations (applications). In such a scenario, there would be reinforce habitat connections, thereby increasing the proba-a sufficient number of user requests for the Digital Ecosystem bility of service migration. If a successful migration occursto find similarity within them to apply the novel DEC approach through multi-hop migration, then a new link can be formedof the EOA. between those habitats. Unsuccessful migrations will lead to

connections (migration probabilities) decreasing, until finallyD. Service Migration the connection is closed.

The inter-habitat connectivity is the migration of services Each cluster of habitats represents a community withinfrom one habitat to another. Each connection between the the user base. The clustering/community formation is likehabitats is bi-directional, and there is a probability associated joining a club; you can be a member of more than one. Inwith moving in either direction across the connection (like the the same way habitats can be clustered into more than one

Authorized licensed use limited to: Heriot-Watt University. Downloaded on April 29, 2009 at 05:40 from IEEE Xplore. Restrictions apply.

Ur g0 0 I-

ax xI0 0,bt Zi # f Ilabitat x x Stx x x x x xxx

Fig5. Habta Clsern ,_XXXX-Community X }W ;x"Formtio Digta Ecosystem

iXN ... _ r HHY / \ n' r r @= L r~~~~~~s XX X X X

c-O ie ta fSArchitectureX

over language, business sector, nationality, geography, etc.Time(userrequestevexns x)

IV.IMULTIONAND ESULS Fi. 6 Gah: Prormanceo h igia Eoytm agis1ataitoa

o 2~~~~~~5 l/SOA bae systemcoysem

ThemuDigitalhEcosystemin wasimulatead toomntiesth efciusency 00 20 30 00

of the system in responding to user requests. Throughout thesimulation we assumed a hundred users, which meant that at hv rae ope ytmt rvd clbesltoany time, the number of users that joined the network was thtieqa otenubro sr thatwoullev.Teusr' t n essence greater than the sum of its parts, and expected

toshow emergent and complex behaviour that cannot behabitats were initially randomly connected. At the start, users peitdutli scetd h xeietlrslsidctdeployed five services each. Services could migrate from one thtudrsmliocniinsheDgalEsyem u-habitat to another thereby affecting the connections between pefom th coprsnsse.ae nataiinlSAthe habitats. Users would submit requests for services. Each Th Diia Ecsse moe a rsne tte20user then deployed a new service after the submission of three Jav On ofrne[9,adaddctdoe-oresm

Jav On ofrne[9] n eXaedoe-oresm

requests. The order in which users submitted requests was uainfaeokhsbe rae oass h nonrandom. investigations and research into the Digital Ecosystem [20].

An important measure for determining the success of the A.O owihti prac sbigapidi h Digital.Digta Ecsytemisitsreatie erfrmncetoa tadtioal Business Ecosystem (DBE), an EU Framework VI project

SOA based system. We simulated a simple SOA with a []TeDEam ob etgnrto nomto ndistributed service registry. A typical simulation run of the communications technology. In essence, the DBE will be anDigital Ecosystem under the above conditions was compared Inent-ae eniomn in whc 'uiess ilb betwith a typical run of the SOA based system. The time available itrc ihec te nvr fetv n fiin asto the SOA based system for the search in the distributed We have created a prototype Digital Ecosystem within theregistry was limited to the time the Digital Ecosystem required DBXaldteEouinr nvrnet(v) osrceto perform the same user request. In Figure 6 we graphed'upon the SOA of the DBE, the Execution Environment (ExE)the percentage match to the user requests for typical runs, as [21 [2] whc wilueteEouinr'optn nJvdetermined by a distance function between the request and the '.

servic dsrpin. (ECJ) library for the evolutionary optimisation. The core ofthe EvE Digital Ecosystem iS currently being implementedBoth the Digital Ecosystem and the SOA based system per- [3i h B rjc ySlbr nvriyo ple

formed as expected, providing better responses to user requests Sine,SnMcoytm brc,TcIES n neas mre srviesbcameavalabl. Te SO refrene sytem Ireland. Soon it will be possible to collect real world data to

initially performed better than the Digital Ecosystem, butwith the increasing number of services the Digital Ecosystem ivsiaefrhrtecmlxntr faDgtlEoytmbegan to outperform the reference system and continued to ACKNOWLEDGMENTdo so. This was anticipated as the the Digital Ecosystem wasexpected to be more effective at larger scales [19]. The authors would like to thank for their encouragement and

suggestions: Paolo Dini of the London School of EconomicsV. CONCLUSION ~~and Political Science, Thomas Heistracher and his group

A hybrid platform combining SOAs with DEC has been of the Salzburg University of Applied Sciences, Jonathancreated that is the digital counterpart of an ecosystem. We Rowe of the University of Birmingham, and Miguel Vidal of

Authorized licensed use limited to: Heriot-Watt University. Downloaded on April 29, 2009 at 05:40 from IEEE Xplore. Restrictions apply.

Sun Microsystems Iberica. The Digital Ecosystem model was ABOUT THE AUTHORSconstructed through interacting with these people and others. Gerard Briscoe is a Research Assistant at the Intelligent

Systems Lab of the School of Mathematical and Computer Sci-REFERENCES ences, Heriot-Watt University, Edinburgh, United Kingdom,

[1] F... Project, "Digital business ecosystems project." [Online]. Available: while working towards his Ph.D. at the Intelligent Systems andhttp://www.digital-ecosystem.org/ Networks group of the Department of Electrical and Electronic

[2] N. Milanovic and M. Malek, "Current solutions for web service com- Engineering, Imperial College London, United Kingdom. Heposition," Internet Computing, IEEE, vol. 8, no. 6, pp. 51-59, 2004. is also a visiting academic at the Digital Ecosystems Lab of

