Resource management is an important component in many distributed clusters. A resource manager handles which server a task should run on and which user’s task that should be allocated. If a system has multiple users with similar demands, all users should have an equal share of the cluster, making the system fair. This is typically done today using a centralized server which has full knowledge of all servers in the cluster and the different users. Having a centralized server brings problems such as single point of failure, and vertical scaling on the resource manager. This thesis focuses on fairness for users during task allocation with a decentralized resource manager. A solution called, Parallel Distributed Gradient-based Dominant Resource Fairness, is proposed. It allows servers to handle a subset of users and to allocate tasks in parallel, while maintaining fairness results close to a centralized server. The solution utilizes a gradient network topology overlay to sort the servers based on their users’ current usage and allows a server to know if it has the user with the currently lowest resource usage. The solution is compared to pre-existing solutions, based on fairness and allocation time. The results show that the solution is more fair than the pre-existing solutions based on the gini-coefficient. The results also show that the allocation time scales based on the number of users in the cluster because it allows more parallel allocations by the servers. It does not scale as well though as existing distributed solutions. With 40 users and over 100 servers the solution has an equal time to a centralized solution and outperforms a centralized solution with more users. / Resurshantering är en viktig komponent i många distribuerade kluster. En resurshanterare bestämmer vilken server som skall exekvera en uppgift, och vilken användares uppgift som skall allokeras. Om ett system har flera användare med liknande krav, bör resurserna tilldelas jämnlikt mellan användarna. Idag implementeras resurshanterare oftast som en centraliserad server som har information om alla servrar i klustret och de olika användarna. En centraliserad server skapar dock problem som driftstopp vid avbrott på ett enda ställe, även enbart vertikal skalning för resurshanteraren. Denna uppsats fokuserar på jämnlikhet för användare med en decentraliserad resurshanterare. En lösning föreslås, Parallel Distributed Gradient-based Dominant Resource Fairness, som tillåter servrar att hantera en delmängd av användare i systemet, detta med en liknande jämnlikhet jämförande med en centraliserad server. Lösningen använder en så kallad gradient network topology overlay för att sortera servrarna baserat på deras användares resursanvändning och tillåter en server att veta om den har användaren med lägst resursanvändning i klustret. Lösningen jämförs med existerande lösningar baserat på jämnlikhet och allokeringstid. Resultaten visar att lösningen ger en mer jämnlik allokering än existerande lösningar utifrån gini-koefficienten. Resultaten visar även att systemets skallbarhet angående allokeringstid är beroende på antalet användare i klustret eftersom det tillåter fler parallella allokeringar. Lösningen skalar inte lika bra dock som existerande distribuerade lösningar. Med 40 användare och över 100 servrar har lösningen liknande tid som en centraliserad server, och är snabbare med fler användare.

