Relating goal scheduling, precedence, and memory management in and-parallel execution of logic programs

The interactions among three important issues involved in the implementation of logic programs in parallel (goal scheduling, precedence, and memory management) are discussed. A simplified, parallel memory management model and an efficient, load-balancing goal scheduling strategy are presented. It is shown how, for systems which support "don't know" non-determinism, special care has to be taken during goal scheduling if the space recovery characteristics of sequential systems are to be preserved. A solution based on selecting only "newer" goals for execution is described, and an algorithm is proposed for efficiently maintaining and determining precedence relationships and variable ages across parallel goals. It is argued that the proposed schemes and algorithms make it possible to extend the storage performance of sequential systems to parallel execution without the considerable overhead previously associated with it. The results are applicable to a wide class of parallel and coroutining systems, and they represent an efficient alternative to "all heap" or "spaghetti stack" allocation models.

Advanced Topics in Resource Analysis: Certification, Incrementality, Concurrency and Array-Sensitivity

El Análisis de Consumo de Recursos o Análisis de Coste trata de aproximar el coste de ejecutar un programa como una función dependiente de sus datos de entrada. A pesar de que existen trabajos previos a esta tesis doctoral que desarrollan potentes marcos para el análisis de coste de programas orientados a objetos, algunos aspectos avanzados, como la eficiencia, la precisión y la fiabilidad de los resultados, todavía deben ser estudiados en profundidad. Esta tesis aborda estos aspectos desde cuatro perspectivas diferentes: (1) Las estructuras de datos compartidas en la memoria del programa son una pesadilla para el análisis estático de programas. Trabajos recientes proponen una serie de condiciones de localidad para poder mantener de forma consistente información sobre los atributos de los objetos almacenados en memoria compartida, reemplazando éstos por variables locales no almacenadas en la memoria compartida. En esta tesis presentamos dos extensiones a estos trabajos: la primera es considerar, no sólo los accesos a los atributos, sino también los accesos a los elementos almacenados en arrays; la segunda se centra en los casos en los que las condiciones de localidad no se cumplen de forma incondicional, para lo cual, proponemos una técnica para encontrar las precondiciones necesarias para garantizar la consistencia de la información acerca de los datos almacenados en memoria. (2) El objetivo del análisis incremental es, dado un programa, los resultados de su análisis y una serie de cambios sobre el programa, obtener los nuevos resultados del análisis de la forma más eficiente posible, evitando reanalizar aquellos fragmentos de código que no se hayan visto afectados por los cambios. Los analizadores actuales todavía leen y analizan el programa completo de forma no incremental. Esta tesis presenta un análisis de coste incremental, que, dado un cambio en el programa, reconstruye la información sobre el coste del programa de todos los métodos afectados por el cambio de forma incremental. Para esto, proponemos (i) un algoritmo multi-dominio y de punto fijo que puede ser utilizado en todos los análisis globales necesarios para inferir el coste, y (ii) una novedosa forma de almacenar las expresiones de coste que nos permite reconstruir de forma incremental únicamente las funciones de coste de aquellos componentes afectados por el cambio. (3) Las garantías de coste obtenidas de forma automática por herramientas de análisis estático no son consideradas totalmente fiables salvo que la implementación de la herramienta o los resultados obtenidos sean verificados formalmente. Llevar a cabo el análisis de estas herramientas es una tarea titánica, ya que se trata de herramientas de gran tamaño y complejidad. En esta tesis nos centramos en el desarrollo de un marco formal para la verificación de las garantías de coste obtenidas por los analizadores en lugar de analizar las herramientas. Hemos implementado esta idea mediante la herramienta COSTA, un analizador de coste para programas Java y KeY, una herramienta de verificación de programas Java. De esta forma, COSTA genera las garantías de coste, mientras que KeY prueba la validez formal de los resultados obtenidos, generando de esta forma garantías de coste verificadas. (4) Hoy en día la concurrencia y los programas distribuidos son clave en el desarrollo de software. Los objetos concurrentes son un modelo de concurrencia asentado para el desarrollo de sistemas concurrentes. En este modelo, los objetos son las unidades de concurrencia y se comunican entre ellos mediante llamadas asíncronas a sus métodos. La distribución de las tareas sugiere que el análisis de coste debe inferir el coste de los diferentes componentes distribuidos por separado. En esta tesis proponemos un análisis de coste sensible a objetos que, utilizando los resultados obtenidos mediante un análisis de apunta-a, mantiene el coste de los diferentes componentes de forma independiente. Abstract Resource Analysis (a.k.a. Cost Analysis) tries to approximate the cost of executing programs as functions on their input data sizes and without actually having to execute the programs. While a powerful resource analysis framework on object-oriented programs existed before this thesis, advanced aspects to improve the efficiency, the accuracy and the reliability of the results of the analysis still need to be further investigated. This thesis tackles this need from the following four different perspectives. (1) Shared mutable data structures are the bane of formal reasoning and static analysis. Analyses which keep track of heap-allocated data are referred to as heap-sensitive. Recent work proposes locality conditions for soundly tracking field accesses by means of ghost non-heap allocated variables. In this thesis we present two extensions to this approach: the first extension is to consider arrays accesses (in addition to object fields), while the second extension focuses on handling cases for which the locality conditions cannot be proven unconditionally by finding aliasing preconditions under which tracking such heap locations is feasible. (2) The aim of incremental analysis is, given a program, its analysis results and a series of changes to the program, to obtain the new analysis results as efficiently as possible and, ideally, without having to (re-)analyze fragments of code that are not affected by the changes. During software development, programs are permanently modified but most analyzers still read and analyze the entire program at once in a non-incremental way. This thesis presents an incremental resource usage analysis which, after a change in the program is made, is able to reconstruct the upper-bounds of all affected methods in an incremental way. To this purpose, we propose (i) a multi-domain incremental fixed-point algorithm which can be used by all global analyses required to infer the cost, and (ii) a novel form of cost summaries that allows us to incrementally reconstruct only those components of cost functions affected by the change. (3) Resource guarantees that are automatically inferred by static analysis tools are generally not considered completely trustworthy, unless the tool implementation or the results are formally verified. Performing full-blown verification of such tools is a daunting task, since they are large and complex. In this thesis we focus on the development of a formal framework for the verification of the resource guarantees obtained by the analyzers, instead of verifying the tools. We have implemented this idea using COSTA, a state-of-the-art cost analyzer for Java programs and KeY, a state-of-the-art verification tool for Java source code. COSTA is able to derive upper-bounds of Java programs while KeY proves the validity of these bounds and provides a certificate. The main contribution of our work is to show that the proposed tools cooperation can be used for automatically producing verified resource guarantees. (4) Distribution and concurrency are today mainstream. Concurrent objects form a well established model for distributed concurrent systems. In this model, objects are the concurrency units that communicate via asynchronous method calls. Distribution suggests that analysis must infer the cost of the diverse distributed components separately. In this thesis we propose a novel object-sensitive cost analysis which, by using the results gathered by a points-to analysis, can keep the cost of the diverse distributed components separate.

