Técnicas e ferramentas (middlewares para desenvolvimento de aplicações distribuídas são utilizadas no caso dos recursos de um único computador não serem suficientes, ou para atender requisitos de qualidade de serviço, ou mesmo para abranger uma maior quantidade de usuários. A Realidade Virtual é uma das áreas da computação a sua complexidade, se beneficia de tais técnicas e middlewares, seja para proporcionar ambientes mais imersivos e/ou possibilitar a interação de múltiplos usuários. A Realidade Virtual engloba uma ampla diversidade de aplicações, incluindo infraestruturas de alta imersão, jogos multi-jogadores, e simuladores de voo, e cada classe de aplicação têm demandas específicas de distribuição, como escalabilidade, ou requisitos de tempo-real, entre outros. Devido à complexidade e a essa diversidade de aplicações de realidade virtual, as alternativas de middlewares disponíveis atendem requisitos específicos de uma única classe, ou, as genéricas, não fornecem modelos de objetos, comunicação ou execução favoráveis às aplicações de realidade virtual. Esse trabalho tem o objetivo de propor um novo modelo de objetos e arquitetura de um middleware para desenvolvimento de aplicações distribuídas de realidade virtual suficientemente flexível para atender diversas classes de aplicações. O modelo de objetos proposto une os modelos publicador/assinante e requisição/ resposta em uma única entidade similar aos objetos do paradigma de programação orientada a objetos, que, além de empregar conceitos intuitivos para desenvolvedores, possibilita o desacoplamento lógico entre os processos, simplificando o projeto e a implementação das aplicações distribuídas. O principal diferencial desta abordagem é a possibilidade de distribuir a implementação de cada membro dos objetos em diferentes processos. / Techniques and middlewares for distributed application development are used in cases the resources available in a single computer are not enough for the demanded complexity, to meet the required quality of services, or even to cover a greater number of users. The Virtual Reality is one of the computer fields that, because of its complexity, benefits from such techniques and tools, either to provide more immersive virtual environments, or to enable multi-user interaction. The Virtual Reality covers a wide range of applications, including high immersion infrastructures, multiplayer games, and full mission flight simulators. Each application class has specifics distribution demands, such as high scalability, or real-time requirements. Because of complexity and the diversity of virtual reality applications, the distribution solutions available meet the specific requirements of a single class, or the generic ones do not provide object, communication, or execution favorable for virtual reality applications. This work proposes a new model of objects and architecture of a middleware for distributed virtual reality applications that is flexible enough to suit several classes of applications. The proposed object model joins the publisher/subscriber and request/response models to a single logic entity similar to the objects of the object-oriented programming paradigm. Besides being an intuitive concept for developers, this entity enable logical decoupling between processes, simplifying the design and implementation of distributed applications. The main feature of this approach is the possibility to distribute the implementation of each member of the objects in different processes.

D’ici 2020, l’univers numérique atteindra 44 zettaoctets puisqu’il double tous les deux ans. Les données se présentent sous les formes les plus diverses et proviennent de sources géographiquement dispersées. L’explosion de données crée un besoin sans précédent en terme de stockage et de traitement de données, mais aussi en terme de logiciels de traitement de données capables d’exploiter au mieux ces ressources informatiques. Ces applications à grande échelle prennent souvent la forme de workflows qui aident à définir les dépendances de données entre leurs différents composants. De plus en plus de workflows scientifiques sont exécutés sur des clouds car ils constituent une alternative rentable pour le calcul intensif. Parfois, les workflows doivent être répartis sur plusieurs data centers. Soit parce qu’ils dépassent la capacité d’un site unique en raison de leurs énormes besoins de stockage et de calcul, soit car les données qu’ils traitent sont dispersées dans différents endroits. L’exécution de workflows multisite entraîne plusieurs problèmes, pour lesquels peu de solutions ont été développées : il n’existe pas de système de fichiers commun pour le transfert de données, les latences inter-sites sont élevées et la gestion centralisée devient un goulet d’étranglement. Cette thèse présente trois contributions qui visent à réduire l’écart entre les exécutions de workflows sur un seul site ou plusieurs data centers. Tout d’abord, nous présentons plusieurs stratégies pour le soutien efficace de l’exécution des workflows sur des clouds multisite en réduisant le coût des opérations de métadonnées. Ensuite, nous expliquons comment la manipulation sélective des métadonnées, classées par fréquence d’accès, améliore la performance des workflows dans un environnement multisite. Enfin, nous examinons une approche différente pour optimiser l’exécution de workflows sur le cloud en étudiant les paramètres d’exécution pour modéliser le passage élastique à l’échelle. / By 2020, the digital universe is expected to reach 44 zettabytes, as it is doubling every two years. Data come in the most diverse shapes and from the most geographically dispersed sources ever. The data explosion calls for applications capable of highlyscalable, distributed computation, and for infrastructures with massive storage and processing power to support them. These large-scale applications are often expressed as workflows that help defining data dependencies between their different components. More and more scientific workflows are executed on clouds, for they are a cost-effective alternative for intensive computing. Sometimes, workflows must be executed across multiple geodistributed cloud datacenters. It is either because these workflows exceed a single site capacity due to their huge storage and computation requirements, or because the data they process is scattered in different locations. Multisite workflow execution brings about several issues, for which little support has been developed: there is no common ile system for data transfer, inter-site latencies are high, and centralized management becomes a bottleneck. This thesis consists of three contributions towards bridging the gap between single- and multisite workflow execution. First, we present several design strategies to eficiently support the execution of workflow engines across multisite clouds, by reducing the cost of metadata operations. Then, we take one step further and explain how selective handling of metadata, classified by frequency of access, improves workflows performance in a multisite environment. Finally, we look into a different approach to optimize cloud workflow execution by studying some parameters to model and steer elastic scaling.

