In this dissertation, we present our work on fault management in distributed systems, with motivating application roots in monitoring fault and abrupt change of large computing systems like the grid and the cloud. Instead of building a complete a priori knowledge of the software and hardware infrastructures as in conventional detection or diagnosis methods, we propose to use appropriate techniques to perform end-to-end monitoring for such large scale systems, leaving the inaccessible details of involved components in a black box.For the fault monitoring of a distributed system, we first model this probe-based application as a static collaborative prediction (CP) task, and experimentally demonstrate the effectiveness of CP methods by using the max margin matrix factorization method. We further introduce active learning to the CP framework and exhibit its critical advantage in dealing with highly imbalanced data, which is specially useful for identifying the minority fault class.Further we extend the static fault monitoring to the sequential case by proposing the sequential matrix factorization (SMF) method. SMF takes a sequence of partially observed matrices as input, and produces predictions with information both from the current and history time windows. Active learning is also employed to SMF, such that the highly imbalanced data can be coped with properly. In addition to the sequential methods, a smoothing action taken on the estimation sequence has shown to be a practically useful trick for enhancing sequential prediction performance.Since the stationary assumption employed in the static and sequential fault monitoring becomes unrealistic in the presence of abrupt changes, we propose a semi-supervised online change detection (SSOCD) framework to detect intended changes in time series data. In this way, the static model of the system can be recomputed once an abrupt change is detected. In SSOCD, an unsupervised offline method is proposed to analyze a sample data series. The change points thus detected are used to train a supervised online model, which gives online decision about whether there is a change presented in the arriving data sequence. State-of-the-art change detection methods are employed to demonstrate the usefulness of the framework.All presented work is verified on real-world datasets. Specifically, the fault monitoring experiments are conducted on a dataset collected from the Biomed grid infrastructure within the European Grid Initiative, and the abrupt change detection framework is verified on a dataset concerning the performance change of an online site with large amount of traffic.