Towards extracting non-strict independent and-parallelism using sharing and freeness information

Logic programming systems which exploit and-parallelism among non-deterministic goals rely on notions of independence among those goals in order to ensure certain efficiency properties. "Non-strict" independence (NSI) is a more relaxed notion than the traditional notion of "strict" independence (SI) which still ensures the relevant efficiency properties and can allow considerable more parallelism than SI. However, all compilation technology developed to date has been based on SI, presumably because of the intrinsic complexity of exploiting NSI. This is related to the fact that NSI cannot be determined "a priori" as SI. This paper fills this gap by developing a technique for compile-time detection and annotation of NSI. It also proposes algorithms for combined compile- time/run-time detection, presenting novel run-time checks for this type of parallelism. Also, a transformation procedure to eliminate shared variables among parallel goals is presented, attempting to perform as much work as possible at compiletime. The approach is based on the knowledge of certain properties about run-time instantiations of program variables —sharing and freeness— for which compile-time technology is available, with new approaches being currently proposed.

GreenDisc: a HW/SW energy optimization framework in globally distributed computation

In recent future, wireless sensor networks (WSNs) will experience a broad high-scale deployment (millions of nodes in the national area) with multiple information sources per node, and with very specific requirements for signal processing. In parallel, the broad range deployment of WSNs facilitates the definition and execution of ambitious studies, with a large input data set and high computational complexity. These computation resources, very often heterogeneous and driven on-demand, can only be satisfied by high-performance Data Centers (DCs). The high economical and environmental impact of the energy consumption in DCs requires aggressive energy optimization policies. These policies have been already detected but not successfully proposed. In this context, this paper shows the following on-going research lines and obtained results. In the field of WSNs: energy optimization in the processing nodes from different abstraction levels, including reconfigurable application specific architectures, efficient customization of the memory hierarchy, energy-aware management of the wireless interface, and design automation for signal processing applications. In the field of DCs: energy-optimal workload assignment policies in heterogeneous DCs, resource management policies with energy consciousness, and efficient cooling mechanisms that will cooperate in the minimization of the electricity bill of the DCs that process the data provided by the WSNs.

GreenDisc: a HW/SW energy optimization framework in globally distributed computation

In recent future, wireless sensor networks ({WSNs}) will experience a broad high-scale deployment (millions of nodes in the national area) with multiple information sources per node, and with very specific requirements for signal processing. In parallel, the broad range deployment of {WSNs} facilitates the definition and execution of ambitious studies, with a large input data set and high computational complexity. These computation resources, very often heterogeneous and driven on-demand, can only be satisfied by high-performance Data Centers ({DCs}). The high economical and environmental impact of the energy consumption in {DCs} requires aggressive energy optimization policies. These policies have been already detected but not successfully proposed. In this context, this paper shows the following on-going research lines and obtained results. In the field of {WSNs}: energy optimization in the processing nodes from different abstraction levels, including reconfigurable application specific architectures, efficient customization of the memory hierarchy, energy-aware management of the wireless interface, and design automation for signal processing applications. In the field of {DCs}: energy-optimal workload assignment policies in heterogeneous {DCs}, resource management policies with energy consciousness, and efficient cooling mechanisms that will cooperate in the minimization of the electricity bill of the DCs that process the data provided by the WSNs.