Este trabalho apresenta uma ferramenta de apoio ao desenvolvimento de aplicações distribuídas, baseadas em RPCs, para o ambiente Windows 95. Discute, ainda, algumas aplicações construídas para validar o sistema, as quais seguem o paradigma cliente-servidor. É feita uma revisão bibliográfica dos assuntos mais relevantes relacionados ao tema e apresentados os detalhes de implementação da ferramenta. Esta foi implementada com técnicas de orientação a objetos, sendo composta por um Gerador Automático de Stubs e uma Biblioteca RPC, além de possibilitar a utilização de um Serviço de Binding. As aplicações construídas procuram explorar as potencialidades da ferramenta, sendo fornecidas também orientações gerais para o desenvolvimento de aplicações distribuídas no ambiente Windows 95. / This work presents a tool to support the development of RPC-based distributed applications in the Windows 95 environment. It also discusses some applications built to validate the system, which follow the client-server model. A theoretical review of the most relevant topics related to the field is done and the tool’s implementation details are presented. This tool was implemented with object-oriented techniques and comprises an automatic Stub Generator and an RPC Library, together with a Binding Service. The distributed applications that were built try to explore the tool’s full potential. General guidelines regarding the development of distributed applications for the Windows 95 environment are also provided.

Cloud computing platforms allow users to deploy geographically distributed applications on servers around the world. Applications may be simple to deploy on these platforms, but it is up to the user and the application to decide which regions and servers to use for application placement. Furthermore, network conditions and routing between the geo-distributed servers change over time, which can lead to sub-optimal performance of applications deployed on such servers. A user could either employ a static deployment configuration of servers, or attempt to use a more dynamic configuration. However, both have inherent limitations. A static configuration will be sub-optimal, as it will be unable to adapt to changing network conditions. A more dynamic approach where an application could switch over or transition to a more suitable server could be beneficial, but this can be very complex in practice. Furthermore, such a solution is more about adapting to change as it happens, and not beforehand. This thesis will investigate the possibility of forecasting impending routing changes between servers, by leveraging messages generated by the Border Gateway Protocol (BGP) and past knowledge about routing changes. BGP routers can delay BGP updates due to factors such as the minimum route advertisement interval (MRAI). Thus, out proposed solution involves forwarding BGP updates downstream in the network, before BGP routers process them. As routing between servers changes, so does the latency, meaning that the latency then could be predicted to some degree. This observation could be applied to realize when the latency to a server increases or decreases past another server. This in turn facilitates the decision process of selecting the most optimal servers in terms of latency for application deployment. The solution presented in this thesis can successfully predict routing changes between end-points in an enclosed environment, and inform users ahead of time that the latency is about to change. The time gained by such predictions depend on factors such as the number of ASs between the end-points, the MRAI, and the update processing delay imposed on BGP routers. Time gains between tens of milliseconds to over 2 minutes has been observed.

