Mes travaux de recherche portent sur les modèles de programmation distribuée, principalement par objets et composants. Dans ce domaine, j'ai travaillé à fournir des outils facilitant la programmation d'applications distribuées à large échelle et vérifiant la correction de leur comportement. Pour faciliter la programmation d'applications distribuées je me suis intéressé à la mise au point de langages avec un fort niveau d'abstraction: objets actifs, squelettes algorithmiques, composants. Afin de vérifier la correction du comportement d'une application j'ai collaboré à la mise au point d'outils de spécification et de vérification comportementales d'applications distribuées. Mes travaux ont pour but de fournir un modèle formel des langages de programmations, des bibliothèques, et des environnements d'exécution fournies au programmeur afin de garantir un comportement sûr des applications distribuées. Ma thèse m'a permis de mettre au point le calcul ASP modélisant lecomportement des objets actifs et des futurs. Depuis, nous avons créé une version fonctionnelle de ce calcul que nous avons modélisé en Isabelle/HOL. Aussi j'ai fortement contribué à la définition d'un modèle à composants distribués -- le GCM (Grid Component model)--, à sa formalisation et à son utilisation pour programmer des composants adaptables ou autonomes. Enfin, j'ai contribué à la spécification et la vérification comportementale des programmes utilisant des objets actifs et des composants afin de garantir la sûreté de leur exécution. Actuellement, nous travaillons à la fois à une extension multi-threadée du modèle à objets actifs mieux adaptée aux machines multi-coeurs, et à l'utilisation de méthodes formelles pour mettre au point et prouver la correction d'un algorithme de diffusion pour réseau pair-à-pair de type CAN (Content Adressable Network). Ce manuscrit fournit une vue d'ensemble de tous ces travaux.

Orientador: Luiz Gustavo Bizarro Mirisola / Dissertação (mestrado) - Universidade Federal do ABC. Programa de Pós Graduação em Ciência da Computação, 2013

Peer-to-peer applications enable a group of users to create a communications framework from scratch without the need of a central service provider. This is achievable via the aggregation of resources each one of them provide, creating a completely distributed collaborative environment based in a flat hierarchy of users, without the need for centralization. Usually, peer-to-peer applications are conceptualized as a global network, without any kind of logical segmentation or segregation as far as resource availability is concerned. At every model, any peer may access any resource available within the network just by being able to reach the peer that provides such resource. Although having a unique huge open network may be desirable for some applications, there are cases in which it might be interesting to create different, but not necessarily disjoint, groups of peers operating under the same global peer-topeer network. In order for peer groups to be able to operate effectively in a global peer-to-peer network, additional security services must be provided. These mechanisms should allow peers to be able to prove group membership to other members of the group, so they can be granted access to group resources, as well as ensuring that resource discovery and message exchange between peer group members remain secure. A group may need to limit membership for various reasons, such as ensuring privacy, anonymity or enforcing that peer group members are up to some specific parameter (data shared, performance, computing power, etc.) The goals of this PhD. thesis are twofold, the reason being the fact that securing a peer group can be divided at two distinct, but interrelated, layers: • Enabling effective group membership, starting from the process by which any peer becomes part of a peer group and then, following, the mechanisms by which such peer may prove its membership to other group members for the rest of the membership's lifecycle (peer group access control). • Providing a secure environment for standard operations within a peer group, which functions once any peer's membership to the group has already been established. Typical operations at this layer are those of resource location and retrieval, or messaging. In order to achieve the former goal, basic group membership and access control scenarios are categorized and formalized as part of the research work in order to assess which are the current challenges. From this study, we present a generic model proposal that fulfils the objectives of autonomy, keeps a pure peer-to-peer model and the possibility to be used in different peer-to-peer frameworks. The later goal focuses in secure mechanisms in order to provide basic security services to both resource discovery and message exchange. However, in contrast with group membership models, where a generic approach is feasible, peer group operation security is intimately tied to each specific peer-to-peer framework, since each one specifies resource location and messaging primitives in a different manner. For that reason, a specific one has been chosen for the research work: JXTA. Such election is due to the fact that JXTA's architecture is entirely based on the concept of peer groups, since it was the one to first define the concept of peer group, providing an excellent testbed for peer group research