Chaos in non lineal Alfven waves using the DNLS equations

The electro-dynamical tethers emit waves in structured denominated Alfven wings. The Derivative Nonlineal Schrödinger Equation (DNLS) possesses the capacity to describe the propagation of circularly polarized Alfven waves of finite amplitude in cold plasmas. The DNLS equation is truncated to explore the coherent, weakly nonlinear, cubic coupling of three waves near resonance, one wave being linearly unstable and the other waves damped. In this article is presented a theoretical and numerical analysis when the growth rate of the unstable wave is next to zero considering two damping models: Landau and resistive. The DNLS equation presents a chaotic dynamics when is consider only three wave truncation. The evolution to chaos possesses three routes: hard transition, period-doubling and intermittence of type I.

A taxonomy of the quality attributes for distributed applications

The software engineering community has paid little attention to non-functional requirements, or quality attributes, compared with studies performed on capture, analysis and validation of functional requirements. This circumstance becomes more intense in the case of distributed applications. In these applications we have to take into account, besides the quality attributes such as correctness, robustness, extendibility, reusability, compatibility, efficiency, portability and ease of use, others like reliability, scalability, transparency, security, interoperability, concurrency, etc. In this work we will show how these last attributes are related to different abstractions that coexist in the problem domain. To achieve this goal, we have established a taxonomy of quality attributes of distributed applications and have determined the set of necessary services to support such attributes.

