MoSE 2010 Proceedings

Valeria de Castro, Juan Manuel Vara, Esperanza Marcos, Mike Papazoglou, Willem-Jan Van den Heuvel (Eds.)

2 ^nd International Workshop on

Model-Driven Service Engineering (MoSE 2010)

Málaga, Spain, 29 June 2010.

MoSE 2010

Proceedings of the 2 ^nd International Workshop on Model-Driven Service

Engineering

In conjunction with the

TOOLS 2010 Federated Conferences

Organized by

Kybele Research Group

Department of Languages and Computing Systems II Rey Juan Carlos University

Supported by

Rey Juan Carlos University

Preface

Model-Driven Engineering (MDE) deals with the provision of models, transformations between them and code generators to address software development. One of the main advantages of model-driven approaches is the provision of a conceptual structure where the models used by business managers and analysts can be traced towards more detailed models used by software developers. This kind of alignment between high level business specifications and the lower level Service Oriented Architectures (SOA) is a crucial aspect in the field of Service-Oriented Development (SOD) where meaningful business services and business process specifications are those that can give support to real business environment usually changing with increasing speed.

SOD has become currently in one of the major research topics in the field of software engineering, leading the appearance of a novel and emerging discipline called Service Engineering (SE), which aim to bring together benefits of SOA and Business Process Management (BPM). SE focuses on the identification of service (a client-provider interaction that creates value for the client) as first class elements for the software construction. The convergence of SE with MDE can holds out the promise of rapid and accurate development of software that serves software users’ goals.

In this context, the 2

Workshop on Model Driven Service Engineering (MoSE 2010) aims to provide a forum to discuss different issues related to SE in conjunction with MDE, boarding open research problems in this area as well as practical experiences. Particular interests include methods, modelling languages, development methodologies and techniques in the field of SOD.

We have received in this edition 10 contributions. All the papers received have been reviewed by, at least, three members of the international Program Committee. As the result of the review process 5 works were accepted as regular papers for their presentation at MoSE 2010 workshop. Moreover, we have also heard Dr. Eelco Visser from the Department of Software Technology at Delft University of Technology who presented the invited lecture: “Service Models for WebDSL and Mobl”.

We wish to thank all the contributors to MoSE 2010, in particular the authors

who submitted papers and likewise, we acknowledge the time and effort

contributed by all the members of the Program Committee who have very carefully reviewed the submitted papers. In closing, we would like to thank the Rey Juan Carlos University for their financial support.

June 2010

Valeria De Castro Juan Manuel Vara Esperanza Marcos Mike Papazoglou Willem-Jan Van den Heuvel Organization Chair

MoSE 2010

Workshop Organization

Workshop Organizers

Valeria De Castro Rey Juan Carlos University, Spain Juan Manuel Vara Rey Juan Carlos University, Spain Esperanza Marcos Rey Juan Carlos University, Spain Mike Papazoglou Tilburg University, Netherlands Willem-Jan Van den Heuvel Tilburg University, Netherlands

Progam Committee

Yuan An Drexel University, USA

Rolv Braek University of Science and Technology, Norway Jorge Cardoso University of Coimbra, Portugal

Alfonso Castro Telefonica I+D, Spain Rafael Corchuelo University of Seville, Spain Marcos Didonet Del Fabro IBM Software Group, France Ruben Fuentes Technical University of Madrid

Nora Koch Ludwig Maximilians University, Germany Guadalupe Ortiz Bellot University of Cadiz, Spain

Genoveva Vargas Solar CNRS, LSR-IMAG, France Organizing Committee

Verónica Bollati Rey Juan Carlos University, Spain

Carlos Cuesta Rey Juan Carlos University, Spain

Elisa Herrmann Rey Juan Carlos University, Spain

Marcos López Rey Juan Carlos University, Spain

Diana Sánchez Rey Juan Carlos University, Spain

Belén Vela Rey Juan Carlos University, Spain

Table of Contents

Modelling Self-Management in Service-Oriented Systems using SelfMML.

Carlos Rodriguez, Jorge Jesus Gomez Sanz and Juan Pavon.

On the Design of a Domain Specific Language for Enterprise Application Integration Solutions. Rafael Z. Frantz, Carlos Molina Jimenez and Rafael Corchuelo.

Tool support for Service Oriented development from Business Processes.

Andrea Delgado, Ignacio García-Rodríguez de Guzmán, Francisco Ruiz and Mario Piattini.

Organic Aggregation Service Engineering Framework (OASEF): A New Model-driven Approach to Service Engineering. Yuanzhi Wang.

Inference of performance annotations in Web Service composition models.

