Ces dernières années ont vu le développement du cloud computing. Le principe fondateur est de déporter la gestion des services informatique des entreprises dans des centres d'hébergement gérés par des entreprise tiers. Ce déport a pour principal avantage une réduction des coûts pour l'entreprise cliente, les moyens nécessaires à la gestion de ces services étant mutualisés entre clients et gérés par l'entreprise hébergeant ces services. Cette évolution implique la gestion de structures d'hébergement à grande échelle, que la dimension et la complexité rendent difficiles à administrer. Avec le développement des infrastructures de calcul de type cluster ou grille ont émergé des système fournissant un support pour l'administration automatisée de ces environnements. Ces systèmes sont désignés sous le terme Systèmes d'Administration Autonomes (SAA). Ils visent à fournir des services permettant d'automatiser les tâches d'administration comme le déploiement des logiciels, la réparation en cas de panne ou leur dimensionnement dynamique en fonction de la charge. Ainsi, il est naturel d'envisager l'utilisation des SAA pour l'administration d'une infrastructure d'hébergement de type clouds. Cependant, nous remarquons que les SAA disponibles à l'heure actuelle ont été pour la plupart conçus pour répondre aux besoins d'un domaine applicatif particulier. Un SAA doit pouvoir être adapté en fonction du domaine considéré, en particulier celui de l'administration d'un cloud. De plus, dans le domaine du cloud, différents besoins doivent être pris en compte : ceux de l'administrateur du centre d'hébergement et ceux de l'utilisateur du centre d'hébergement qui déploie ses applications dans le cloud. Ceci implique qu'un SAA doit pouvoir être adapté pour répondre à ces besoins divers. Dans cette thèse, nous étudions la conception et l'implantation d'un SAA adaptable. Un tel SAA doit permettre d'adapter les services qu'il offre aux besoins des domaines dans lesquels il est utilisé. Nous montrons ensuite comment ce SAA adaptable peut être utilisé pour l'administration autonome d'un environnement de cloud. / Last years have seen the development of cloud computing. The main underlying principle of to externalize the management of companies' IT services in hosting centers which are managed by third party companies. This externalization allows saving costs for the client company, since the resources required to manage these services are mutualized between clients and managed by the hosting company. This orientation implies the management of large scale hosting centers, whose dimension and complexity make them difficult to manage. With the developement of computing infrastructures such as clusters or grids, researchers investigated the design of systems which provides support of an automatized management of these environments. We refer to these system as Autonomic Management Systems (AMS). They aim at providing services which automate administration tasks such as software deployment, fault repair or dynamic dimensioning according to a load. Therefore, in this context, it is natural to consider the use of AMS for the administration of a cloud infrastructure. However, we observe that currently available AMS have been designed to address the requirements of a particular application domain. It should be possible to adapt an AMS according to the considered domain, in particular that of the cloud. Moreover, in the cloud computing area, different requirements have to be accounted : those of the administrator of the hosting center and those of the user of the hosting center (who deploys his application in the cloud). Therefore, an AMS should be adaptable to fulfill such various needs. In this thesis, we investigate the design and implementation of an adaptable AMS. Such an AMS must allow adaptation of all the services it provides, according to the domains where it is used. We next describe the application of this adaptable AMS for the autonomic management of a cloud environment.