Time-resolved evolution of coherent structures in turbulent channels

El objetivo de esta tesis es estudiar la dinámica de la capa logarítmica de flujos turbulentos de pared. En concreto, proponemos un nuevo modelo estructural utilizando diferentes tipos de estructuras coherentes: sweeps, eyecciones, grupos de vorticidad y streaks. La herramienta utilizada es la simulación numérica directa de canales turbulentos. Desde los primeros trabajos de Theodorsen (1952), las estructuras coherentes han jugado un papel fundamental para entender la organización y dinámica de los flujos turbulentos. A día de hoy, datos procedentes de simulaciones numéricas directas obtenidas en instantes no contiguos permiten estudiar las propiedades fundamentales de las estructuras coherentes tridimensionales desde un punto de vista estadístico. Sin embargo, la dinámica no puede ser entendida en detalle utilizando sólo instantes aislados en el tiempo, sino que es necesario seguir de forma continua las estructuras. Aunque existen algunos estudios sobre la evolución temporal de las estructuras más pequeñas a números de Reynolds moderados, por ejemplo Robinson (1991), todavía no se ha realizado un estudio completo a altos números de Reynolds y para todas las escalas presentes de la capa logarítmica. El objetivo de esta tesis es llevar a cabo dicho análisis. Los problemas más interesantes los encontramos en la región logarítmica, donde residen las cascadas de vorticidad, energía y momento. Existen varios modelos que intentan explicar la organización de los flujos turbulentos en dicha región. Uno de los más extendidos fue propuesto por Adrian et al. (2000) a través de observaciones experimentales y considerando como elemento fundamental paquetes de vórtices con forma de horquilla que actúan de forma cooperativa para generar rampas de bajo momento. Un modelo alternativo fué ideado por del Álamo & Jiménez (2006) utilizando datos numéricos. Basado también en grupos de vorticidad, planteaba un escenario mucho más desorganizado y con estructuras sin forma de horquilla. Aunque los dos modelos son cinemáticamente similares, no lo son desde el punto de vista dinámico, en concreto en lo que se refiere a la importancia que juega la pared en la creación y vida de las estructuras. Otro punto importante aún sin resolver se refiere al modelo de cascada turbulenta propuesto por Kolmogorov (1941b), y su relación con estructuras coherentes medibles en el flujo. Para dar respuesta a las preguntas anteriores, hemos desarrollado un nuevo método que permite seguir estructuras coherentes en el tiempo y lo hemos aplicado a simulaciones numéricas de canales turbulentos con números de Reynolds lo suficientemente altos como para tener un rango de escalas no trivial y con dominios computacionales lo suficientemente grandes como para representar de forma correcta la dinámica de la capa logarítmica. Nuestros esfuerzos se han desarrollado en cuatro pasos. En primer lugar, hemos realizado una campaña de simulaciones numéricas directas a diferentes números de Reynolds y tamaños de cajas para evaluar el efecto del dominio computacional en las estadísticas de primer orden y el espectro. A partir de los resultados obtenidos, hemos concluido que simulaciones con cajas de longitud 2vr y ancho vr veces la semi-altura del canal son lo suficientemente grandes para reproducir correctamente las interacciones entre estructuras coherentes de la capa logarítmica y el resto de escalas. Estas simulaciones son utilizadas como punto de partida en los siguientes análisis. En segundo lugar, las estructuras coherentes correspondientes a regiones con esfuerzos de Reynolds tangenciales intensos (Qs) en un canal turbulento han sido estudiadas extendiendo a tres dimensiones el análisis de cuadrantes, con especial énfasis en la capa logarítmica y la región exterior. Las estructuras coherentes han sido identificadas como regiones contiguas del espacio donde los esfuerzos de Reynolds tangenciales son más intensos que un cierto nivel. Los resultados muestran que los Qs separados de la pared están orientados de forma isótropa y su contribución neta al esfuerzo de Reynolds medio es nula. La mayor contribución la realiza una familia de estructuras de mayor tamaño y autosemejantes cuya parte inferior está muy cerca de la pared (ligada a la pared), con una geometría compleja y dimensión fractal « 2. Estas estructuras tienen una forma similar a una ‘esponja de placas’, en comparación con los grupos de vorticidad que tienen forma de ‘esponja de cuerdas’. Aunque el número de objetos decae al alejarnos de la pared, la fracción de esfuerzos de Reynolds que contienen es independiente de su altura, y gran parte reside en unas pocas estructuras que se extienden más allá del centro del canal, como en las grandes estructuras propuestas por otros autores. Las estructuras dominantes en la capa logarítmica son parejas de sweeps y eyecciones uno al lado del otro y con grupos de vorticidad asociados que comparten las dimensiones y esfuerzos con los remolinos ligados a la pared propuestos por Townsend. En tercer lugar, hemos estudiado la evolución temporal de Qs y grupos de vorticidad usando las simulaciones numéricas directas presentadas anteriormente hasta números de Reynolds ReT = 4200 (Reynolds de fricción). Las estructuras fueron identificadas siguiendo el proceso descrito en el párrafo anterior y después seguidas en el tiempo. A través de la interseción geométrica de estructuras pertenecientes a instantes de tiempo contiguos, hemos creado gratos de conexiones temporales entre todos los objetos y, a partir de ahí, definido ramas primarias y secundarias, de tal forma que cada rama representa la evolución temporal de una estructura coherente. Una vez que las evoluciones están adecuadamente organizadas, proporcionan toda la información necesaria para caracterizar la historia de las estructuras desde su nacimiento hasta su muerte. Los resultados muestran que las estructuras nacen a todas las distancias de la pared, pero con mayor probabilidad cerca de ella, donde la cortadura es más intensa. La mayoría mantienen tamaños pequeños y no viven mucho tiempo, sin embargo, existe una familia de estructuras que crecen lo suficiente como para ligarse a la pared y extenderse a lo largo de la capa logarítmica convirtiéndose en las estructuras observas anteriormente y descritas por Townsend. Estas estructuras son geométricamente autosemejantes con tiempos de vida proporcionales a su tamaño. La mayoría alcanzan tamaños por encima de la escala de Corrsin, y por ello, su dinámica está controlada por la cortadura media. Los resultados también muestran que las eyecciones se alejan de la pared con velocidad media uT (velocidad de fricción) y su base se liga a la pared muy rápidamente al inicio de sus vidas. Por el contrario, los sweeps se mueven hacia la pared con velocidad -uT y se ligan a ella más tarde. En ambos casos, los objetos permanecen ligados a la pared durante 2/3 de sus vidas. En la dirección de la corriente, las estructuras se desplazan a velocidades cercanas a la convección media del flujo y son deformadas por la cortadura. Finalmente, hemos interpretado la cascada turbulenta, no sólo como una forma conceptual de organizar el flujo, sino como un proceso físico en el cual las estructuras coherentes se unen y se rompen. El volumen de una estructura cambia de forma suave, cuando no se une ni rompe, o lo hace de forma repentina en caso contrario. Los procesos de unión y rotura pueden entenderse como una cascada directa (roturas) o inversa (uniones), siguiendo el concepto de cascada de remolinos ideado por Richardson (1920) y Obukhov (1941). El análisis de los datos muestra que las estructuras con tamaños menores a 30η (unidades de Kolmogorov) nunca se unen ni rompen, es decir, no experimentan el proceso de cascada. Por el contrario, aquellas mayores a 100η siempre se rompen o unen al menos una vez en su vida. En estos casos, el volumen total ganado y perdido es una fracción importante del volumen medio de la estructura implicada, con una tendencia ligeramente mayor a romperse (cascada directa) que a