Antonio García-Domínguez, Inmaculada Medina-Bulo and Mariano Marcos-

Bárcena.

Servie-Oriented Systems using SelfMML

CarlosRodríguez-Fernández

1

,JorgeJ.Gómez-Sanz

1

,andJuanPavón

1

FaultaddeInformátia,

C/Prof.JoséGaríaSantesmases,s/n,

28040Madrid,Spain

{arlosro,jjgomez,jpavon}fdi. um.e s

Abstrat. This paper introdues a language alled Self-Management

Modelling Language (SelfMML) whih supports the modelling of self-

management apability requirements.The paper presentsa ase study

relatedtotheautomatire-bindingfeaturesinservieswhihillustrates

andanalysesthelanguageusageinservie-orientedsystems.

1 Introdution

Software Systems intended to provide servies usually should be operative at

peak performane 24/7 to meet the end-user needs and the business require-

ments. Therefore, it is desired that suh systems have the apability of man-

aging themselves withouthuman intervention, sineasystem ould donesuh

operations in a faster and more aurate way. This apability is alled self-

management[1℄.

Theself-managementdenitionisdetailedbyitsfouraspets:self-optimisation,

self-onguration,self-healingandself-protetion[1℄.Self-optimisation istheau-

tomatilookingandndingofopportunitiesto tunethesystemtoimprovethe

performaneandtheeieny;Self-ongurationistheautomationguringof

thesystemfollowinghighlevelpoliies;Self-healing istheautomatireovering

from unhealthy state; and Self-protetion is the proteting itself from attaks

andasadingerrors,withantiipatoryorreativeations[1℄.

Obtainingsuhapabilitiesin aservie-orientedsystemisnotatrivialwork

and require a ostly engineering eort. This work ould be failitated if the

developerhad speialised toolswhih support thedenition of those apabili-

ties.Tosupport this laim,this paperstudies theimpat ofapplying theSelf-

ManagementModellingLanguage(SelfMML)forthemodellingofself-management

apabilityrequirementsinaservie-orientedsystem.

TheSelf-ManagementModelling Language(SelfMML) is alanguagewhih

intendstoassistintheengineeringofself-managementapabilityrequirements,

providing visual representations related to the speiation of them. A visual

editortoolforthislanguageisprovidedtoreate,view,editandstoreSelfMML

speiations.Thistoolanbedownloadedat http://selfmml.sf.net.

Thehosenasestudyforapplying SelfMMLis basedonawellknown se-

management apability requirement that will be modelled using the proposed

language.

It is important to remark that we fous on the modelling aspets, not in

the requirementengineering proess. Hene, we intend to provide amodelling

language that permits a developer to eetively apture and speify a self-

managementrequirement,butwedonotprovideassistanefortherequirement

engineeringproessthatusesthislanguage.TheSelfMMLsopeislimitedtothe

self-managementrequirementspeiation.Aself-managementapabilityrepre-

sents an expeted behaviour from the system, whih should be implemented

in some way during the hosendevelopment proess. Duringthe realisationof

these,theengineeranidentifyelementsanddesignthearhitetureofthenal

system aording to the speied expeted behaviour, analogous to the reali-

sation of useases.The developerould onsider someframework orreferene

arhitetureforself-managementorjustdevelopanad-hosolution.A method

torealiseandverifyself-managementrequirementsmodelledbySelfMMLisnot

studyin thiswork.

Thisworkisstruturedasfollows.Setion2desribesthelanguageSelfMML.

Setion 3 presents a ase study used as illustrative example of the language

usage.Setion 4showssomerelatedwork thathasbeentakenintoaountfor

thislanguage.Setion5introduestheonlusions.

2 The Self-Management Modelling Language (SelfMML)

TheSelf-ManagementModellingLanguage(SelfMML)isalanguageto beused

in the modelling of self-management apabilities that a system should have,

that is, self-management apability requirements. This languageis madefrom

UML2.2Superstruture,onretely,opiesallelementsfromtheUseCases and

Ativitiespakagesandextendsthemwithnewelements.Alsoimportselements

fromtheKernel pakageforthedenitionoftheelements.

The language has a meta-model that denes its abstrat syntax. Further

information about the meta-model an be found in http://selfmml.sf.net.

Thelanguageisdesribedbelow(seegures1,2and3).

Fig.1.SelfMML(1)

Self-ManagementCapability.Self-ManagementCapabilitiesaretheabil-

itiesofsystemsto domanagement operationbythemselvesonthemselves.

Thiselementisprovidedfortherepresentationofself-managementapability

requirementsin asystem.