Cyber-Physical Systems (CPS) entail the tight integration of and coordination between computational and physical resources. These systems are increasingly becoming vital to modernizing the national critical infrastructure systems ranging from healthcare, to transportation and energy, to homeland security and national defense. Advances in CPS technology are needed to help improve their current capabilities as well as their adaptability, autonomicity, efficiency, reliability, safety and usability.  Due to the proliferation of increasingly sophisticated cyber threats with exponentially destructive effects, CPS defense systems must systematically evolve their detection, understanding, attribution, and mitigation capabilities. Unfortunately most of the current CPS defense systems fall short to adequately provision defense services while maintaining operational continuity and stability of the targeted CPS applications in presence of advanced persistent attacks. Most of these defense systems use un-coordinated combinations of disparate tools to provision defense services for the cyber and physical components. Such isolation and lack of awareness of and cooperation between defense tools may lead to massive resource waste due to unnecessary redundancy, and potential conflicts that can be utilized by a resourceful attacker to penetrate the system.   Recent research argued against the suitability of the current security solutions to CPS environments. We assert the need for new defense platforms that effectively and efficiently manage dynamic defense missions and toolsets in real-time with the following goals: 1) Achieve asymmetric advantage to CPS defenders, prohibitively increasing the cost for attackers; 2) Ensure resilient operations in presence of persistent and evolving attacks and failures; and 3) Facilitate defense alliances, effectively and efficiently diffusing defense intelligence and operations transcending organizational boundaries. Our proposed solution comprehensively addresses the aforementioned goals offering an evolutionary CPS defense system. The presented CPS defense platform, termed CyPhyCARD (Cooperative Autonomous Resilient Defenses for Cyber-Physical systems) presents a unified defense platform to monitor, manage, and control the heterogeneous composition of CPS components. CyPhyCARD relies on three interrelated pillars to construct its defense platform. CyPhyCARD comprehensively integrates these pillars, therefore building a large scale, intrinsically resilient, self- and situation-aware, cooperative, and autonomous defense cloud-like platform that provisions adequate, prompt, and pervasive defense services for large-scale, heterogeneously-composed CPS. The CyPhyCARD pillars are: 1) Autonomous management platform (CyberX) for CyPhyCARD's foundation. CyberX enables application elasticity and autonomic adaptation to changes by runtime diversity employment, enhances the application resilience against attacks and failures by multimodal recovery mechanism, and enables unified application execution on heterogeneously composed platforms by a smart employment of a fine-grained environment-virtualization technology. 2) Diversity management system (ChameleonSoft) built on CyberX. ChameleonSoft encrypts software execution behavior by smart employment of runtime diversity across multiple dimensions to include time, space, and platform heterogeneity inducing a trace-resistant moving-target defense that works on securing CyPhyCARD platform against software attacks. 3) Evolutionary Sensory system (EvoSense) built on CyberX. EvoSense realizes pervasive, intrinsically-resilient, situation-aware sense and response system to seamlessly effect biological-immune-system like defense. EvoSense acts as a middle layer between the defense service provider(s) and the Target of Defense (ToD) creating a uniform defense interface that hides ToD's scale and heterogeneity concerns from defense-provisioning management. CyPhyCARD is evaluated both qualitatively and quantitatively. The efficacy of the presented approach is assessed qualitatively, through a complex synthetic CPS attack scenario. In addition to the presented scenario, we devised multiple prototype packages for the presented pillars to assess their applicability in real execution environment and applications. Further, the efficacy and the efficiency of the presented approach is comprehensively assessed quantitatively by a set of custom-made simulation packages simulating each CyPhyCARD pillar for performance and security evaluation.  The evaluation illustrated the success of CyPhyCARD and its constructing pillars to efficiently and effectively achieve its design objective with reasonable overhead. / Ph. D.

Critical networked services established between service provider and customers are expected to operate respecting Service Level Agreements (SLAs). An interesting possibility to monitor such SLAs is using active measurement mechanisms. However, these mechanisms are expensive in terms of network devices resource consumption and also increase the network load because of the injected traffic. In addition, if the number of SLA violations in a given time is higher than the number of available measurement sessions (common place in large and complex network infrastructures), certainly some violations will be missed. The current best practice, the observation of just a subset of network destinations driven by human administrators expertise, is error prone, does not scale well, and is ineffective on dynamic network conditions. This practice can lead to SLA violations being missed, which invariably affect the performance of several applications. In the present thesis, we advocated the use of Peer-to-Peer (P2P) technology to improve the detection of SLA violations. Such use is described using principles to control active measurement mechanisms. These principles are accomplished through strategies to activate measurement sessions. In this context, the major contributions of this thesis are: i) An approach to improve the detection of SLA violations through the steering of the activation of active measurement sessions using local and remote past service level measurement results and resource utilization constraints; ii) The concept of destination rank as an approach to autonomically prioritize destinations for the activation of active measurement sessions using destination scores; iii) The concept of correlated peers to enable the autonomic provisioning of a P2P measurement overlay for the exchange of relevant active measurement results; iv) The concept of virtual measurement sessions to enable the sharing of measurement results among correlated peers in order to save network devices resources and to improve SLA monitoring coverage; v) The definition of decentralized strategies to steer the activation of active measurement sessions using P2P principles. The method used on the investigation started with the execution of literature reviews on the networkwide control of measurement mechanisms and the employment of P2P technology on network management. After that, the proposed principles to control active measurement mechanisms and the strategies to activate measurement sessions were described. Finally, experiments were performed to evaluate the performance as well as to highlight properties of such principles and strategies. The findings showed properties which improve the detection of SLA violations in terms of the number of detected violations and the adaptivity to network dynamics. We expect that such findings can lead to better SLA monitoring tools and methods.