Alors que l'informatique est devenue omniprésente dans notre société actuelle, assurer la qualité d'un logiciel revêt une importance grandissante. Pour accroître cette qualité, l'une des conditions à respecter est la correction du système. Dans cette thèse, nous nous intéressons plus particulièrement aux systèmes distribués mettant en œuvre un ou plusieurs programmes exécutés sur plusieurs machines qui communiquent entre elles à travers le réseau. Dans ce contexte, assurer leur correction est rendu plus difficile par leur hétérogénéité mais également par leurs spécificités communes. Les algorithmes correspondants sont parfois complexes et la prédiction de leur comportement difficilement réalisable sans une étude avancée. Les travaux réalisés au cours de cette thèse mettent en œuvre la vérification dynamique formelle de propriétés temporelles sur des applications distribuées. Cette approche consiste à vérifier l'implémentation réelle d'une application à travers son exécution. L'enjeu majeur est de réussir à appliquer les techniques associées au Model checking dans le cadre d'une vérification sur des implémentations réelles d'applications distribuées et non plus sur des modèles abstraits. Pour cela, nous proposons dans un premier temps une analyse sémantique dynamique par introspection mémoire d'un état système permettant de détecter des états sémantiquement identiques. Puis, nous mettons en œuvre la vérification dynamique formelle de certaines propriétés temporelles : les propriétés de vivacité, formulées à l'aide de la logique LTL_X, et le déterminisme des communications dans les applications MPI. Une évaluation de chacune de ces contributions est réalisée à travers plusieurs expériences / While computers have become ubiquitous in our current society, ensuring the software quality takes on an increasing importance. One of the requirements to enhance this quality is the system correctness. In this thesis, we are particularly interested in distributed systems implementing one or more programs executed on several machines which communicate with each other through a network. Ensuring the system correctness is more difficult in this context, due to their heterogeneity but also their common characteristics. Corresponding algorithms are sometimes complex and the prediction of their behavior may be difficult to realize without an advanced study. The work done during this thesis implement the dynamic formal verification of some temporal properties on legacy distributed applications. This approach consists of checking the real implementation of an application by its systematic execution. The challenge in this approach is how to apply the methods derived from Model checking in the context of the verification of legacy distributed applications (without access to source code) and no longer on abstract models. For that, we propose in a first step a dynamic semantic analysis of a system state permitting the detection of identical states. Then, we implement the dynamic formal verification of some temporal properties: liveness properties, specified with the LTL_X logic, and the communications determinism in MPI applications. These contributions are experimentaly validated and evaluated with different series of experiments

Grids, multi-clusters, NUMA systems, and ad-hoc collections of distributed computing devices all present diverse environments in which distributed computing applications can be run. Due to the diversity of features provided by these environments a distributed application that is to perform well must be specifically designed and optimised for the environment in which it is deployed. Such optimisations generally affect the application's communication structure, its consistency protocols, and its communication protocols. This thesis explores approaches to improving the ability of distributed applications to share consistent data efficiently and with improved functionality over wide-area and diverse environments. We identify a fundamental separation of concerns for distributed applications. This is used to propose a new model, called the view model, which is a hybrid, cost-conscious approach to remote data sharing. It provides the necessary mechanisms and interconnects to improve the flexibility and functionality of data sharing without defining new programming models or protocols. We employ the view model to adapt distributed applications to their run-time environment without modifying the application or inventing new consistency or communication protocols. We explore the use of view model properties on several programming models and their consistency protocols. In particular, we focus on programming models used in distributed-shared-memory middleware and applications, as these can benefit significantly from the properties of the view model. Our evaluation demonstrates the benefits, side effects and potential short-comings of the view model by comparing our model with traditional models when running distributed applications across several multi-clusters scenarios. In particular, we show that the view model improves the performance of distributed applications while reducing resource usage and communication overheads.

Grids, multi-clusters, NUMA systems, and ad-hoc collections of distributed computing devices all present diverse environments in which distributed computing applications can be run. Due to the diversity of features provided by these environments a distributed application that is to perform well must be specifically designed and optimised for the environment in which it is deployed. Such optimisations generally affect the application's communication structure, its consistency protocols, and its communication protocols. This thesis explores approaches to improving the ability of distributed applications to share consistent data efficiently and with improved functionality over wide-area and diverse environments. We identify a fundamental separation of concerns for distributed applications. This is used to propose a new model, called the view model, which is a hybrid, cost-conscious approach to remote data sharing. It provides the necessary mechanisms and interconnects to improve the flexibility and functionality of data sharing without defining new programming models or protocols. We employ the view model to adapt distributed applications to their run-time environment without modifying the application or inventing new consistency or communication protocols. We explore the use of view model properties on several programming models and their consistency protocols. In particular, we focus on programming models used in distributed-shared-memory middleware and applications, as these can benefit significantly from the properties of the view model. Our evaluation demonstrates the benefits, side effects and potential short-comings of the view model by comparing our model with traditional models when running distributed applications across several multi-clusters scenarios. In particular, we show that the view model improves the performance of distributed applications while reducing resource usage and communication overheads.