Beaucoup de travaux de recherche autour des systèmes d'intégration de données se sont concentrés sur les problémes liés aux schémas, aux langages et au traitement de requÍtes. Dans cette thèse, nous nous intéressons à la problématique de construction de tels systèmes. Nous appliquons les concepts architecturaux des systèmes répartis, notamment de canevas et de composant logiciel, pour proposer un intergiciel d'intégration de données offrant quatre niveaux d'adaptabilité. Au niveau architectural, nous définissons le canevas de domaine de données qui est le composant central de l'intégration. Un système d'intégration de données est défini par la composition de domaines de données. Au niveau modèle de données, nous adoptons un modèle hybride doté d'un typage flexible, associé au langage d'expression des vues de domaines, qui permet de prendre en compte l'hétérogénéité structurelle des sources (ou domaines) de données à intégrer. Au niveau langage de requêtes, nous proposons un canevas d'expression qui permet de décrire les requêtes à la fois pour l'intégration et pour l'interrogation de domaines, indépendamment d'une syntaxe particulière. Au niveau optimisation, nous proposons un canevas d'optimisation de requêtes adaptable, dans le sens ou il permet de construire et de supporter diverses stratÈgies de recherche. Le canevas d'expression et le canevas d'optimisation de requÍte ont été implantés dans le cadre du consortium ObjectWeb. Ces implantations ont été utilisÈes dans la mise en oeuvre de deux standards de persistance d'objets Java, à savoir JDO (Java Data Object) et EJB-CMP (Container Managed Persistance). Dans le contexte de recherche de composants pour des besoins d'administration des systèmes à large échelle, nous avons proposé une utilisation de notre intergiciel pour proposer un service de requête qui permette de rechercher des composants dans un gisement et dans des systèmes en cours d'exécution, par introspection associative.

Made available in DSpace on 2014-12-17T15:47:50Z (GMT). No. of bitstreams: 1 JoseDSS.pdf: 1294518 bytes, checksum: aa9617f35bc3b6c43b4ee10931265d07 (MD5) Previous issue date: 2009-02-26 / Middleware platforms have been widely used as an underlying infrastructure to the development of distributed applications. They provide distribution and heterogeneity transparency and a set of services that ease the construction of distributed applications. Nowadays, the middlewares accommodate an increasing variety of requirements to satisfy distinct application domains. This broad range of application requirements increases the complexity of the middleware, due to the introduction of many cross-cutting concerns in the architecture, which are not properly modularized by traditional programming techniques, resulting in a tangling and spread of theses concerns in the middleware code. The presence of these cross-cutting concerns limits the middleware scalability and aspect-oriented paradigm has been used successfully to improve the modularity, extensibility and customization capabilities of middleware. This work presents AO-OiL, an aspect-oriented (AO) middleware architecture, based on the AO middleware reference architecture. This middleware follows the philosophy that the middleware functionalities must be driven by the application requirements. AO-OiL consists in an AO refactoring of the OiL (Orb in Lua) middleware in order to separate basic and crosscutting concerns. The proposed architecture was implemented in Lua and RE-AspectLua. To evaluate the refactoring impact in the middleware architecture, this paper presents a comparative analysis of performance between AO-OiL and OiL / As plataformas de middlewares t?m sido amplamente utilizadas como infra-estrutura subjacente para o desenvolvimento de sistemas distribu?dos. Elas fornecem transpar?ncia de localiza??o e de heterogeneidade e um conjunto de servi?os que facilitam a constru??o de aplica??es distribu?das. Atualmente, os middlewares acomodam uma variedade crescente de requisitos a fim de atender aos mais variados dom?nios de aplica??o. Essa ampla variedade de requisitos procovou um aumento na complexidade dos middleware, devido a introdu??o de v?rios conceitos transversais na arquitetura. Tais conceitos n?o s?o modularizados adequadamente pelas t?cnicas de programa??o tradicionais, resultando no espalhamento e entrela?amento desses conceitos no c?digo do middleware. A presen?a de conceitos transversais limitam a escalabilidade do middleware. O paradigma orientado a aspectos tem sido utilizado com sucesso para melhorar a extensibilidade, a modularidade e a capacidade de personaliza??o das plataformas de middleware atrav?s da separa??o dos conceitos transversais do c?digo base da middleware. Este trabalho apresenta o AO-OiL, uma arquitetura de um middleware orientado a aspectos (OA) dinamicamente adapt?vel, baseado na arquitetura de refer?ncia para middlewares OA. O AO-OiL consiste em uma refatora??o orientada a aspectos do middleware OiL (Orb in Lua) para separa??o de conceitos b?sicos e conceitos transversais e segue a filosofia de que as funcionalidades do middleware devem ser ditadas pelos requisitos da aplica??o. A arquitetura proposta foi implementada em Lua e RE-AspectLua. Para avaliar o impacto da refatora??o na arquitetura, esse trabalho apresenta uma an?lise comparativa de desempenho entre o AO-OiL e o OiL

No description available.