Critical networked services established between service provider and customers are expected to operate respecting Service Level Agreements (SLAs). An interesting possibility to monitor such SLAs is using active measurement mechanisms. However, these mechanisms are expensive in terms of network devices resource consumption and also increase the network load because of the injected traffic. In addition, if the number of SLA violations in a given time is higher than the number of available measurement sessions (common place in large and complex network infrastructures), certainly some violations will be missed. The current best practice, the observation of just a subset of network destinations driven by human administrators expertise, is error prone, does not scale well, and is ineffective on dynamic network conditions. This practice can lead to SLA violations being missed, which invariably affect the performance of several applications. In the present thesis, we advocated the use of Peer-to-Peer (P2P) technology to improve the detection of SLA violations. Such use is described using principles to control active measurement mechanisms. These principles are accomplished through strategies to activate measurement sessions. In this context, the major contributions of this thesis are: i) An approach to improve the detection of SLA violations through the steering of the activation of active measurement sessions using local and remote past service level measurement results and resource utilization constraints; ii) The concept of destination rank as an approach to autonomically prioritize destinations for the activation of active measurement sessions using destination scores; iii) The concept of correlated peers to enable the autonomic provisioning of a P2P measurement overlay for the exchange of relevant active measurement results; iv) The concept of virtual measurement sessions to enable the sharing of measurement results among correlated peers in order to save network devices resources and to improve SLA monitoring coverage; v) The definition of decentralized strategies to steer the activation of active measurement sessions using P2P principles. The method used on the investigation started with the execution of literature reviews on the networkwide control of measurement mechanisms and the employment of P2P technology on network management. After that, the proposed principles to control active measurement mechanisms and the strategies to activate measurement sessions were described. Finally, experiments were performed to evaluate the performance as well as to highlight properties of such principles and strategies. The findings showed properties which improve the detection of SLA violations in terms of the number of detected violations and the adaptivity to network dynamics. We expect that such findings can lead to better SLA monitoring tools and methods.

Critical networked services established between service provider and customers are expected to operate respecting Service Level Agreements (SLAs). An interesting possibility to monitor such SLAs is using active measurement mechanisms. However, these mechanisms are expensive in terms of network devices resource consumption and also increase the network load because of the injected traffic. In addition, if the number of SLA violations in a given time is higher than the number of available measurement sessions (common place in large and complex network infrastructures), certainly some violations will be missed. The current best practice, the observation of just a subset of network destinations driven by human administrators expertise, is error prone, does not scale well, and is ineffective on dynamic network conditions. This practice can lead to SLA violations being missed, which invariably affect the performance of several applications. In the present thesis, we advocated the use of Peer-to-Peer (P2P) technology to improve the detection of SLA violations. Such use is described using principles to control active measurement mechanisms. These principles are accomplished through strategies to activate measurement sessions. In this context, the major contributions of this thesis are: i) An approach to improve the detection of SLA violations through the steering of the activation of active measurement sessions using local and remote past service level measurement results and resource utilization constraints; ii) The concept of destination rank as an approach to autonomically prioritize destinations for the activation of active measurement sessions using destination scores; iii) The concept of correlated peers to enable the autonomic provisioning of a P2P measurement overlay for the exchange of relevant active measurement results; iv) The concept of virtual measurement sessions to enable the sharing of measurement results among correlated peers in order to save network devices resources and to improve SLA monitoring coverage; v) The definition of decentralized strategies to steer the activation of active measurement sessions using P2P principles. The method used on the investigation started with the execution of literature reviews on the networkwide control of measurement mechanisms and the employment of P2P technology on network management. After that, the proposed principles to control active measurement mechanisms and the strategies to activate measurement sessions were described. Finally, experiments were performed to evaluate the performance as well as to highlight properties of such principles and strategies. The findings showed properties which improve the detection of SLA violations in terms of the number of detected violations and the adaptivity to network dynamics. We expect that such findings can lead to better SLA monitoring tools and methods.