Les architectures modulaires intégrées (IMA) sont une évolution majeure de l'architecture des systèmes avioniques. Elles permettent à plusieurs systèmes de se partager des ressources matérielles sans interférer dans leur fonctionnement grâce à un partitionnement spatial (zones mémoires prédéfinies) et temporel (ordonnancement statique) dans les processeurs ainsi qu'une réservation des ressources sur les réseaux empruntés. Ces allocations statiques permettent de vérifier le déterminisme général des différents systèmes: chaque système doit respecter des exigences de bout-en-bout dans une architecture asynchrone. Une étude pire cas permet d'évaluer les situations amenant aux limites du système et de vérifier que les exigences de bouten- bout sont satisfaites dans tous les cas. Les architectures IMA utilisés dans les avions centralisent physiquement des modules de calcul puissants dans des baies avioniques. Dans le cadre d'une étude de cas hélicoptère, ces baies ne sont pas envisageables pour des raisons d'encombrement: des processeurs moins puissants, utilisés à plus de 80%, composent ces architectures. Pour ajouter de nouvelles fonctionnalités ainsi que de nouveaux équipements, le souhait est de distribuer la puissance de traitement sur un plus grand nombre de processeurs dans le cadre d'une architecture globale asynchrone. Deux problématiques fortes ont été mises en avant tout au long de cette thèse. La première est la répartition des fonctions avioniques associée à une contrainte d'ordonnancement hors-ligne sur les différents processeurs. La deuxième est la satisfaction des exigences de communication de bout-en-bout, dépendantes de l'allocation et l'ordonnancement des fonctions ainsi que des latences de communication sur les réseaux. La contribution majeure de cette thèse est la recherche d'un compromis entre la distribution des architectures IMA sur un plus grand nombre de processeurs et la satisfaction des exigences de communication de bout-en-bout. Nous répondons à cet enjeu de la manière suivante: - Nous formalisons dans un premier temps un modèle de partitions communicantes tenant en compte des contraintes d'allocation et d'ordonnancement des partitions d'une part et des contraintes de communication de bout-en-bout entre partitions d'autre part. - Nous présentons dans un deuxième temps une recherche exhaustive des architectures valides. Nous proposons l'allocation successive des fonctions avioniques en considérant au même niveau la problématique d'ordonnancement et la satisfaction des exigences de bout-en-bout avec des latences de communication figées. Cette méthode itérative permet de construire des allocations de partitions partiellement valides. La construction des ordonnancements dans chacun des processeurs est cependant une démarche coûteuse dans le cadre d'une recherche exhaustive. - Nous avons conçu dans un troisième temps une heuristique gloutonne pour réduire l'espace de recherche associé aux ordonnancements. Elle permet de répondre aux enjeux de distribution d'une architecture IMA dans un contexte hélicoptère. - Nous nous intéressons dans un quatrième temps à l'impact des latences de communication de bout-en-bout sur des architectures distribuées données. Nous proposons pour celles-ci les choix de réseaux basés sur les latences de communication admissibles entre les différentes fonctions avioniques. Les méthodes que nous proposons répondent au besoin industriel de l'étude de cas hélicoptère, ainsi qu'à celui de systèmes de plus grande taille. / Integrated Modular Architectures (IMA) is a major evolution of avionics systems. A spatial (predefined memory zones) and temporal (off-line scheduling) partitioning as well as communication resources reservation permit several systems not to interfere in this architecture. The determinism of systems is proved thanks to these static allocations: each system must respect end-to-end requirements in an asynchronous architecture. A worst-case study permits to assess the bounds of systems in order to verify that end-to-end requirements are satisfied in all the cases. IMA architectures physically centralize powerful computing resources in avionics bays in aircraft. These aren't feasible in helicopters due to size reasons: powerless processors, at least 80% used, set these architectures. In order to add new functionalities and equipment, the aim is to distribute processing power over a larger number of processors in the context of a globally asynchronous architecture. Two strong issues have been advanced throughout this thesis. The first one is the distribution of avionics functions with an off-line scheduling constraint on the different processors. The second one is the satisfaction of end-to-end requirements, depending on allocation and scheduling of functions as well as communication latencies over the networks. This thesis proposes a trade-off between the distribution of IMA architectures on a larger number of processors and the satisfaction of end-to-end communication requirements. We answer at this topic as follows: - First, we formalize a communicating partitions model based on the partitions allocation and scheduling constraints on the one hand and end-to-end communication constraints on the other hand. - Second, we present an exhaustive search of valid architectures. We introduce a successive allocation of avionics functions considering altogether the scheduling and the satisfaction of end-to-end constraints with fixed communication latencies. This iterative method allows the building of partially valid allocations schemes, but the scheduling search is expensive here. - Third, we create a greedy heuristic to reduce the scheduling search space. It permits to meet the challenges of the distribution of IMA architecture in a helicopter context. - Finally, we focus on the impact of end-to-end communication latencies on given distributed architectures. We define for them the networks based on eligible communication latencies between the different avionics functions. Our methods answer the industrial case study needs as well as bigger size systems needs.

Mobile Ad Hoc Networks (MANETs) are a class of infrastructure less network architecture which are formed by a collection of mobile nodes that communicate with each other using multi-hop wireless links. They eliminate the need for central management, hence each node must operate cooperatively to successfully maintain the network. Each node performs as a source, a sink and a router. Future applications of MANETs are expected to be based on all-IP architecture, carrying a multitude of real-time multimedia applications such as voice, video and data. It would be necessary for MANETs to have an efficient routing and quality of service (QoS) mechanism to support diverse applications. This thesis proposes a set of cooperative protocols that provide support for QoS routing. The first is the on-demand, Non-Disjoint Multiple Routes Discovery protocol (NDMRD). NDMRD allows the establishment of multiple paths with node non-disjoint between source and destination node. It returns to the source a collection of routes with the QoS parameters. The second part of the protocol is the Node State Monitoring protocol for the purpose of monitoring, acquisition, dissemination and accumulation of QoS route information. The third part of the protocol implements the QoS route selection based on a Genetic Algorithm. The GA is implemented online with predetermined initial population and weighted-sum fitness function which operates simultaneously on the node bandwidth, media access delay, end to end delay and the node connectivity index (NCI). The term node connectivity index is a numerical value designed to predict comparatively the longest time a node-pair might be connected wirelessly.