Self-ManagementCapabilitiesareusuallyrelatedtoqualityrequirementsas

maintainability,portability,reliability,usability,availability,amongothers[2,3℄,

inawaythatontributetothesatisfationofthem.ThelanguagehastheQuality

GoalandQualityLevelelementsfortherepresentationofqualityrequirements

(seegure1).

Quality Goal. Thiselementrepresentsaqualityrequirementdesribed as

agoalthatthesystemshouldmaintain.Ithasaharateristipropertythat

desribestheharateristiorfatorrelatedtothequalityrequirements,and

alsohasasatisfationpropertythatdesribeshowthequalitygoalissatised

by the using of some expression in natural language or another language

suhasObjetConstraintLanguage(OCL).Suhexpressionsoulddesribe

what are aeptable values for the quality metris related to the quality

requirement,followingtheIEEE1062-1998Standardreommendations.

QualityLevel.Thiselementrepresentsaqualitylevelwhihgroupsquality

goalsthatthesystemshouldmaintain.

The language lets developers model how self-management apabilities are

relatedtoQualityRequirementsandUseCases,bytheusingoftherelationships

ontributeforqualitygoalsandrequireforuseases.Alsotheinluderelationship

isprovidedtodesribewhatqualitygoalsareinludedinaspeiqualitylevel

(seegure2).QualityGoalsanbeonnetedtodesribeontributionwiththe

ontribute relationshiptoo.

Problemsare thetarget of self-protetionand self-healingapabilities. The

languageprovideselementsto model possibleproblemsin thesystemfollowing

thephilosophyfromtheFailureModesandEetAnalysismethods(FMEA)[4 ℄.

Theseelementsare:

Failure.Thiselementdesribesafailurethatouldhappeninthesystem.

Failure Case.Thiselementdesribeswhat is thewrong behaviour thata

systemshowswhensomefailurehashappened(failuremodesinFMEA).

Misuse Case. This element desribes a misuse of a system that an user

A failure, amisuse aseoran attakan be related to aqualitygoal with

thethreaten relationshipindiatingthat therstonesthreatenthesatisfation

of theseondones.Failuresanbeonnetedto failureaseswiththeativate

relationshipindiatingthattherstonesativatethewrongbehaviourdesribed

in the seond ones.Alsofailures anbe onnetedto use aseswith theaet

relationship indiating that the rst ones aet the normal operation of the

funtionality desribed in the seond ones. Failures, misuse ases and attaks

anbeonnetedtofailureswiththeauserelationshipindiatingthattherst

onesausetheseondones.

A self-management apability, as a self-protetion apability, an be on-

neted to problems (failures, misuse aseand attaks)to indiate that the a-

pability takesantiipatoryations to avoid, redue the onsequeneor redue

thelikelihoodofsuhproblemsusingtheavoid,redueConsequeneandredue-

Likelihood relationships respetively. Also, as a self-protetion apability, the

elementsan be onneted to attakswith the defend relationshipto indiate

that the systemreatsto suh attaks;and anbe relatedto failureswith the

isolate relationshiptoindiatethatthesystemisolatessuhfailurestoavoidthe

asadingfailuresthat theyouldause.

Aself-managementapability,asaself-healingapability,anbeonneted

to a failure with the reoverFrom relationship indiating that the apability

reoversthesystemfromthefailure.Also,itanbeonnetedtoafailurease

with the deativate relationship indiating that the apability deativates the

wrongbehaviourofthesystem.

Aording to the self-onguration and self-optimising aspet of the self-

management, aself-management apability ouldhave theintentionto tune a

ertainongurationparameterstoimprovetheperformaneofthesystem,and

ouldalsohavetheintentionto ompleteorupdate aertainongurationpa-

rametersfollowinghighlevelpoliies,inreationtohangeeventsinthesystem

orin theenvironment.SelfMML providestwoelementsto modelsuhongu-

rationsandhangeevents:

Change Event.Thiselementdesribesahangeeventinasystem.

Conguration.Thiselementdesribeaongurationdesription.

A self-management apability ould be related to aonguration with the

followingrelationships:tune,ompleteandupdate,indiatingthattheapability

trytotune,ompeteorupdate theonguration.Aapabilityouldberelated

to ahange event with treat relationship indiating that the apability treats

theevent.

SelfMML providesaset of ativity nodesto speifytheabstrat proess of

a self-management apability using ativities diagrams (see gure 3). A self-

managementproess usuallyhasastrutureaordingtofour phases[1℄:moni-

toring,analysis,planning,andplanexeution.Inmonitoringphasetheinterest-

inginformation isgathered;in theanalysisphase,theinformationis proessed

theplanexeutionphasetheseletedorbuiltplanisexeuted.