This thesis investigates the application of autonomic management to a distributed storage system. Effects on performance and resource consumption were measured in experiments, which were carried out in a local area test-bed. The experiments were conducted with components of one specific distributed storage system, but seek to be applicable to a wide range of such systems, in particular those exposed to varying conditions. The perceived characteristics of distributed storage systems depend on their configuration parameters and on various dynamic conditions. For a given set of conditions, one specific configuration may be better than another with respect to measures such as resource consumption and performance. Here, configuration parameter values were set dynamically and the results compared with a static configuration. It was hypothesised that under non-changing conditions this would allow the system to converge on a configuration that was more suitable than any that could be set a priori. Furthermore, the system could react to a change in conditions by adopting a more appropriate configuration. Autonomic management was applied to the peer-to-peer (P2P) and data retrieval components of ASA, a distributed storage system. The effects were measured experimentally for various workload and churn patterns. The management policies and mechanisms were implemented using a generic autonomic management framework developed during this work. The motivation for both groups of experiments was to test management policies with the objective to avoid unsatisfactory situations with respect to resource consumption and performance. Such unsatisfactory situations occur when either the P2P layer or the data retrieval mechanism is configured statically. In a statically configured P2P system two unsatisfactory situations can be identified. The first arises when the frequency with which P2P node states are verified is low and membership churn is high. The P2P node state becomes inaccurate due to a high membership churn, leading to errors during the routing process and a reduction in performance. In this situation it is desirable to increase the frequency to increase P2P state accuracy. The converse situation arises when the frequency is high and churn is low. In this situation network resources are used unnecessarily, which may also reduce performance, making it desirable to decrease the frequency. In ASA’s data retrieval mechanism similar unsatisfactory situations can be identified with respect to the degree of concurrency (DOC). The DOC controls the eagerness with which multiple redundant replicas are retrieved. An unsatisfactory situation arises when the DOC is low and there is a large variation in the times taken to retrieve replicas. In this situation it is desirable to increase the DOC, because by retrieving more replicas in parallel a result can be returned to the user sooner. The converse situation arises when the DOC is high, there is little variation in retrieval time and there is a network bottleneck close to the requesting client. In this situation it is desirable to decrease the DOC, since the low variation removes any benefit in parallel retrieval, and the bottleneck means that decreasing parallelism reduces both bandwidth consumption and elapsed time for the user. The experimental evaluations of autonomic management show promising results, and suggest several future research topics. These include optimisations of the managed mechanisms, alternative management policies, different evaluation methods, and the application of developed management mechanisms to other facets of a distributed storage system. The findings of this thesis could be exploited in building other distributed storage systems that focus on harnessing storage on user workstations, since these are particularly likely to be exposed to varying, unpredictable conditions.

