The Publish/Subscribe (pub/sub) model has been widely applied in a variety of application scenarios which demand loose-coupling and asynchronous communication between a large number of information sources and sinks. In this model, clients are granted the flexibility to specify their interests at a high level and rely on the pub/sub middleware for delivery of their publications of interest. This increased flexibility and ease of use on the client side results in substantial complexity on part of the pub/sub middleware implementation. Furthermore, for several reasons including improved scalability, availability and avoiding a single point of failure, the pub/sub middleware is commonly composed of a set of collaborating message routers, a.k.a. brokers. The distributed nature of this design further introduces new challenges in ensuring end-to-end reliability as well as efficiency of operation. These challenges are largely unique to the pub/sub model and hence absent in both point-to-point or multicast protocols. This thesis develops solutions that ensure the dependable operation of the pub/sub system by exploiting the notion of overlay neighborhoods in a formal manner. More specifically, brokers maintain information about their neighbors within a configurable distance in the pub/sub overlay and exploit this knowledge to construct alternative forwarding paths or make smart forwarding decisions that improves efficiency, bandwidth utilization and delivery delay, all at the same time. Furthermore, in the face of failures overlay neighborhoods enable fast reconstruction of forwarding paths in the system without compromising its reliability and availability. Finally, as an added benefit of overlay neighborhoods, this thesis develops large-scale algorithms that bring the advantages of the pub/sub model to the domain of file sharing and bulk content dissemination applications. Experimental evaluation results with deployments as large as 1000 nodes illustrate that the pub/sub system scales well and outperforms the traditional BitTorrent protocol in terms of content dissemination delay.

The Publish/Subscribe (pub/sub) model has been widely applied in a variety of application scenarios which demand loose-coupling and asynchronous communication between a large number of information sources and sinks. In this model, clients are granted the flexibility to specify their interests at a high level and rely on the pub/sub middleware for delivery of their publications of interest. This increased flexibility and ease of use on the client side results in substantial complexity on part of the pub/sub middleware implementation. Furthermore, for several reasons including improved scalability, availability and avoiding a single point of failure, the pub/sub middleware is commonly composed of a set of collaborating message routers, a.k.a. brokers. The distributed nature of this design further introduces new challenges in ensuring end-to-end reliability as well as efficiency of operation. These challenges are largely unique to the pub/sub model and hence absent in both point-to-point or multicast protocols. This thesis develops solutions that ensure the dependable operation of the pub/sub system by exploiting the notion of overlay neighborhoods in a formal manner. More specifically, brokers maintain information about their neighbors within a configurable distance in the pub/sub overlay and exploit this knowledge to construct alternative forwarding paths or make smart forwarding decisions that improves efficiency, bandwidth utilization and delivery delay, all at the same time. Furthermore, in the face of failures overlay neighborhoods enable fast reconstruction of forwarding paths in the system without compromising its reliability and availability. Finally, as an added benefit of overlay neighborhoods, this thesis develops large-scale algorithms that bring the advantages of the pub/sub model to the domain of file sharing and bulk content dissemination applications. Experimental evaluation results with deployments as large as 1000 nodes illustrate that the pub/sub system scales well and outperforms the traditional BitTorrent protocol in terms of content dissemination delay.