Orientador: Mauricio Ferreira Magalhães / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Eletrica e de Computação / Made available in DSpace on 2018-08-07T01:33:38Z (GMT). No. of bitstreams: 1 Duarte_RafaelLuiz_M.pdf: 1020338 bytes, checksum: 55a286156672abedc8e5b20a444d8a7c (MD5) Previous issue date: 2006 / Resumo: Para o provisionamento de conexões fim-a-fim em domínios ópticos é desejável uma solução de arquitetura que permita o estabelecimento automático destas conexões: A tecnologia GMPLS ofe­rece um plano de controle que especifica mecanismos que estendem o roteamento e a sinalização do mundo IP contribuindo para a configuração automática de conexões em domínios ópticos. Este traba­lho propõe uma arquitetura para o provisionamento de conexões fim-a-fim em redes ápticas GMPLS. Tal arquitetura é baseada na tecnologia Web services e permite o estabelecimento de dois tipos de conexões. A primeira, conhecida como SPC (Soft Permanent Connection), é utilizada pelo gerente do domínio óptico. O SPC conecta dois elementos de rede pertencentes ao mesmo domínio. A se­gunda, é uma conexão fim-a-fim na qual um usuário cliente pode enviar dados através de múltiplos domínios. Neste último caso, o cliente precisa negociar com cada domínio para verificar a disponibi­lidade de recursos. Para tal arquitetura foram criados módulos responsáveis pelo provisionamento e gerenciamento das conexões além da utilização do simulador GLASS para validação da arquitetura proposta / Abstract: For the provisioning of end-to-end connections in optical domains it is desirable to have an archi­tecture that supports the automatic establishment of such connections. The GMPLS technology offers a control plan that specifies mechanisms which extend the routing and the signalling of IP world con­tributing for the automatic configuration of end-to-end connections in optical domains. This work proposes an architecture for the provisioning of end-to-end connections in GMPLS optical networks. Such architecture is based on the Web services technology and allows the establishment of two kinds of connections. The first one, known as SPC (Soft Permanent Connection), is used by the manager of the optical domain. The SPC connects two network elements belonging to the same domain. The se­cond one is an end-to-end connection by which a given client can send data across multiple domains. In this last case the client needs to negotiate with each domain to verify the availability of resources. For such architecture we defined modules responsibIe for the provisioning and management of the connections. The GLASS simulator was used for validation ofthe proposed architecture / Mestrado / Engenharia de Computação / Mestre em Engenharia Elétrica

A contração muscular gerada por impulsos elétricos provenientes de duas fontes nervosas distintas pode ser alternativa no tratamento de lesões do plexo braquial e na paralisia facial. O objetivo desta tese foi avaliar e comparar diferentes técnicas de reinervação dupla com a técnica de reinervação única do músculo gastrocnêmio em ratos. Cinquenta ratos Wistar adultos, após terem seu nervo fibular direito seccionado, foram divididos em cinco grupos com relação ao procedimento realizado no nervo tibial: controle (C); seccionado (S); neurorrafia término-terminal (TT); neurorrafia primária associada à transferência nervosa fibular para tibial de maneira término-lateral (TL); e neurorrafia término-terminal convergente entre os cotos proximais dos nervos tibial e fibular com o coto distal do nervo tibial (TTC). Os resultados foram avaliados 12 semanas após o experimento por meio do teste da marcha, eletromiografia, índice de massa do músculo gastrocnêmio e contagem axonal no coto distal do nervo tibial. Os grupos de reinervação dupla (TL e TTC) revelaram maiores resultados funcionais (p < 0,05) em relação ao grupo de reinervação única (TT). O grupo TTC apresentou maior amplitude (p=0,006) e maior latência (p=0,041) do que o grupo TT. Em relação ao índice de massa muscular, não houve diferença entre os grupos de reinervação (p > 0,705). A análise histológica revelou maior densidade axonal no grupo TTC em relação ao grupo TT (p=0,001) e ao grupo TL (p=0,002). Ambas técnicas de dupla reinervação revelaram recuperação funcional do músculo gastrocnêmio mais precoce e maior quando comparadas à técnica de reinervação única (TT). Os animais do grupo TTC apresentaram maior número de axônios regenerados no coto distal do nervo tibial do que os do grupo TT e TL / Muscle contraction generated by electrical impulses simultaneously originating from two different neural sources may be an interesting treatment alternative for facial palsy and brachial plexus injury. The purpose of this thesis was to evaluate and compare distinct double reinnervation techniques with single reinnervation technique of gastrocnemius muscle in rats. Fifty adult Wistar rats underwent transection of their right peroneal nerve and were divided into five groups related to tibial nerve procedure: the control group (C), tibial nerve section group (S), tibial nerve end-to-end neurorrhaphy (EE) group, tibial nerve primary repair associated with end-to-side peroneal-to-tibial nerve transfer (ES) group, and tibial nerve repair by convergent end-to-end neurorrhaphy between the proximal stumps of the tibial and peroneal nerves to the distal stump of the tibial nerve (CEE) group. The outcomes were assessed 12 weeks after the experiment by use of a walking track, electromyography, gastrocnemius muscle mass index, and histomorphometric analysis of the distal tibial nerve. The double reinnervation groups (ES and CEE) showed greater functional recovery (p < 0.05) than the single reinnervation group (EE). The CEE group showed greater amplitude (p=0.006) and higher latency (p=0.041) than the EE group. There was no difference in the muscle mass index among the reinnervation groups (p > 0.705). Histologic analysis revealed greater axonal density in the CEE group than EE group (p=0.001) and ES group (p=0.002). The double reinnervation techniques showed earlier and greater functional recovery of the gastrocnemius muscle than did the single reinnervation technique. The CEE group showed a higher number of regenerated axons in the distal tibial nerve stump