Les formines de la famille des DRF sont des puissants nucleateurs d'actine. Précédemment, nous avons montré que DRF1 participe à la capture des microtubules (MTs) au niveau du cortex cellulaire, en aval du récepteur ErbB2. Ceci impliquait le recrutement d'APC et ACF7. Dans cette étude, nous avons examiné la contribution de DRF1, DRF2 et DRF3 à la capture des MT corticaux et à la migration cellulaire ErbB2- dépendante. La déplétion individuelle de DRF1/2 ou 3 à l'aide de siRNA perturbe fortement la migration chimiotactique ErbB2-dépendante. Les DRF sont toutes trois requises pour la capture des MT au niveau du cortex cellulaire. Des mutants de DRF1 déficients pour leur association avec l'actine sont toujours actifs pour la capture des MT. Nous avons aussi pu montrer qu'une construction limitée au domaine FH2 des DRF était parfaitement fonctionnelle. Nous avons alors procéder à une recherche systématique des protéines se liant au domaine FH2, par purification d'affinité et spectrométrie de masse. Nous avons observé que les domaines FH2 de DRF1, DRF2 et DRF3 se lient à des groupes de partenaires distincts. Ainsi, seul le domaine FH2 de DRF1 lie la protéine Rab6-Interacting Protein 2 (RB6IP2). De plus, DRF1 contrôle le recrutement de RB6IP2 au cortex cellulaire et la déplétion concomitante de RB6IP2 et d'IQGAP1 perturbe fortement la capture des MT. Ces résultats démontrent l'implication de l'interaction entre DRF1 et RB6IP2 dans la capture des MT dans les cellules en migration. / Diaphanous-related formins (DRF) nucleate single linear filaments, binding to and protecting from capping their growing barbed ends. We have previously found that DRF1 participated to the tethering of microtubules (MTs) to the cell cortex, downstream of the ErbB2 receptor tyrosine kinase. This involved the recruitment of APC and ACF7. We have now further investigated the contribution of DRF1, and of the closely related DRF2 and DRF3, to the capture of cortical MTs and ErbB2-dependent breast carcinoma cell migration.Using siRNA to knock down individual DRFs, we found that depletion of DRF1/2 or3 strongly disturbed ErbB2-dependent chemotaxis. All three DRFs were required for the formation of cortical MTs, in a non-redundant manner. DRF1 mutant proteins defective for actin binding were still active for MT capture. We also found that, upon truncation of the Formin Homology (FH) 1 domain, the FH2 domain remained fully functional. In a systematic search for proteins binding to the FH2 domains via affinity purification and mass spectrometry analysis, we observed that the FH2 domains of DRF1, DRF2 and DRF3 engaged with distinct sets of proteins. For instance, only FH2 domain of DRF1 pulled down Rab6-Interacting Protein 2 (RB6IP2). Interestingly, DRF1 controlled the cortical localization of RB6IP2 and concomitant depletion of RB6IP2 and IQGAP1 strongly disturbed capture of cortical MTs, showing the involvement of the DRF1/IQGAP1/RB6IP2 interaction in MT tethering at the cell leading edge.

Controladores PID são largamente utilizados no controle de processos industriais, tanto em sistemas monovariáveis como em sistemas multivariáveis. No entanto, muitos dos controladores encontrados na indústria são mal sintonizados. Um dos métodos mais simples de sintonia de controladores PID consiste em identificar algumas grandezas, as quais se relacionam com características do processo a controlar, e em seguida aplicar fórmulas para os parâmetros dos controladores baseadas nestas grandezas. Estas grandezas são o ganho e o período críticos do processo, os quais se relacionam diretamente com o limite de estabilidade do sistema. Uma característica bastante interessante deste método é que o mesmo pode facilmente ser implementado por um controle auto-ajustável. Por este fato, métodos de auto-ajuste destes controladores têm sido bastante utilizados em sistemas monovariáveis, através do uso do método do relé como procedimento de obtenção das grandezas críticas, necessárias à sintonia dos controladores. O método do relé consiste em aplicar um controle bang-bang em malha fechada com o processo do qual se quer identificar as grandezas críticas. Este procedimento, dentro de certas condições, provoca uma oscilação sustentada na saída do processo, da qual as grandezas críticas são obtidas. Tendo como objetivo o auto-ajuste de controladores PID em sistemas multivariáveis, o método do relé também pode ser utilizado para obtenção das grandezas críticas, devidamente adequado de maneira a obter as grandezas