16<strong>CoreGRID</strong>InstitutesComponentsand services arebecoming evercloser and commonperspectives maybe exploited.Marco DaneluttoLeader <strong>of</strong> the Programming Model Institute,Università di PisaPROGRAMMING MODEL▼Programming Model Institute: making the programming<strong>of</strong> Grid infrastructures as simple, transparentand efficient as possible.While Grid tool technology is rapidlyconverging towards Services andService Oriented Architectures,advanced programming environmentsare still needed that may simplifythe Grid programmer’s task andshorten the application design-toproductiontime while outper<strong>for</strong>mingapplications written using plain,low-level Grid middleware tools.The Programming Model Instituteaims to investigate new ProgrammingModels capable <strong>of</strong> fully supportingefficient Grid programming, andto raise the level <strong>of</strong> abstractionprovided to the Grid applicationprogrammer as recommended inthe NGG documents.The Programming Model Institutebrings together 13 full partnersand 2 associate partners, with atotal <strong>of</strong> more that 45 researchersand about 30 PhD students.Altogether, these researchershave contributed to an extensiveprogramme <strong>of</strong> short visits, to theresearcher exchange programme,to <strong>CoreGRID</strong> fellowships <strong>of</strong>teninvolving other <strong>CoreGRID</strong> Institutes,and have published a significantnumber <strong>of</strong> joint research papersin international journals andpresented them at conferences.The Programming Model Instituteinvestigates suitable Grid programmingmodels and techniques based oncomponent technology. Componentsand services are becoming evercloser and common perspectivesmay be exploited in the design<strong>of</strong> advanced component-basedprogramming frameworks supportingfull interoperability with services.Within the Institute, the GCM (GridComponent Model) has been definedand is currently being refined. A spin<strong>of</strong>fSTREP project has been started(Jun06 – Nov08) whose main goal isthe implementation <strong>of</strong> an open sourceprototype implementation <strong>of</strong> GCM.GCM builds on top <strong>of</strong> the Fractalcomponent model and <strong>of</strong>fers tothe Grid application programmerhierarchical component composition,autonomic controllers, advanced andcollective component communicationpatterns, as well as an XML based ADL.Within the Programming ModelInstitute, ISTI/CNR, HLRS, IC, INRIA,QUB, WWU Muenster, UCHILE, UNI-PASSAU, UNIPI, EIA/FR, UOW, UPC,VUA, UoS and UoL have activelyparticipated in the activities <strong>of</strong>the three main research topicscovered by the Institute: “BasicProgramming Models”, investigatingthe Programming Models suitable<strong>for</strong> programming primitive GCMcomponents; “Componentsand hierarchical composition”,investigating the basic features <strong>of</strong>GCM; and “Advanced ProgrammingModels”, investigating the possibility<strong>of</strong> building advanced ProgrammingModels on top <strong>of</strong> GCM.In 2007, the activities in theProgramming Model Institute havebeen mainly centred on a couple <strong>of</strong><strong>CoreGRID</strong> events (the workshop coorganisedby the Architectural Issues:Scalability, Dependability, AdaptabilityInstitute, by the Programming ModelInstitute and by the Grid Systems,Tools and Environments Institute inHeraklion in June; and the <strong>CoreGRID</strong>Symposium in Rennes in August)and on a quite large number <strong>of</strong>
17short visits and Research ExchangeProgramme events involving Instituteresearchers from different partners. All<strong>of</strong> the Institute partners participatedin the Heraklion workshop and most<strong>of</strong> them also participated in the<strong>CoreGRID</strong> Symposium, presentingseveral important research resultsand discussing with partners bothexisting and new research topicsin the Institute. Overall, severalnotable results have been achievedin 2007, concerning GCM mechanismand structure refinement andrelated to GCM-based advancedprogramming models and GCMverification/modelling techniques.The advances in Programming ModelInstitute activities contributed to the<strong>CoreGRID</strong> goal <strong>of</strong> providing solidmethodological and technologicalfoundations <strong>for</strong> Grid and peer-to-peer.The GCM model is being adopted andused in several other Institutes withinthe NoE. GCM-related results have beenpresented and appreciated in a number<strong>of</strong> international contexts, includingSuperComputing07 and EuroPar07conferences, OGF (Manchester07) andthe <strong>CoreGRID</strong> Scientific Advisory Board.The results achieved during 2007contributed to the well-focusedobjective <strong>of</strong> the Programming ModelInstitute that may be summarised asbeing able to deliver a definition <strong>of</strong> acomponent programming model thatcan be usefully exploited to design,implement and run Grid applicationswhile efficiently and transparentlyfacing the new challenges in terms<strong>of</strong> programmability, interoperability,code reuse and efficiency that derivefrom the peculiar Grid features suchas heterogeneity and dynamicity.During 2007, partners <strong>of</strong> theProgramming Model Institute:• Demonstrated autonomicmanagers taking complete care<strong>of</strong> per<strong>for</strong>mance related (nonfunctional)aspects <strong>of</strong> notablecomposite components.