Fig.3.SelfMML(3)

Theelementsformonitoringandanalysisarethefollowing(seegure3):

Monitoring Node. This element desribes the monitoring ativities and

itontinuallygeneratestokensafter eahmonitoringyle. Themethod to

obtaininformationbymonitoringouldbepulling,pushing,intereptionor

anyotherkindorvariant;thenodedoesnotimplytheusingofanypartiular

method.

AnalysisNode.Thiselementdesribestheanalysisativitytoinferknowl-

edgefromthemonitoringreports.

Theprovided elements to model the seletion oronstrution of plans are

thefollowing(seegure3):

PlanSeletionNode.ThiselementisfortheseletionofplansusingOCL

expression,itreeivestokensfromtheinomingedgeandopyoneforeah

outgoingedges. Theontinuity ofthetokendepends onthePlan Seletion

GuardNode.

Plan SeletionGuard Node.Thiselementis toonstraintheexeution

of plans, it reeives tokens from the inoming edge and presents them to

theoutgoingedgesonlyiftheOCLexpressiondesribedinthespeiation

propertyisevaluatedto true.

Plan Seletion Proess Node.This elementdesribesaseletionwhih

isdonebyamoresophistiatedproessthatannotbeexpressedbyOCL.It

opiesallinomingtokensto alloutgoingedges,but theontinuityofsuh

tokensdependsontheevaluationoftheguardsontheoutgoingedges.Then,

asimpletermsin guardsanbeusedtodesribewhatplanisseleted.

PlanConstrution Node.Thiselementdesribesaplanonstrutiona-

tivity.

Inorder to speify the plan exeution ativitiesthe language provides the

followingelements(seegure3):

Plan Step Exeution Node. This element desribesa stepof a ertain

exeuted(usefulwhentheplanisonstruted).

3 Case Study: The Servie Re-binding

Thestudiedsystemisablogsystemsuh asblogger.om. Publisheransubmit

postsusing apublishingservie. Thispublishing servieletsuserswriteapost

and attah to thepost anykindof les. It will usean externalstorageservie

tostoretheattahedles.

Thisservieisonstrainedbytwoqualityrequirementsrelatedtotheavail-

abilityandthereliability.Bothrequirementsareinludedinanaeptablequal-

itylevel forstandardusers.Therst requirementonstrains theavailabilityof

thepublishingserviein avaluethatshouldbeequalorgreaterthan98%.The

seondrequirementonstrainsthereliabilityinafaultresponselikelihoodvalue

equalorlessthan0.05(seegure4).

Severalproblemsanaettherequirementsfullment,thispaperidentied

onlyfourofthem.Therearetwofailuresthatanaetthenormaloperationof

theservie:theused storageserviebeomesunavailable;andtheusedstorage

servie givestoomany fault responses, beoming unreliable. Bothfailures an

auseothertwo:theunavailabilityandtheunreliabilityofthepublishingservie

respetively, and an threaten the quality requirements satisfation desribed

before(seegure4).

Fig.4.FailuresandQualityGoalsdiagram

Bothfailures in thepublishing servie ativatetwofailure ases (seegure

4). The Frequent Error Responses Submitting Posts desribes that randomly

thesystemanswerswithanunexpetederrorwhenthepublisherissubmittinga

newpost.ThePublishingServieUnavailabledesribesthatwhenthepublisher

aessestotheserviethesystemshowsthemessageThePublishingServieis

have the apability of deteting suh problems and automatially re-bind to

analternativeStorage Servie,followingagivenseletionriteriabasedonthe

quality levels oered by them. It assumes that exist a Registry that have a

full desriptionof StorageServiesthat anbeusedbythePublishing Servie.

Thesystemshould trytoseletthe serviethatoersthehighestqualitylevel

related to the availability and the reliability. But, when the seletion is not

lear(beauseexistanalternativewiththehighestavailabilitybut withoutthe

highest reliability, orotherwise)the system should follow the poliy desribed

byanadministrator.Inordertoavoidthere-seletionofaproblematiservie

the system will manage a blak list of them and will use the list to subtrat

problemati serviesfrom thelistgivenbytheRegistry.Thisapabilityshould

bepresentintheprovidersoftwareagentsofthePublishingServie.

Fig.5.ThePublishingServieRe-bindingCapabilitytreatingproblemsdiagram

Theapabilitywill try to reover the systemfrom theStorage Servie Un-

available andStorageServieBeomesUnreliablefailures.Italsodeativates the

failure asesaused indiretlybythe failures.These intentions arerelated to

theself-healingaspetoftheapability(seegure5).