críticas multivariáveis do processo. Diferentes procedimentos de ensaios com relés podem ser aplicados em sistemas multivariáveis, porém somente um destes procedimentos é capaz de identificar as grandezas críticas multivariáveis do processo diretamente: o ensaio descentralizado com relés. Mesmo assim, a sintonia proposta na literatura é baseada nas fórmulas monovariáveis de Ziegler-Nichols e afins, mostrando-se, muitas vezes, inadequada. Desta forma, este trabalho tem o objetivo de apresentar um método de ajuste multivariável para controladores PID descentralizados, baseado nas grandezas críticas do processo. Este método estende o método do ponto crítico para sistemas monovariáveis aos sistemas multivariáveis, através da análise multivariável do problema. A análise do método do ponto crítico para sistemas monovariáveis mostra que um controlador PI ou PID sintonizado pelas fórmulas baseadas nas grandezas críticas do processo sempre irá deslocar o ponto crítico do processo para um outro ponto no plano complexo, determinado pelas fórmulas utilizadas. Da mesma forma, o método de sintonia proposto neste trabalho desloca o ponto crítico do processo para outro ponto no plano complexo, escolhido a priori, alterando a freqüência crítica do sistema. / PID controllers are widely used in process control, in singlevariable systems as well as in multivariable ones. Yet, many of the controllers found in industry are poorly tuned. One of the simplest tuning method of PID controllers consists in identifying some values which are related to the process characteristics, and simply apply some formulae based on these quantities to determine the parameters of the controllers. Theses quantities are the ultimate gain and the ultimate period of the process, which are directly related to the system stability limit. A very interesting characteristic of this method is that it is easily implemented by an auto-tuning control. Thus, auto-tuning methods of this kind of controllers have been largely used in singlevariable systems, using the relay feedback experiment in order to obtain the ultimate quantities, which are needed to tune the controllers. The relay feedback experiment consists in applying a bang-bang control to the process from which the ultimate quantities are to be identified. This procedure, under some conditions, provides a sustained oscillation in the process’ output, from which the ultimate quantities are obtained. Aiming at auto-tuning of PID controllers in multivariable systems, the relay feedback experiment can also be used in order to get the ultimate quantities. Different relay feedback procedures can be applied to multivariable processes, but only one of these can identify the real multivariable ultimate quantities, formally considering the multivariable nature of the process: the decentralized relay feedback (DRF). However, the tuning of the controllers proposed in the literature is based on Ziegler-Nichols like formulae, what seems to be, many times, inappropriate. This work presents a multivariable tuning method of decentralized PID controllers, based on the process’ ultimate quantities. This method extends the ultimate point method used in SISO systems to multivariable ones, through multivariable analysis of the problem. The analysis of the ultimate point method used in singlevariable systems shows that a PI or PID controller tuned through formulae based on the process’ ultimate quantities will always dislocate the ultimate point to another point in the complex plane, determined by the used formulae. The tuning method proposed in this work dislocates the process’ ultimate point of a multivariable process to another point in the complex plane, chosen a priori, modifying the system’s ultimate frequency.

Controladores PID são largamente utilizados no controle de processos industriais, tanto em sistemas monovariáveis como em sistemas multivariáveis. No entanto, muitos dos controladores encontrados na indústria são mal sintonizados. Um dos métodos mais simples de sintonia de controladores PID consiste em identificar algumas grandezas, as quais se relacionam com características do processo a controlar, e em seguida aplicar fórmulas para os parâmetros dos controladores baseadas nestas grandezas. Estas grandezas são o ganho e o período críticos do processo, os quais se relacionam diretamente com o limite de estabilidade do sistema. Uma característica bastante interessante deste método é que o mesmo pode facilmente ser implementado por um controle