• Refined several <strong>for</strong>mal andsemi-<strong>for</strong>mal techniques thatallow reasoning about GCMprogrammes and pro<strong>of</strong> <strong>of</strong>properties <strong>of</strong> these programmes.• Achieved further results relatedto the usage <strong>of</strong> “data sharing”and “data flow stream” portsamong GMC components.• Demonstrated the possibility <strong>of</strong>combining classical parallelisationtechniques (such as looptrans<strong>for</strong>mations) with structuredprogramming environments that caneventually be built on top <strong>of</strong> GCM.• Started to demonstrateinteroperability <strong>of</strong> GCMconcepts with s<strong>of</strong>tware servicesand SOA in general.• Finally, during the GRIDS@work event in Beijing (end<strong>of</strong> October) the preliminaryprototype developed byGridCOMP demonstrated theability to efficiently deploy andrun component programmeson hundreds <strong>of</strong> distributedGrid resources. This is notdirectly a <strong>CoreGRID</strong> result,although most GridCOMPpartners are also partners <strong>of</strong> theProgramming Model Institute.These results have been achieved bypartners <strong>of</strong> the Institute cooperatingvia all the mechanisms providedby <strong>CoreGRID</strong> (short visits, REPs,fellowships) and clearly contributeto the Programming Model Instituteobjective. In turn, the main objectivein the Programming Model Institutecontributes to the overall <strong>CoreGRID</strong>roadmap both in the field <strong>of</strong> toolsand environments <strong>for</strong> efficient Gridprogramming and in the field <strong>of</strong>methodologies <strong>for</strong> the development<strong>of</strong> suitable Grid s<strong>of</strong>tware. The GCMis in fact being used as a referencemodel in the Grid Systems, Toolsand Environments Institute. GCM isalso being considered as a suitableprogramming model to developseveral kinds <strong>of</strong> (system) applicationsin other Institutes in the NoE.Through participation in GridCOMP,several industrial partners havebeen involved in the GCM designand development process. TheProgramming Model Institutebuilds on these experiences insuch a way that the final GCMcould be immediately used intypical industrial applicationswithout any further tuning.Also, a standardisation process hasbeen initiated through ETSI that willeventually result in a complete GCMstandard. In 2007, the first steps havebeen per<strong>for</strong>med and the GCM ADLis close to being an ETSI standard.We expect that Programming ModelInstitute activities, and thoseconcerned with GCM in particular havea tw<strong>of</strong>old impact on industry: on onehand, the availability <strong>of</strong> a complete,advanced, component-based Gridprogramming environment will allowindustry to cut the development timeand cost <strong>of</strong> Grid applications. On theother hand, complete interoperabilitywith the services framework (one <strong>of</strong>the objectives <strong>of</strong> the ProgrammingModel Institute yet to be achieved)will allow industry to reuse andexploit a large base <strong>of</strong> alreadydeveloped s<strong>of</strong>tware services in a moreconvenient programming framework.There are still several openchallenges to be tackled in theProgramming Model Institute:• Complete integration <strong>of</strong> GCM withthe s<strong>of</strong>tware service framework.Although Institute partners alreadypartially demonstrated the feasibility<strong>of</strong> porting GCM concepts on top <strong>of</strong>the Service Component Architecture,much work has to be done toguarantee that the advancedconcepts introduced in GCM (suchas collective communicationpatterns or autonomic managers)can be migrated to thes<strong>of</strong>tware service scenario.• Implementation <strong>of</strong> advancedprogramming models on top <strong>of</strong> GCM.There are several kinds <strong>of</strong> advancedprogramming models considered t<strong>of</strong>urther raise the level <strong>of</strong> abstractionpresented to Grid applicationprogrammes, ranging from skeletonbasedstructured programmingparadigms to component paradigmscombining spatial (componentlike)and temporal (à la workflow)composition <strong>of</strong> components.• Development <strong>of</strong> a complete set <strong>of</strong><strong>for</strong>mal (or semi-<strong>for</strong>mal) tools that canbe used to reason about and proveproperties <strong>of</strong> GCM programmes.Partners in the Programming ModelInstitute recognise these are significantchallenges still to be addressed. Duringthe Institute meeting in Heraklion(June) partners also agreed that theactivities needed to tackle thesechallenges must be continued after theend <strong>of</strong> the NoE, either on a volunteerbasis or within possible furtherinitiatives and projects following the(<strong>for</strong>mal) end <strong>of</strong> the <strong>CoreGRID</strong> NoE.<strong>Annual</strong>Report2007