[3] M. Bossardt, R. Antink, A. Moser, and B. Plattner, Lecture Notes inComputer Science. Berlin, Heidelberg: Springer, 2004, vol. 2982, ch. the Department of Media and Communications at the LondonChameleon: Realizing Automatic Service Composition for Extensible School of Economics and Political Science, United Kingdom.Active Routers, pp. 163-177. His research interests focus on computational biomimicry,

[4] S. Mcllraith, C. Son, and H. Zeng, "Semantic web services," IntelligentSystems, IEEE, vol. 16, no. 2, pp. 46-53, 2001. which iS the applicaton of models, theories and ideas from

[5] J. Rao and X. Su, Lecture Notes in Computer Science. Berlin, Hei- biology to software systems. This includes evolutionary com-delberg: Springer-Verlag, 2005, vol. 3387, ch. A Survey of Automated puting and Biological Design Patterns (BDPs). He received hisWeb Service Composition Methods, pp. 43-54. puteng a n giclesig Patterns Dps recv his

[6] D. Norfolk, "Avoiding soa standards-based chaos." Masters of Engineering degree from the Department of Coi-[7] T. Erl, Service-Oriented Architecture, Concepts, Technology and Design. puting, Imperial College London, in 2002. He is a member of

Prentice Hall PTR, 2005. the Association for Computing Machinery.[8] G. Briscoe and P. D. Wilde, "D6.6 high-level design specification of

the distributed intelligence system," Digital Business Ecosystem, vol. Philippe De Wilde iS a Professor at the Intelligent Sys-Contract no 507953, 2006. tems Lab, School of Mathematical and Computer Sciences,

[9] M. D. Tommasi, "D15.3 BML framework 2nd release," Digital Business Heriot-Watt University, Edinburgh, United Kingdom. ResearchEcosystem, 2005. interests: stability scalability and evolution of multi-agent

[10] D. Goldberg, Genetic Algorithms in Search, Optimization and Machine iLearning. Addison-Wesley, 1989. systems; networked populations; coordination mechanisms for

[11] M. Ward, "Life offers lessons for business." [Online]. Available: populations; group decision making under uncertainty; neuralhttp://news.bbc.co.uk/2/hi/technology/3752725.stmhttp:Hnws.bbc.o.uk/2/i/technlogy/372725.st networks, neuro-economics. He tries to discover biological and[12] M. Begon, J. Harper, and C. Townsend, Ecology: Individuals, Popula- n no H t t dtions and Communities. Third Edition. Blackwell Publishing, 1996. sociological principles that can improve the design of decision

[13] X. Yang, "Chaos in small-world networks," PHYSICAL REVIEW E, making and of networks. Research Fellow, British Telecom,vol. 63, no. 046206, 2001. ,, 1994. Laureate, Royal Academy of Sciences, Letters and Fine

[14] Censis, "Deliverable 27.1: Territorial social capital and driver smes," Arts of Belgium, 1988. Senior Member of IEEE, Member ofDigital Business Ecosystems Project, 2005.

[15] 0. of Europeans SMEs, "Smes and cooperation," European Commission, IEEE Computational Intelligence Society and Systems, Man2003. and Cybernetics Society, ACM, and British Computer Society.

[16] F. Nachira, "Towards a network of digital business ecosystemsdigital Associate Editor, IEEE Transactions on Systems, Man, andbusiness ecosystems fostering the local development," European Com-mission DG INFSO, Tech. Rep., 2002. Cybernetics, Part B, Cybernetics.

[17] D. J. Watts and S. H. Strogatz, "Collective dynamics of 'small-world'networks," Nature, vol. 393, no. 6684, pp. 440-442, 1998.

[18] M. E. J. Newman and D. J. Watts, "Scaling and percolation in the small-world network model," Physical Review E 60, vol. 6, pp. 7332-7342,1999.

[19] G. Briscoe, M. Chli, and M. Vidal, "BOF-0759 Creating aDigital Ecosystem: Service Orientated Architectures with DistributedEvolutionary Computing," in JavaOne Conference, 2006. [Online].Available: https://www28.cplan.com/javaoneO6\-cv\A124\A1/session\details.jsp?isid=277759\&ilocation\id=124- 1\&ilanguage=english

[20] T. Kurz. [Online]. Available: http://evesim.sourceforge.net/[21] Digital Business Ecosystem Project, "Execution Environment." [Online].

Available: http://swallow.sourceforge.net/[22] T. Heistracher, T. Kurz, C. Masuch, P. Ferronato, M. Vidal, A. Corallo,

P. Dini, and G. Briscoe, "Pervasive service architecture for a digitalbusiness ecosystem," in ECOOP First International Workshop on Co-ordination and Adaptation Techniques for Software Entities (WCAT04),June 2004, http://wcatO4.unex.es/.

[23] C. Masuch, "Evolutionary environment network." [Online]. Available:http://evenet.sourceforge.net/

[24] B. Sleeper and B. Robins, "Defining Web Services."[Online]. Available: http://www.perfectxml.com/Xanalysis/TSG/TSG\DefiningWebServices.pdf

[25] W. G. Hale, J. P. Margham, and V. A. Saunders, Dictionary ofBIOLOGY,2nd ed. HarperCollins Publishers Inc., 1995.

[26] A. Redmore and M. Griffen, Longman Reference Guides: Biology,7th ed. Longman Group Limited, 1994.

[27] W. Beck, K. Liem, and G. Simpson, LIFE: An Introduction to Biology,3rd ed. HarperCollins Publishers Inc., 1991.

[28] W3C, "Web Services Activity," 2004. [Online]. Available: http:llwww.w3.org/2002/ws/

Authorized licensed use limited to: Heriot-Watt University. Downloaded on April 29, 2009 at 05:40 from IEEE Xplore. Restrictions apply.