auto-ajustável. Por este fato, métodos de auto-ajuste destes controladores têm sido bastante utilizados em sistemas monovariáveis, através do uso do método do relé como procedimento de obtenção das grandezas críticas, necessárias à sintonia dos controladores. O método do relé consiste em aplicar um controle bang-bang em malha fechada com o processo do qual se quer identificar as grandezas críticas. Este procedimento, dentro de certas condições, provoca uma oscilação sustentada na saída do processo, da qual as grandezas críticas são obtidas. Tendo como objetivo o auto-ajuste de controladores PID em sistemas multivariáveis, o método do relé também pode ser utilizado para obtenção das grandezas críticas, devidamente adequado de maneira a obter as grandezas críticas multivariáveis do processo. Diferentes procedimentos de ensaios com relés podem ser aplicados em sistemas multivariáveis, porém somente um destes procedimentos é capaz de identificar as grandezas críticas multivariáveis do processo diretamente: o ensaio descentralizado com relés. Mesmo assim, a sintonia proposta na literatura é baseada nas fórmulas monovariáveis de Ziegler-Nichols e afins, mostrando-se, muitas vezes, inadequada. Desta forma, este trabalho tem o objetivo de apresentar um método de ajuste multivariável para controladores PID descentralizados, baseado nas grandezas críticas do processo. Este método estende o método do ponto crítico para sistemas monovariáveis aos sistemas multivariáveis, através da análise multivariável do problema. A análise do método do ponto crítico para sistemas monovariáveis mostra que um controlador PI ou PID sintonizado pelas fórmulas baseadas nas grandezas críticas do processo sempre irá deslocar o ponto crítico do processo para um outro ponto no plano complexo, determinado pelas fórmulas utilizadas. Da mesma forma, o método de sintonia proposto neste trabalho desloca o ponto crítico do processo para outro ponto no plano complexo, escolhido a priori, alterando a freqüência crítica do sistema. / PID controllers are widely used in process control, in singlevariable systems as well as in multivariable ones. Yet, many of the controllers found in industry are poorly tuned. One of the simplest tuning method of PID controllers consists in identifying some values which are related to the process characteristics, and simply apply some formulae based on these quantities to determine the parameters of the controllers. Theses quantities are the ultimate gain and the ultimate period of the process, which are directly related to the system stability limit. A very interesting characteristic of this method is that it is easily implemented by an auto-tuning control. Thus, auto-tuning methods of this kind of controllers have been largely used in singlevariable systems, using the relay feedback experiment in order to obtain the ultimate quantities, which are needed to tune the controllers. The relay feedback experiment consists in applying a bang-bang control to the process from which the ultimate quantities are to be identified. This procedure, under some conditions, provides a sustained oscillation in the process’ output, from which the ultimate quantities are obtained. Aiming at auto-tuning of PID controllers in multivariable systems, the relay feedback experiment can also be used in order to get the ultimate quantities. Different relay feedback procedures can be applied to multivariable processes, but only one of these can identify the real multivariable ultimate quantities, formally considering the multivariable nature of the process: the decentralized relay feedback (DRF). However, the tuning of the controllers proposed in the literature is based on Ziegler-Nichols like formulae, what seems to be, many times, inappropriate. This work presents a multivariable tuning method of decentralized PID controllers, based on the process’ ultimate quantities. This method extends the ultimate point method used in SISO systems to multivariable ones, through multivariable analysis of the problem. The analysis of the ultimate point method used in singlevariable systems shows that a PI or PID controller tuned through formulae based on the process’ ultimate quantities will always dislocate the ultimate point to another point in the complex plane, determined by the used formulae. The tuning method proposed in this work dislocates the process’ ultimate point of a multivariable process to another point in the complex plane, chosen a priori, modifying the system’s ultimate frequency.