unirse (cascade inversa). La mayor parte de interacciones entre ramas se debe a roturas o uniones de fragmentos muy pequeños en la escala de Kolmogorov con estructuras más grandes, aunque el efecto de fragmentos de mayor tamaño no es despreciable. También hemos encontrado que las roturas tienen a ocurrir al final de la vida de la estructura y las uniones al principio. Aunque los resultados para la cascada directa e inversa no son idénticos, son muy simétricos, lo que sugiere un alto grado de reversibilidad en el proceso de cascada. ABSTRACT The purpose of the present thesis is to study the dynamics of the logarithmic layer of wall-bounded turbulent flows. Specifically, to propose a new structural model based on four different coherent structures: sweeps, ejections, clusters of vortices and velocity streaks. The tool used is the direct numerical simulation of time-resolved turbulent channels. Since the first work by Theodorsen (1952), coherent structures have played an important role in the understanding of turbulence organization and its dynamics. Nowadays, data from individual snapshots of direct numerical simulations allow to study the threedimensional statistical properties of those objects, but their dynamics can only be fully understood by tracking them in time. Although the temporal evolution has already been studied for small structures at moderate Reynolds numbers, e.g., Robinson (1991), a temporal analysis of three-dimensional structures spanning from the smallest to the largest scales across the logarithmic layer has yet to be performed and is the goal of the present thesis. The most interesting problems lie in the logarithmic region, which is the seat of cascades of vorticity, energy, and momentum. Different models involving coherent structures have been proposed to represent the organization of wall-bounded turbulent flows in the logarithmic layer. One of the most extended ones was conceived by Adrian et al. (2000) and built on packets of hairpins that grow from the wall and work cooperatively to gen- ´ erate low-momentum ramps. A different view was presented by del Alamo & Jim´enez (2006), who extracted coherent vortical structures from DNSs and proposed a less organized scenario. Although the two models are kinematically fairly similar, they have important dynamical differences, mostly regarding the relevance of the wall. Another open question is whether such a model can be used to explain the cascade process proposed by Kolmogorov (1941b) in terms of coherent structures. The challenge would be to identify coherent structures undergoing a turbulent cascade that can be quantified. To gain a better insight into the previous questions, we have developed a novel method to track coherent structures in time, and used it to characterize the temporal evolutions of eddies in turbulent channels with Reynolds numbers high enough to include a non-trivial range of length scales, and computational domains sufficiently long and wide to reproduce correctly the dynamics of the logarithmic layer. Our efforts have followed four steps. First, we have conducted a campaign of direct numerical simulations of turbulent channels at different Reynolds numbers and box sizes, and assessed the effect of the computational domain in the one-point statistics and spectra. From the results, we have concluded that computational domains with streamwise and spanwise sizes 2vr and vr times the half-height of the channel, respectively, are large enough to accurately capture the dynamical interactions between structures in the logarithmic layer and the rest of the scales. These simulations are used in the subsequent chapters. Second, the three-dimensional structures of intense tangential Reynolds stress in plane turbulent channels (Qs) have been studied by extending the classical quadrant analysis to three dimensions, with emphasis on the logarithmic and outer layers. The eddies are identified as connected regions of intense tangential Reynolds stress. Qs are then classified according to their streamwise and wall-normal fluctuating velocities as inward interactions, outward interactions, sweeps and ejections. It is found that wall-detached Qs are isotropically oriented background stress fluctuations, common to most turbulent flows, and do not contribute to the mean stress. Most of the stress is carried by a selfsimilar family of larger wall-attached Qs, increasingly complex away from the wall, with fractal dimensions « 2. They have shapes similar to ‘sponges of flakes’, while vortex clusters resemble ‘sponges of strings’. Although their number decays away from the wall, the fraction of the stress that they carry is independent of their heights, and a substantial part resides in a few objects extending beyond the centerline, reminiscent of the very large scale motions of several authors. The predominant logarithmic-layer structures are sideby- side pairs of sweeps and ejections, with an associated vortex cluster, and dimensions and stresses similar to Townsend’s conjectured wall-attached eddies. Third, the temporal evolution of Qs and vortex clusters are studied using time-resolved DNS data up to ReT = 4200 (friction Reynolds number). The eddies are identified following the procedure presented above, and then tracked in time. From the geometric intersection of structures in consecutive fields, we have built temporal connection graphs of all the objects, and defined main and secondary branches in a way that each branch represents the temporal evolution of one coherent structure. Once these evolutions are properly organized, they provide the necessary information to characterize eddies from birth to death. The results show that the eddies are born at all distances from the wall, although with higher probability near it, where the shear is strongest. Most of them stay small and do not last for long times. However, there is a family of eddies that become large enough to attach to the wall while they reach into the logarithmic layer, and become the wall-attached structures previously observed in instantaneous flow fields. They are geometrically self-similar, with sizes and lifetimes proportional to their distance from the wall. Most of them achieve lengths well above the Corrsin’ scale, and hence, their dynamics are controlled by the mean shear. Eddies associated with ejections move away from the wall with an average velocity uT (friction velocity), and their base attaches very fast at the beginning of their lives. Conversely, sweeps move towards the wall at -uT, and attach later. In both cases, they remain attached for 2/3 of their lives. In the streamwise direction, eddies are advected and deformed by the local mean velocity. Finally, we interpret the turbulent cascade not only as a way to conceptualize the flow, but as an actual physical process in which coherent structures merge and split. The volume of an eddy can change either smoothly, when they are not merging or splitting, or through sudden changes. The processes of merging and splitting can be thought of as a direct (when splitting) or an inverse (when merging) cascade, following the ideas envisioned by Richardson (1920) and Obukhov (1941). It is observed that there is a minimum length of 30η (Kolmogorov units) above which mergers and splits begin to be important. Moreover, all eddies above 100η split and merge at least once in their lives. In those cases, the total volume gained and lost is a substantial fraction of the average volume of the structure involved, with slightly more splits (direct cascade) than mergers. Most branch interactions are found to be the shedding or absorption of Kolmogorov-scale fragments by larger structures, but more balanced splits or mergers spanning a wide range of scales are also found to be important. The results show that splits are more probable at the end of the life of the eddy, while mergers take place at the beginning of the life. Although the results for the direct and the inverse cascades are not identical, they are found to be very symmetric, which suggests a high degree of reversibility of the cascade process.