The thesis is focused on simulating of protocols from standard IEC 61850 in simulation tool OMNeT++. The theoretical part in the thesis deals with description of the field of operating technologies, supervisory control and data acquisition and protocols Tase-2/ICCP, IEC 61850, IEC 60870-5-104, DNP 3 and DLMS/COSEM. Next part deals with parameters influencing connection and description of simulation tools NS2/NS3, OPNET and OMNeT++. In the practical part there is created a network containing protocols from the standard IEC 61850 in the simulation program OMNeT++. Then in the network a end to end delay and packet loss with increasing traffic is measured.

Dans les systèmes embarqués critiques, assurer la sûreté de fonctionnement est primordial du fait, à la fois, des exigences en sûreté dictées par les autorités de certification et des contraintes en sûreté de ces systèmes où des défaillances pourraient conduire à des évènements catastrophiques, voire la perte de vies humaines. Les architectures de ces systèmes sont aujourd'hui de plus en plus distribuées, s'appuyant sur des réseaux numériques complexes, ce qui pose la problématique de l'intégrité des communications. Dans ce contexte, nous proposons une approche bout en bout pour l'intégrité des communications, basée sur le concept du " canal noir " introduit par l'IEC 61508. Elle utilise les codes détecteurs d'erreurs CRC, Adler et Fletcher. Selon le niveau de redondance des systèmes, nous proposons une approche multi-codes (intégrité jugée sur un lot de messages) pour les systèmes dotés d'un niveau de redondance important et une approche mono-code (intégrité jugée sur chaque message) pour les autres cas. Nous avons validé ces propositions par des expérimentations évaluant le pouvoir de détection intrinsèque de chaque code détecteur et la complémentarité entre ces code en termes de pouvoir de détection, ainsi que leurs coûts de calcul avec une analyse de l'impact du type de leur implémentation et de l'environnement matériel (standard et embarqué : processeurs i7, STM32, TMS320C6657 et P2020). L'approche mono-code a été appliquée à un cas d'étude industriel : les futurs systèmes de commande de vol d'Airbus Helicopters. / In critical embedded systems, ensuring dependability is crucial given both dependability requirements imposed by certification authorities and dependability constraints of these systems where failures could lead to catastrophic events even loss of human lives. The architectures of these systems are increasingly distributed deploying complex digital networks which raise the issue of communication integrity. In this context, we propose an end to end approach for communication integrity. This approach is based on the "black channel" concept introduced by IEC 61508. It uses error detection codes particularly CRC, Adler and Fletcher. Depending on the redundancy level of targeted systems, we propose a multi-codes approach (integrity of a set of messages) for systems with an important redundancy level and a single- code approach (integrity per message) for the other cases. We validated our proposals through experiments in order to evaluate intrinsic error detection capability of each error detection code, their complementarity in terms of error detection and their computational costs by analyzing the impact of the type of implementation and the hardware environment (standard or embedded: i7, STM32, TMS320C6657 and P2020 processors). The single-code approach was applied to an industrial case study: future flight control systems of Airbus Helicopters.