Controladores PID são largamente utilizados no controle de processos industriais, tanto em sistemas monovariáveis como em sistemas multivariáveis. No entanto, muitos dos controladores encontrados na indústria são mal sintonizados. Um dos métodos mais simples de sintonia de controladores PID consiste em identificar algumas grandezas, as quais se relacionam com características do processo a controlar, e em seguida aplicar fórmulas para os parâmetros dos controladores baseadas nestas grandezas. Estas grandezas são o ganho e o período críticos do processo, os quais se relacionam diretamente com o limite de estabilidade do sistema. Uma característica bastante interessante deste método é que o mesmo pode facilmente ser implementado por um controle auto-ajustável. Por este fato, métodos de auto-ajuste destes controladores têm sido bastante utilizados em sistemas monovariáveis, através do uso do método do relé como procedimento de obtenção das grandezas críticas, necessárias à sintonia dos controladores. O método do relé consiste em aplicar um controle bang-bang em malha fechada com o processo do qual se quer identificar as grandezas críticas. Este procedimento, dentro de certas condições, provoca uma oscilação sustentada na saída do processo, da qual as grandezas críticas são obtidas. Tendo como objetivo o auto-ajuste de controladores PID em sistemas multivariáveis, o método do relé também pode ser utilizado para obtenção das grandezas críticas, devidamente adequado de maneira a obter as grandezas críticas multivariáveis do processo. Diferentes procedimentos de ensaios com relés podem ser aplicados em sistemas multivariáveis, porém somente um destes procedimentos é capaz de identificar as grandezas críticas multivariáveis do processo diretamente: o ensaio descentralizado com relés. Mesmo assim, a sintonia proposta na literatura é baseada nas fórmulas monovariáveis de Ziegler-Nichols e afins, mostrando-se, muitas vezes, inadequada. Desta forma, este trabalho tem o objetivo de apresentar um método de ajuste multivariável para controladores PID descentralizados, baseado nas grandezas críticas do processo. Este método estende o método do ponto crítico para sistemas monovariáveis aos sistemas multivariáveis, através da análise multivariável do problema. A análise do método do ponto crítico para sistemas monovariáveis mostra que um controlador PI ou PID sintonizado pelas fórmulas baseadas nas grandezas críticas do processo sempre irá deslocar o ponto crítico do processo para um outro ponto no plano complexo, determinado pelas fórmulas utilizadas. Da mesma forma, o método de sintonia proposto neste trabalho desloca o ponto crítico do processo para outro ponto no plano complexo, escolhido a priori, alterando a freqüência crítica do sistema. / PID controllers are widely used in process control, in singlevariable systems as well as in multivariable ones. Yet, many of the controllers found in industry are poorly tuned. One of the simplest tuning method of PID controllers consists in identifying some values which are related to the process characteristics, and simply apply some formulae based on these quantities to determine the parameters of the controllers. Theses quantities are the ultimate gain and the ultimate period of the process, which are directly related to the system stability limit. A very interesting characteristic of this method is that it is easily implemented by an auto-tuning control. Thus, auto-tuning methods of this kind of controllers have been largely used in singlevariable systems, using the relay feedback experiment in order to obtain the ultimate quantities, which are needed to tune the controllers. The relay feedback experiment consists in applying a bang-bang control to the process from which the ultimate quantities are to be identified. This procedure, under some conditions, provides a sustained oscillation in the process’ output, from which the ultimate quantities are obtained. Aiming at auto-tuning of PID controllers in multivariable systems, the relay feedback experiment can also be used in order to get the ultimate quantities. Different relay feedback procedures can be applied to multivariable processes, but only one of these can identify the real multivariable ultimate quantities, formally considering the multivariable nature of the process: the decentralized relay feedback (DRF). However, the tuning of the controllers proposed in the literature is based on Ziegler-Nichols like formulae, what seems to be, many times, inappropriate. This work presents a multivariable tuning method of decentralized PID controllers, based on the process’ ultimate quantities. This method extends the ultimate point method used in SISO systems to multivariable ones, through multivariable analysis of the problem. The analysis of the ultimate point method used in singlevariable systems shows that a PI or PID controller tuned through formulae based on the process’ ultimate quantities will always dislocate the ultimate point to another point in the complex plane, determined by the used formulae. The tuning method proposed in this work dislocates the process’ ultimate point of a multivariable process to another point in the complex plane, chosen a priori, modifying the system’s ultimate frequency.