Les réseaux mobiles opportunistes (ou OppNets, pour Opportunistic Networks) constituent une solution séduisante pour compléter les réseaux fixes d’infrastructure, voire compenser leur absence dans des zones sinistrées ou défavorisées. Les recherches menées ces dernières années ont principalement visé à permettre les transmissions dans les OppNets, mais ceci ne peut être qu’un premier pas vers une réelle exploitation de tels environnements contraints. L’informatique opportuniste (Opportunistic Computing) dépasse le cadre des seules transmissions, et introduit un nouveau paradigme d’exécution de tâches collaboratives dans de tels environnements. Dans ce domaine qu’est l’informatique opportuniste, la conception, la mise en œuvre et le déploiement d’applications distribuées sont des objectifs majeurs. Une application pour OppNet doit pouvoir fonctionner et assurer un niveau de service satisfaisant, tout en supportant les diverses contraintes propres aux OppNets, telles qu’une connectivité fluctuante, un partitionnement chronique du réseau, de longs délais de transmissions, de fréquents échecs de transmission, et des équipements hétérogènes offrant des ressources limitées. La complexité et le coût du développement d’applications pour OppNets peuvent être réduits de manière significative en utilisant des modèles de programmation appropriés. De tels modèles peuvent être fournis par des systèmes intergiciels capables de supporter de manière transparente les contraintes évoquées plus haut. Le travail rapporté dans ce mémoire a porté sur l’étude des contraintes inhérentes aux OppNets, et sur la proposition de solutions appropriées. Parmi les modèles de programmation usuels, certains ont été identifiés comme pouvant être utilisés dans le cadre des OppNets. Sur la base de ces divers modèles de programmation, des systèmes intergiciels opportunistes ont été mis en oeuvre. Ces systèmes supportent respectivement le modèle de messagerie distribuée (sur la base de files d’attentes et de "topics"), le modèle du tuple-space, et la résolution de consensus. Des implémentations complètes ont été réalisées, et le code source est distribué sous licence GPL (GNU General Public License). Ces systèmes ont été évalués par le biais d’expérimentations menées en conditions réelles et par simulation. / Opportunistic networks (OppNets) constitute an appealing solution to complement fixed network infrastructures –or make up for the lack thereof– in challenged areas. Researches in the last few years have mostly addressed the problem of supporting networking in OppNets, yet this can only be a first step towards getting real benefit from these networks. Opportunistic computing goes beyond the concept of opportunistic networking, and provides a new paradigm to enable collaborative computing tasks in such environments. In the realm of opportunistic computing, properly designing, implementing and deploying distributed applications are important tasks. An OppNet-dedicated application must be able to operate and maintain an acceptable level of service while addressing the many problems that can occur in these networks, such as disconnections, partitioning, long transmission delays, transmission failures, resource constraints, frequent changes in topology, and heterogeneous devices. Much of the complexity and cost of building OppNet-dedicated applications can be alleviated by the use of high-level programming models. Such models can be supported by middleware systems capable of transparently addressing all the above-mentioned problems. The work reported in this dissertation focused on providing insight into the fundamental problems posed by OppNets, so as to analyze and solve the problems faced by application developers while dealing with these environments. The research focused on identifying well-known high-level programming models that can be satisfactorily implemented for OppNets, and that can prove useful for application developers. In order to demonstrate the feasibility of application development for OppNets, while assessing the benefits brought about by carefully designed middleware systems, a couple of such systems have been designed, implemented, and evaluated as part of this work. These middleware systems respectively support distributed messaging (through message queues and topics), the tuple-space model, and consensus solving in OppNets. They are supplemented with fully-functional implementations, that can be used in real settings, and that are all distributed under the terms of the GNU General Public License (GPL). Real-life experiments and simulations have been realized so as to evaluate the effectiveness and efficiency of these systems in real conditions.

No description available.

Made available in DSpace on 2015-03-03T15:47:44Z (GMT). No. of bitstreams: 1 TassiaAVF.pdf: 1514001 bytes, checksum: a6dbf5832432db823e99d729ccf7fa2b (MD5) Previous issue date: 2009-02-20 / Nowadays, there are many aspect-oriented middleware implementations that take advantage of the modularity provided by the aspect oriented paradigm. Although the works always present an assessment of the middleware according to some quality attribute, there is not a specific set of metrics to assess them in a comprehensive way, following various quality attributes. This work aims to propose a suite of metrics for the assessment of aspect-oriented middleware systems at different development stages: design, refactoring, implementation and runtime. The work presents the metrics and how they are applied at each development stage. The suite is composed of metrics associated to static properties (modularity, maintainability, reusability, exibility, complexity, stability, and size) and dynamic properties (performance and memory consumption). Such metrics are based on existing assessment approaches of object-oriented and aspect-oriented systems. The proposed metrics are used in the context of OiL (Orb in Lua), a middleware based on CORBA and implemented in Lua, and AO-OiL, the refactoring of OIL that follows a reference architecture for aspect-oriented middleware systems. The case study performed in OiL and AO-OiL is a system for monitoring of oil wells. This work also presents the CoMeTA-Lua tool to automate the collection of coupling and size metrics in Lua source code / Atualmente, ha diversas implementa??es de sistemas de middleware orientado a aspectos que aproveitam o suporte a modulariza??o do paradigma de orienta??o a aspectos. Apesar desses trabalhos sempre apresentarem uma avalia??o do middleware de acordo com algum atributo de qualidade, n?o ha ainda um conjunto de metricas especificamente de nidas para avali?-los de forma abrangente, seguindo v?rios atributos de qualidade. Este trabalho tem como objetivo propor um conjunto de metricas para avalia??o de sistemas de middleware orientado a aspectos em diferentes fases de desenvolvimento: design, refatora??o, implementa??o e execu??o. O trabalho apresenta as metricas e como elas s?o aplicadas em cada uma das fases de desenvolvimento. O conjunto e composto por metricas associadas a propriedades estaticas (modularidade, manutenibilidade, reusabilidade, exibilidade, complexidade, estabilidade e tamanho) e din?micas (desempenho e consumo de mem oria). Tais metricas s?o baseadas em abordagens existentes de avalia??o de sistemas orientados a aspectos e a objetos. As metricas propostas s?o utilizadas no contexto do OiL (Orb in Lua), um middleware baseado em CORBA e implementado em Lua, e AO-OiL, uma refatora??o do OIL que segue uma arquitetura de refer?ncia para sistemas de middleware orientados a aspectos. O estudo de caso executado no OiL e no AO-OiL e um sistema de monitoramento de po?os de petroleo. Esse trabalho apresenta ainda a ferramenta CoMeTA-Lua para automatizar a coleta das metricas de tamanho e acoplamento em codigo-fonte Lua