Au cours de la dernière décennie, la complexité des applications a considérablement évolué afin de répondre aux besoins métiers émergeants. Leur conception implique une composition distribuée de composants logiciels. Ces applications fournissent des services à travers les interactions métiers maintenues par leurs composants. De telles applications sont intrinsèquement en évolution dynamique en raison de la dynamicité de leurs contextes. En effet, elles évoluent dans des environnements qui changent tout en présentant des conditions très dynamiques durant leur cycle de vie d’exécution. De tels contextes représentent une lourde charge pour les développeurs aussi bien pour leurs tâches de conception que de gestion. Cela a motivé́ le besoin de renforcer l’autonomie de gestion des applications pour les rendre moins dépendantes de l’intervention humaine en utilisant les principes de l’Informatique Autonomique. Les Systèmes Informatiques Autonomes (SIA) impliquent l’utilisation des boucles autonomiques, dédiées aux systèmes afin de les aider à accomplir leurs tâches de gestion. Ces boucles ont pour objectif d’adapter leurs systèmes à la dynamicité de leurs contextes, en se basant sur une logique d’adaptation intégrée. Cette logique est souvent donnée par des règles statiques codées manuellement. La construction de ces règles demande beaucoup de temps tout en exigeant une bonne expertise. En fait, elles nécessitent une compréhension approfondie de la dynamicité du système afin de prédire les adaptations précises à apporter à celui-ci. Par ailleurs, une telle logique ne peut envisager tous les scénarios d’adaptation possibles, donc, ne sera pas en mesure de prendre en compte des adaptations pour des situations précédemment inconnues. Les SIA devraient donc être assez sophistiqués afin de pouvoir faire face à la nature dynamique de leurs contextes et de pouvoir apprendre par eux-mêmes afin d’agir correctement dans des situations inconnues. Les SIA devraient également être capables d’apprendre de leur propre expérience passée afin de modifier leur logique d’adaptation en fonction de la dynamicité de leurs contextes. Dans ce manuscrit, nous abordons les lacunes décrites en utilisant les techniques d’Apprentissage par Renforcement (AR) afin de construire notre logique d’adaptation. Cependant, les approches fondées sur l’AR sont connues pour leur mauvaise performance lors des premières phases d’apprentissage. Cette mauvaise performance entrave leur utilisation dans le monde réel des systèmes déployés. Par conséquent, nous avons amélioré cette logique d’adaptation avec des capacités d’apprentissage plus performantes avec une approche AR en multi-pas. Notre objectif est d’optimiser la performance de l’apprentissage et de le rendre plus efficace et plus rapide, en particulier durant les premières phases d’apprentissage. Nous avons aussi proposé́ un cadriciel générique visant à aider les développeurs dans la construction d’applications auto-adaptatives. Nous avons donc proposé de transformer des applications existantes en ajoutant des capacités d’autonomie et d’apprentissage à leurs composants. La transformation consiste en l’encapsulation des composants dans des conteneurs autonomiques pour les doter du comportement auto-adaptatif nécessaire. Notre objectif est d’alléger la charge des tâches de gestion des développeurs et de leur permettre de se concentrer plus sur la logique métier de leurs applications. Les solutions proposées sont destinées à être génériques, granulaires et basées sur un standard connu, à savoir l’Architecture de Composant de Service. Enfin, nos propositions ont été évaluées et validées avec des résultats expérimentaux. Ils ont démontré leur efficacité en montrant un ajustement dynamique des applications transformées face aux dynamicités de leurs contextes en un temps beaucoup plus court comparé aux approches existantes / During the past decade, the complexity of applications has significantly scaled to satisfy the emerging business needs. Their design entails a composition of distributed and interacting software components. They provide services by means of the business interactions maintained by their components. Such applications are inherently in a dynamic evolution due to their context dynamics. Indeed, they evolve in changing environments while exhibiting highly dynamic conditions during their execution life-cycle (e.g., their load, availability, performance, etc.). Such contexts have burdened the applications developers with their design and management tasks. Subsequently, motivated the need to enforce the autonomy of their management to be less dependent on human interventions with the Autonomic Computing principles. Autonomic Computing Systems (ACS) implies the usage of autonomic loops, dedicated to help the system to achieve its management tasks. These loops main role is to adapt their associated systems to the dynamic of their contexts by acting upon an embedded adaptation logic. Most of time, this logic is given by static hand-coded rules, often concern-specific and potentially error-prone. It is undoubtedly time and effort-consuming while demanding a costly expertise. Actually, it requires a thorough understanding of the system design and dynamics to predict the accurate adaptations to bring to the system. Furthermore, such logic cannot envisage