As cloud-native applications become more distributed, complex, and unpredictable with the adoption of microservices and other new architectural components, traditional monitoring solutions are inadequate in providing end-to-end visibility and proactively identifying deviations from expected behaviour before they become disruptive to services. In response to these challenges, observability-driven development (ODD) is proposed as a new methodology that leverages tools and practices to observe the state and detect the behaviour of systems. Unlike the leading IT giants developing their proprietary tools and platforms, some non-IT companies and smaller organizations still have difficulty adopting observability-driven development. Proprietary development demands extensive resources and manpower, while connecting to third-party platforms may compromise data security. This thesis proposed an end-to-end observability pipeline that is composed of merely open-source components. The pipeline collects and correlates metrics, logs, and traces to facilitate software development and help troubleshoot in production. The pipeline is designed to be adaptive and extensible so that companies can adopt it as the first step towards observability-driven development, and customize it to meet their specific requirements. / Molnbaserade applikationer blir alltmer distribuerade, komplexa och oförutsägbara med införandet av mikrotjänster och andra nya arkitektoniska komponenter. Detta resulterar i att traditionella övervakningslösningar blir alltmer inadekvata. De traditionella lösningarna tillhandahåller inte tillräcklig överskådlighet över dessa applikationer (end-to-end) för proaktiv identifiering av avvikelser från förväntat beteende innan de börjar påverka tjänsterna negativt. Som svar på dessa utmaningar föreslås observerbarhetsdriven utveckling (ODD) som en ny metod som utnyttjar verktyg och praxis för att observera tillståndet och upptäcka systemens beteende. Till skillnad från de ledande IT-jättarna som utvecklar sina egna verktyg och plattformar, har vissa icke-IT-företag och mindre organisationer fortfarande svårt att ta till sig observerbarhetsdriven utveckling. Egenutvecklad mjukvara kräver omfattande resurser och arbetskraft, medan anslutning till tredjepartsplattformar kan äventyra datasäkerheten. Den här avhandlingen bidrar med en end-to-end lösning som enbart baserats på öppen källkod. Pipelinen samlar in data från loggar och korrelerar dessa mätvärden för att underlätta mjukvaruutveckling och hjälpa till att felsöka i produktionen. Pipelinen är designad för att vara anpassningsbar och utvidgningsbar så att företag kan använda den som ett första steg mot observerbarhetsdriven utveckling och anpassa den för att möta deras specifika krav.

Wireless sensor networks is a broad subject with many applications and interesting research areas, such as optimization within connectivity and energy efficiency. One problem is that most published articles in this field use customized simulation environments and do not provide source code of their implementation. By not including aspects of implementation, it becomes difficult to determine how the results are achieved, which questions the validity and reliability of the works. This thesis aims to reproduce one of these researched methods, an algorithm that balances battery life with efficient routing within a network using fuzzy logic, with the goal to increase the reliability of the methodology within its field. The research question constructed on the foundation of these premises is thus “Is reproducibility satisfactory in a research work on a multi-objective routing algorithm, using fuzzy logic, in wireless sensor networks?, a case study by Minhas et al”. Two additional research questions emerge from the first one: “How does the reproduced algorithm perform in comparison to a selection of dif erent routing algorithms?” and “Is the reproduced algorithm, as is, applicable to a less idealistic environment?” To answer the research questions a computer simulation method is used to build, execute, and analyze the output of the algorithms. The results show that the implemented algorithm performs noticeably better in both lifetime and ratio to the shortest path compared with the original implementation, hinting towards the implementation and reproducibility deviating from expected results. The reproduced algorithm is also compared to two other algorithms under a different simulation environment, where it performs better in lifetime and packet delivery rate whilst performing slightly worse in energy efficiency and total energy consumption. Due to the significant differences in performance against the reproduced article’s implementation the study concludes that the reproducibility is not satisfactory. Lastly, it concludes that it does not perform well in a less idealistic simulation environment, making it less applicable.

No description available.