[pt] A implantação de aplicações baseadas em componentes distribuídos é composta por um conjunto de atividades geridas por uma Infraestrutura de Implantação. Aplicações atuais estão se tornando cada vez mais complexas, necessitando de um ambiente alvo dinâmico e multi-plataforma. Assim, a atividade de planejamento de uma implantação é o passo mais crítico, pois define a configuração da infraestrutura de execução de forma a atender os requisitos do ambiente alvo de uma aplicação. Por outro lado, o modelo de serviço na nuvem chamado Infraestrutura como Serviço(IaaS) oferece recursos computacionais sob demanda, com características dinâmicas, escaláveis e elásticas. Nesta dissertação nós estendemos a Infraestrutura de Implantação para componentes SCS de forma a permitir o uso de nuvens privadas ou públicas como o ambiente alvo de uma implantação, através do uso de uma cloud API e políticas flexíveis para especificar um ambiente alvo personalizado. Além disso, hospedamos a infraestrutura de implantação na nuvem. Isto permitiu-nos usar recursos computacionais sob demanda para instanciar os serviços da Infraestrutura de Implantação, produzindo uma Plataforma como Serviço(PaaS) experimental. / [en] Deployment of distributed component-based applications is composed of a set of activities managed by a Deployment Infrastructure. Current applications are becoming increasingly more complex, requiring a multi-platform and a dynamic target environment. Thus, the planning activity is the most critical step because it defines the configuration of the execution infrastructure in order to satisfy the requirements of the application’s target environment. On the other hand, the cloud service model called Infrastructure as a Service (IaaS) offers on-demand computational resources with dynamic, scalable, and elastic features. In this work we have extended the Deployment Infrastructure for SCS componentes to support private or public clouds as its target environment, through the use of a cloud API and flexible policies to specify a customized target environment. Additionally, we host the Deployment Infrastructure on the cloud, which allow us to use on-demand computational resources to instantiate Deployment Infrastructure services, creating an experimental Platform as a Service (PaaS).

The use of distributed systems, involving multiple components, has become a common industry practice. However, modeling the behaviour of such systems is a challenge, especially when the behavior consists of several collaborations of different parties, each involving possibly several starting (input) and ending (output) events of the involved components. Furthermore, the global behavior should be described as a composition of several sub-behaviours, in the following called collaborations, and each collaboration may be further decomposed into several sub-collaborations. We assume that the performance of the elementary sub-collaborations is known, and that the performance of the global behavior should be determined from the performance of the contained elementary collaborations and the form of the composition. A collaboration, in this thesis, is characterized by a partial order of input and output events, and the performance of the collaboration is defined by the minimum delays required for a given output event with respect to an input event. This is a generalization of the semantics of UML Activities, where all input events are assumed to occur at the same time, and all output events occur at the same time. We give a semantic definition of the dynamic behavior of composed collaborations using the composition operators for control flow from UML Activity diagrams, in terms of partial order relationships among the involved input and output events. Based on these semantics, we provide formulas for calculating the performance of composed collaborations in terms of the performance of the sub-collaborations, where each delay is characterized by (a) a fixed value, (b) a range of values, and (c) a distribution (in the case of stochastic behaviours). We also propose approximations for the case of stochastic behavior with Normal distributions, and discuss the expected errors that may be introduced due to ignoring of shared resources or possible dependencies in the case of stochastic behaviours. A tool has been developed for evaluating the performance of complex collaborations, and examples and case studies are discussed to illustrate the applicability of the performance analysis and the visual notation which we introduced for representing the partial-order relationships of the input and output events.

The use of distributed systems, involving multiple components, has become a common industry practice. However, modeling the behaviour of such systems is a challenge, especially when the behavior consists of several collaborations of different parties, each involving possibly several starting (input) and ending (output) events of the involved components. Furthermore, the global behavior should be described as a composition of several sub-behaviours, in the following called collaborations, and each collaboration may be further decomposed into several sub-collaborations. We assume that the performance of the elementary sub-collaborations is known, and that the performance of the global behavior should be determined from the performance of the contained elementary collaborations and the form of the composition. A collaboration, in this thesis, is characterized by a partial order of input and output events, and the performance of the collaboration is defined by the minimum delays required for a given output event with respect to an input event. This is a generalization of the semantics of UML Activities, where all input events are assumed to occur at the same time, and all output events occur at the same time. We give a semantic definition of the dynamic behavior of composed collaborations using the composition operators for control flow from UML Activity diagrams, in terms of partial order relationships among the involved input and output events. Based on these semantics, we provide formulas for calculating the performance of composed collaborations in terms of the performance of the sub-collaborations, where each delay is characterized by (a) a fixed value, (b) a range of values, and (c) a distribution (in the case of stochastic behaviours). We also propose approximations for the case of stochastic behavior with Normal distributions, and discuss the expected errors that may be introduced due to ignoring of shared resources or possible dependencies in the case of stochastic behaviours. A tool has been developed for evaluating the performance of complex collaborations, and examples and case studies are discussed to illustrate the applicability of the performance analysis and the visual notation which we introduced for representing the partial-order relationships of the input and output events.