Distributed, layered and reliable computing nets to represent neuronal receptive fields

Abstract. Receptive fields of retinal and other sensory neurons show a large variety of spatiotemporal linear and non linear types of responses to local stimuli. In visual neurons, these responses present either asymmetric sensitive zones or center-surround organization. In most cases, the nature of the responses suggests the existence of a kind of distributed computation prior to the integration by the final cell which is evidently supported by the anatomy. We describe a new kind of discrete and continuous filters to model the kind of computations taking place in the receptive fields of retinal cells. To show their performance in the analysis of diferent non-trivial neuron-like structures, we use a computer tool specifically programmed by the authors to that efect. This tool is also extended to study the efect of lesions on the whole performance of our model nets.

Cognitive strategies for reducing energy consumption in wireless sensor networks

El consumo energético de las Redes de Sensores Inalámbricas (WSNs por sus siglas en inglés) es un problema histórico que ha sido abordado desde diferentes niveles y visiones, ya que no solo afecta a la propia supervivencia de la red sino que el creciente uso de dispositivos inteligentes y el nuevo paradigma del Internet de las Cosas hace que las WSNs tengan cada vez una mayor influencia en la huella energética. Debido a la tendencia al alza en el uso de estas redes se añade un nuevo problema, la saturación espectral. Las WSNs operan habitualmente en bandas sin licencia como son las bandas Industrial, Científica y Médica (ISM por sus siglas en inglés). Estas bandas se comparten con otro tipo de redes como Wi-Fi o Bluetooth cuyo uso ha crecido exponencialmente en los últimos años. Para abordar este problema aparece el paradigma de la Radio Cognitiva (CR), una tecnología que permite el acceso oportunista al espectro. La introducción de capacidades cognitivas en las WSNs no solo permite optimizar su eficiencia espectral sino que también tiene un impacto positivo en parámetros como la calidad de servicio, la seguridad o el consumo energético. Sin embargo, por otra parte, este nuevo paradigma plantea algunos retos relacionados con el consumo energético. Concretamente, el sensado del espectro, la colaboración entre los nodos (que requiere comunicación adicional) y el cambio en los parámetros de transmisión aumentan el consumo respecto a las WSN clásicas. Teniendo en cuenta que la investigación en el campo del consumo energético ha sido ampliamente abordada puesto que se trata de una de sus principales limitaciones, asumimos que las nuevas estrategias deben surgir de las nuevas capacidades añadidas por las redes cognitivas. Por otro lado, a la hora de diseñar estrategias de optimización para CWSN hay que tener muy presentes las limitaciones de recursos de estas redes en cuanto a memoria, computación y consumo energético de los nodos. En esta tesis doctoral proponemos dos estrategias de reducción de consumo energético en CWSNs basadas en tres pilares fundamentales. El primero son las capacidades cognitivas añadidas a las WSNs que proporcionan la posibilidad de adaptar los parámetros de transmisión en función del espectro disponible. La segunda es la colaboración, como característica intrínseca de las CWSNs. Finalmente, el tercer pilar de este trabajo es teoría de juegos como algoritmo de soporte a la decisión, ampliamente utilizado en WSNs debido a su simplicidad. Como primer aporte de la tesis se presenta un análisis completo de las posibilidades introducidas por la radio cognitiva en materia de reducción de consumo para WSNs. Gracias a las conclusiones extraídas de este análisis, se han planteado las hipótesis de esta tesis relacionadas con la validez de usar capacidades cognitivas como herramienta para la reducción de consumo en CWSNs. Una vez presentada las hipótesis, pasamos a desarrollar las principales contribuciones de la tesis: las dos estrategias diseñadas para reducción de consumo basadas en teoría de juegos y CR. La primera de ellas hace uso de un juego no cooperativo que se juega mediante pares de jugadores. En la segunda estrategia, aunque el juego continúa siendo no cooperativo, se añade el concepto de colaboración. Para cada una de las estrategias se presenta el modelo del juego, el análisis formal de equilibrios y óptimos y la descripción de la estrategia completa donde se incluye la interacción entre nodos. Con el propósito de probar las estrategias mediante simulación e implementación en dispositivos reales hemos desarrollado un marco de pruebas compuesto por un simulador cognitivo y un banco de pruebas formado por nodos cognitivos capaces de comunicarse en tres bandas ISM desarrollados en el B105 Lab. Este marco de pruebas constituye otra de las aportaciones de la tesis que permitirá el avance en la investigación en el área de las CWSNs. Finalmente, se presentan y discuten los resultados derivados de la prueba de las estrategias desarrolladas. La primera estrategia proporciona ahorros de energía mayores al 65% comparados con una WSN sin capacidades cognitivas y alrededor del 25% si la comparamos con una estrategia cognitiva basada en el sensado periódico del espectro para el cambio de