all the possible adaptation scenarios, hence, not able to take appropriate adaptations for previously unknown situations. ACS should be sophisticated enough to cope with the dynamic nature of their contexts and be able to learn on their own to properly act in unknown situations. They should also be able to learn from their past experiences and modify their adaptation logic according to their context dynamics. In this thesis manuscript, we address the described shortcomings by using Reinforcement Learning (RL) techniques to build our adaptation logic. Nevertheless, RL-based approaches are known for their poor performance during the early stages of learning. This poor performance hinders their usage in real-world deployed systems. Accordingly, we enhanced the adaptation logic with sophisticated and better-performing learning abilities with a multi-step RL approach. Our main objective is to optimize the learning performance and render it timely-efficient which considerably improves the ACS performance even during the beginning of learning phase. Thereafter, we pushed further our work by proposing a generic framework aimed to support the application developers in building self-adaptive applications. We proposed to transform existing applications by dynamically adding autonomic and learning abilities to their components. The transformation entails the encapsulation of components into autonomic containers to provide them with the needed self-adaptive behavior. The objective is to alleviate the burden of management tasks on the developers and let them focus on the business logic of their applications. The proposed solutions are intended to be generic, granular and based on a well known standard (i.e., Service Component Architecture). Finally, our proposals were evaluated and validated with experimental results. They demonstrated their effectiveness by showing a dynamic adjustment to the transformed application to its context changes in a shorter time as compared to existing approaches

La motivation principale de cette thèse est de faire face à l'accroissement de la complexité des systèmes informatiques, qui, dans un futur proche ( de l'ordre de quelques années) risque fort d'être le principal frein à leur évolution et à leur développement. Aujourd'hui la tendance s'inverse et le coût de gestion humaine dépasse le coût des infrastructures matérielles et logicielles. De plus, l'administration manuelle de grands systèmes (applications distribuées, réseaux de capteurs, équipements réseaux) est non seulement lente mais aussi sujette à de nombreuses erreurs humaines. Un des domaines de recherche émergent est celui de l'informatique autonomique qui a pour but de rendre ces systèmes auto-gérés. Nous proposons une approche qui permet de décrire des politiques de gestion autonomiques de haut niveau. Ces politiques permettent au système d'assurer quatre propriétés fondamentales de l'auto-gestion: l'auto-guérison, l'auto-configuration, l'auto-protection et l'auto-optimisation. Nos contributions portent sur la spécification de diagrammes de description de politiques de gestion autonomiques appelés (S)PDD "(Sensor) Policy Description Diagrams". Ces diagrammes sont implémentés dans le gestionnaire autonomique TUNe et l'approche a été validée sur de nombreux systèmes: simulation électromagnétique répartie sur grille de calcul, réseaux de capteurs SunSPOT, répartiteur de calcul DIET. Une deuxième partie présente une modélisation mathématique de l’auto-optimisation pour un « datacenter ». Nous introduisons un problème de minimisation d’un critère intégrant d’une part la consommation électrique des équipements du réseau du « datacenter » et d’autre part la qualité de service des applications déployées sur le « datacenter ». Une heuristique permet de prendre en compte les contraintes dues aux fonctions de routage utilisées. / The main challenge of this thesis is to cope with the growing complexity of IT systems. In a near future (mainly the next few years) this complexity will prevent new developments and system evolutions. Today the trend is reversing and the managing costs are overtaking the hardware and software costs. Moreover, the manual administration of large systems (distributed applications, sensor networks, and network equipment) is not only slow but error-prone. An emerging research field called autonomic computing tries to bring up self-managed systems. We introduce an approach that enable the description of high level autonomic management policies. These policies allow the system to ensure four fundamental properties for self-management: self-healing, self-self-configuring, self-protecting and self-optimizing. We specify autonomic management Policy Description Diagrams (PDD) and implement them in Toulouse University Network (TUNe). We validated our approach on many systems: electromagnetic simulations distributed on computer grids (grid’5000), wireless sensor networks with SunSPOTs and the computing scheduler DIET. A second part of this thesis presents a mathematical modeling for self-optimizing datacenters. We introduce a minimization problem with a criterion integrating both the electrical consumption of the datacenter networking equipment and the quality of service of the deployed applications. A heuristic takes into account the routing functions used on the network.