Sinetheapabilityseletstheserviewhih oersthehighestqualitylevel

inavailabilityandreliabilityattributes,itreduesthelikelihoodofthePublish-

ingServieUnavailable andPublishing ServieBeomesUnreliable failures(see

gure5). Therefore,theapabilityontributestothesatisfationof bothqual-

ityrequirements:thehighavailabilityandthehighreliabilityoftheservie(see

gure6).

Theapabilityrequiresthedevelopmentofsomeuseasesinordertooperate

properly(seegure6). ThealternativestorageserviesarefoundinaRegistry,

thus the apability needsthe loation of this registryas input. TheCongure

RegistryLoation useasedesribesthefuntionalityofonguringtheloation

oftheregistry.Theapabilitywillmanageablaklistofservies,butproblemati

servies ould be healedand the administrator ould need removeit from the

blaklist atrun-time.Buteven,theadministratorouldneedadd problemati

ityrequirementsdiagram

Blak List useasedesribesthisfuntionality. Alsotheapabilityrequiresthe

CongureQoSCriteriafor Seletion useasetolettheadministratorongure

whatQoSriteriatheapabilityshould takeintoaounttoseletalternatives;

and also requires theView Re-binding Result Report use asewhih desribes

thefuntionalityof viewing,bythe administrator,the re-bindinglogsthat are

produedat run-time.

Thisapabilityis detailed by aself-managementproessmodel(gure 7,8

and 9). Thereare twomonitoringativities: Storage Servie FaultsMonitoring

and Storage Servie Availability Monitoring. The formermonitors thenumber

offaults inaperiodoftime inorderto updatetheurrentvalueof thequality

metrifault-response-likelihoodoftheStorageServie;andthelattermonitors

the urrent operational status (available, unavailable) of the Storage Servie

too. These monitoring ativities generate reports whih are onsumed by the

AnalysisOftheNeedstoRebind.TheStorageServieFaultsMonitoring ativity

will monitor theusing of the StorageServie in order to ath faults, and the

StorageServie Availability Monitoring ouldmonitortheusing,butalsoould

do themonitoringeither diretly askingto the Storage Servie, orsubsribing

itselfto someheartbeatsignal.

The Analysis Of the Needs to Rebind ativity deides if the rebinding is

neededornot.Thedeisionismadeusingthemonitoringreportsandthequal-

ityrequirementsof thestorageservie. Ifthefault-response-likelihood metri

of the Storage Servie is equalorgreater than 0.05 thereliabilityof the Pub-

lishingServiewillderease,andiftheStorageServiebeomesunavailablethe

Publishing Servie will getinto failure. Therefore, in these asestherebinding

will be needed. However,the Storage Servie ould havea notiationsystem

thatinformtoitslientsoftemporaryunavailabilityforadministrationpurpose.

Then the analysis ativity ould take into aountsuh informations to make

abetterdeisionalulatinghowtheestimatedperiodoftimein unavailability

an aet the quality level of the Publishing Servie. Another ase where the

rebindingisnotneedediswhenthereisalreadyarebindinginexeution.

Ifthe rebindingis needed, thenthe proess getinto aplan seletionphase

andidates or not. There are twoplans: update the blak list, blok the using

of the publishing servie, and do nothing,beause there are no andidates; or

updatetheblaklistandrebind,beausethereisat lessoneandidate.

Fig.8.ThePublishingServieRe-bindingProess(2)diagram

The gure8 shows the diagram of the part orresponding to the seletion

andexeutionoftheplanwhentherearenoandidates.Therstelementsinthe

owistheplanseletionguardnodethat ontainstheOCLspeiationwhih

onstrains the opy of the token to the outgoing edge. In this ase, the on-

straintis:iftheandidateslistisemptythenletsthetokenontinue.Therest

ofelementsintheowaretheupdatingoftheblaklistwith theurrentstor-

ageservie,theunavailabilitynotiationandtheunregisteringtoisolateitself

fromtherestofthesystem.TheFinalNodeindiatesthattheself-management

proesswillstopompletely,eventhemonitoringativities.

Thegure9showsthediagramofthepartorrespondingtotheseletionand

exeutionoftheplanwhenthereisatlessoneandidate.Asbefore,therstele-

mentsintheowistheplanseletionguardthatonstrainstheopyofthetoken

totheoutgoingedge.Whenthereisatlessoneandidateinthelistthetokenis

opyandtheowontinuethroughtherestofplanstepativities.Theplanrst

updatestheblaklistandnotiesatemporaryunavailabilityforadministration

purpose. Afterthat,seletsoneservieaordingtothegivenseletionriteria

and rebinds the publishing servie to it. Finally, noties the availability. The