canal de acuerdo a un nivel de ruido fijado. Este algoritmo se comporta de forma similar independientemente del nivel de ruido siempre que éste sea espacialmente uniformemente. Esta estrategia, a pesar de su sencillez, nos asegura el comportamiento óptimo en cuanto a consumo energético debido a la utilización de teoría de juegos en la fase de diseño del comportamiento de los nodos. La estrategia colaborativa presenta mejoras respecto a la anterior en términos de protección frente al ruido en escenarios de ruido más complejos donde aporta una mejora del 50% comparada con la estrategia anterior. ABSTRACT Energy consumption in Wireless Sensor Networks (WSNs) is a known historical problem that has been addressed from different areas and on many levels. But this problem should not only be approached from the point of view of their own efficiency for survival. A major portion of communication traffic has migrated to mobile networks and systems. The increased use of smart devices and the introduction of the Internet of Things (IoT) give WSNs a great influence on the carbon footprint. Thus, optimizing the energy consumption of wireless networks could reduce their environmental impact considerably. In recent years, another problem has been added to the equation: spectrum saturation. Wireless Sensor Networks usually operate in unlicensed spectrum bands such as Industrial, Scientific, and Medical (ISM) bands shared with other networks (mainly Wi-Fi and Bluetooth). To address the efficient spectrum utilization problem, Cognitive Radio (CR) has emerged as the key technology that enables opportunistic access to the spectrum. Therefore, the introduction of cognitive capabilities to WSNs allows optimizing their spectral occupation. Cognitive Wireless Sensor Networks (CWSNs) do not only increase the reliability of communications, but they also have a positive impact on parameters such as the Quality of Service (QoS), network security, or energy consumption. These new opportunities introduced by CWSNs unveil a wide field in the energy consumption research area. However, this also implies some challenges. Specifically, the spectrum sensing stage, collaboration among devices (which requires extra communication), and changes in the transmission parameters increase the total energy consumption of the network. When designing CWSN optimization strategies, the fact that WSN nodes are very limited in terms of memory, computational power, or energy consumption has to be considered. Thus, light strategies that require a low computing capacity must be found. Since the field of energy conservation in WSNs has been widely explored, we assume that new strategies could emerge from the new opportunities presented by cognitive networks. In this PhD Thesis, we present two strategies for energy consumption reduction in CWSNs supported by three main pillars. The first pillar is that cognitive capabilities added to the WSN provide the ability to change the transmission parameters according to the spectrum. The second pillar is that the ability to collaborate is a basic characteristic of CWSNs. Finally, the third pillar for this work is the game theory as a decision-making algorithm, which has been widely used in WSNs due to its lightness and simplicity that make it valid to operate in CWSNs. For the development of these strategies, a complete analysis of the possibilities is first carried out by incorporating the cognitive abilities into the network. Once this analysis has been performed, we expose the hypotheses of this thesis related to the use of cognitive capabilities as a useful tool to reduce energy consumption in CWSNs. Once the analyses are exposed, we present the main contribution of this thesis: the two designed strategies for energy consumption reduction based on game theory and cognitive capabilities. The first one is based on a non-cooperative game played between two players in a simple and selfish way. In the second strategy, the concept of collaboration is introduced. Despite the fact that the game used is also a non-cooperative game, the decisions are taken through collaboration. For each strategy, we present the modeled game, the formal analysis of equilibrium and optimum, and the complete strategy describing the interaction between nodes. In order to test the strategies through simulation and implementation in real devices, we have developed a CWSN framework composed by a CWSN simulator based on Castalia and a testbed based on CWSN nodes able to communicate in three different ISM bands. We present and discuss the results derived by the energy optimization strategies. The first strategy brings energy improvement rates of over 65% compared to WSN without cognitive techniques. It also brings energy improvement rates of over 25% compared with sensing strategies for changing channels based on a decision threshold. We have also seen that the algorithm behaves similarly even with significant variations in the level of noise while working in a uniform noise scenario. The collaborative strategy presents improvements respecting the previous strategy in terms of noise protection when the noise scheme is more complex where this strategy shows improvement rates of over 50%.

Agreement Abstractions in Anonymous and Homonymous Distributed Systems Prone to Failures