Yes / Background Previous research has shown that use of a dynamic-response prosthetic foot (DRF) that incorporates a small passive hydraulic ankle device (hyA-F), provides certain biomechanical benefits over using a DRF that has no ankle mechanism (rigA-F). This study investigated whether use of a hyA-F in unilateral trans-tibial amputees (UTA) additionally provides metabolic energy expenditure savings and increases the symmetry in walking kinematics, compared to rigA-F. Methods Nine active UTA completed treadmill walking trials at zero gradient (at 0.8, 1.0, 1.2, 1.4, and 1.6 of customary walking speed) and for customary walking speed only, at two angles of decline (5° and 10°). The metabolic cost of locomotion was determined using respirometry. To gain insights into the source of any metabolic savings, 3D motion capture was used to determine segment kinematics, allowing body centre of mass dynamics (BCoM), differences in inter-limb symmetry and potential for energy recovery through pendulum-like motion to be quantified for each foot type. Results During both level and decline walking, use of a hyA-F compared to rigA-F significantly reduced the total mechanical work and increased the interchange between the mechanical energies of the BCoM (recovery index), leading to a significant reduction in the metabolic energy cost of locomotion, and hence an associated increase in locomotor efficiency (p < 0.001). It also increased inter-limb symmetry (medio-lateral and progression axes, particularly when walking on a 10° decline), highlighting the improvements in gait were related to a lessening of the kinematic compensations evident when using the rigA-F. Conclusions Findings suggest that use of a DRF that incorporates a small passive hydraulic ankle device will deliver improvements in metabolic energy expenditure and kinematics and thus should provide clinically meaningful benefits to UTAs’ everyday locomotion, particularly for those who are able to walk at a range of speeds and over different terrains. / Engineering and Physical Sciences Research Council(EPSRC, reference EP/H010491/1).

Utilização de experimentos com relés para identificação de pontos da resposta em freqüência de sistemas multivariáveis quadrados e sintonia de controladores PID descentralizados. / It use relay experiments for square multivariable system frequency response and decentralized PID controller tunning.

Utilização de experimentos com relés para identificação de pontos da resposta em freqüência de sistemas multivariáveis quadrados e sintonia de controladores PID descentralizados. / It use relay experiments for square multivariable system frequency response and decentralized PID controller tunning.

Concurrent programs are pervasive owing to the increasing adoption of multi-core systems across the entire computing spectrum. However, the large set of possible program behaviors make it difficult to write correct and efficient con-current programs. This also makes the formal and automated analysis of such programs a hard problem. Thus, concurrent programs provide fertile grounds for a large class of insidious defects. Static analysis techniques infer semantic properties of programs without executing them. They are attractive because they are sound (they can guarantee the absence of bugs), can execute with a fair degree of automation, and do not depend on test cases. However, current static analyses techniques for concurrent programs are either precise and prohibitively slow, or fast but imprecise. In this thesis, we partially address this problem by designing efficient static analyses for concurrent programs. In the first part of the thesis, we provide a framework for designing and proving the correctness of data flow analysis for race free multi-threaded programs. The resulting analyses are in the same spirit as the \sync-CFG" analysis, originally proposed in De et al, 2011. Using novel thread-local semantics as starting points, we devise abstract analyses which treat a concurrent program as if it were sequential. We instantiate these abstractions to devise efficient relational analyses for race free programs, which we have implemented in a prototype tool called RATCOP. On the benchmarks, RATCOP was fairly precise and fast. In a comparative study with a recent concurrent static analyzer, RATCOP was up to 5 orders of magnitude faster. In the second part of the thesis, we propose a technique for detecting all high-level data races in a system library, like the kernel API of a real-time operating system (RTOS) that relies on ag-based scheduling and synchronization. Such races are good indicators of atomicity violations. Using our technique, a user is able to soundly disregard 99:8% of an estimated 41; 000 potential high-level races. Our tool detected 38 high-level data races in FreeRTOS (a popular OS in the embedded systems domain), out of which 16 were harmful.

Utilização de experimentos com relés para identificação de pontos da resposta em freqüência de sistemas multivariáveis quadrados e sintonia de controladores PID descentralizados. / It use relay experiments for square multivariable system frequency response and decentralized PID controller tunning.