The distributed computing models typically assume every process in the system has a distinct identifier (ID) or each process is programmed differently, which is named as eponymous system. In such kind of distributed systems, the unique ID is helpful to solve problems: it can be incorporated into messages to make them trackable (i.e., to or from which process they are sent) to facilitate the message transmission; several problems (leader election, consensus, etc.) can be solved without the information of network property in priori if processes have unique IDs; messages in the register of one process will not be overwritten by others process if this process announces; it is useful to break the symmetry. Hence, eponymous systems have influenced the distributed computing community significantly either in theory or in practice. However, every thing in the world has its own two sides. The unique ID also has disadvantages: it can leak information of the network(size); processes in the system have no privacy; assign unique ID is costly in bulk-production(e.g, sensors). Hence, homonymous system is appeared. If some processes share the same ID and programmed identically is called homonymous system. Furthermore, if all processes shared the same ID or have no ID is named as anonymous system. In homonymous or anonymous distributed systems, the symmetry problem (i.e., how to distinguish messages sent from which process) is the main obstacle in the design of algorithms. This thesis is aimed to propose different symmetry break methods (e.g., random function, counting technique, etc.) to solve agreement problem. Agreement is a fundamental problem in distributed computing including a family of abstractions. In this thesis, we mainly focus on the design of consensus, set agreement, broadcast algorithms in anonymous and homonymous distributed systems. Firstly, the fault-tolerant broadcast abstraction is studied in anonymous systems with reliable or fair lossy communication channels separately. Two classes of anonymous failure detectors AΘ and AP∗ are proposed, and both of them together with a already proposed failure detector ψ are implemented and used to enrich the system model to implement broadcast abstraction. Then, in the study of the consensus abstraction, it is proved the AΩ′ failure detector class is strictly weaker than AΩ and AΩ′ is implementable. The first implementation of consensus in anonymous asynchronous distributed systems augmented with AΩ′ and where a majority of processes does not crash. Finally, a general consensus problem– k-set agreement is researched and the weakest failure detector L used to solve it, in asynchronous message passing systems where processes may crash and recover, with homonyms (i.e., processes may have equal identities), and without a complete initial knowledge of the membership.

Agreement Abstractions in Anonymous and Homonymous Distributed Systems Prone to Failures

The distributed computing models typically assume every process in the system has a distinct identifier (ID) or each process is programmed differently, which is named as eponymous system. In such kind of distributed systems, the unique ID is helpful to solve problems: it can be incorporated into messages to make them trackable (i.e., to or from which process they are sent) to facilitate the message transmission; several problems (leader election, consensus, etc.) can be solved without the information of network property in priori if processes have unique IDs; messages in the register of one process will not be overwritten by others process if this process announces; it is useful to break the symmetry. Hence, eponymous systems have influenced the distributed computing community significantly either in theory or in practice. However, every thing in the world has its own two sides. The unique ID also has disadvantages: it can leak information of the network(size); processes in the system have no privacy; assign unique ID is costly in bulk-production(e.g, sensors). Hence, homonymous system is appeared. If some processes share the same ID and programmed identically is called homonymous system. Furthermore, if all processes shared the same ID or have no ID is named as anonymous system. In homonymous or anonymous distributed systems, the symmetry problem (i.e., how to distinguish messages sent from which process) is the main obstacle in the design of algorithms. This thesis is aimed to propose different symmetry break methods (e.g., random function, counting technique, etc.) to solve agreement problem. Agreement is a fundamental problem in distributed computing including a family of abstractions. In this thesis, we mainly focus on the design of consensus, set agreement, broadcast algorithms in anonymous and homonymous distributed systems. Firstly, the fault-tolerant broadcast abstraction is studied in anonymous systems with reliable or fair lossy communication channels separately. Two classes of anonymous failure detectors AΘ and AP∗ are proposed, and both of them together with a already proposed failure detector ψ are implemented and used to enrich the system model to implement broadcast abstraction. Then, in the study of the consensus abstraction, it is proved the AΩ′ failure detector class is strictly weaker than AΩ and AΩ′ is implementable. The first implementation of consensus in anonymous asynchronous distributed systems augmented with AΩ′ and where a majority of processes does not crash. Finally, a general consensus problem– k-set agreement is researched and the weakest failure detector L used to solve it, in asynchronous message passing systems where processes may crash and recover, with homonyms (i.e., processes may have equal identities), and without a complete initial knowledge of the membership.

A Novel Approach to Context-Aware Content-Based Middleware

La computación ubicua está extendiendo su aplicación desde entornos específicos hacia el uso cotidiano; el Internet de las cosas (IoT, en inglés) es el ejemplo más brillante de su aplicación y de la complejidad intrínseca que tiene, en comparación con el clásico desarrollo de aplicaciones. La principal característica que diferencia la computación ubicua de los otros tipos está en como se emplea la información de contexto. Las aplicaciones clásicas no usan en absoluto la información de contexto o usan sólo una pequeña parte de ella, integrándola de una forma ad hoc con una implementación específica para la aplicación. La motivación de este tratamiento particular se tiene que buscar en la dificultad de compartir el contexto con otras aplicaciones. En realidad lo que es información de contexto depende del tipo de aplicación: por poner un ejemplo, para un editor de imágenes, la imagen es la información y sus metadatos, tales como la hora de grabación o los ajustes de la cámara, son el contexto, mientras que para el sistema de ficheros la imagen junto con los ajustes de cámara son la información, y el contexto es representado por los metadatos externos al fichero como la fecha de modificación o la de último acceso. Esto significa que es difícil compartir la información de contexto, y la presencia de un middleware de comunicación que soporte el contexto de forma explícita simplifica el desarrollo de aplicaciones para computación ubicua. Al mismo tiempo el uso del contexto no tiene que ser obligatorio, porque si no se perdería la compatibilidad con las aplicaciones que no lo usan, convirtiendo así dicho middleware en un middleware de contexto. SilboPS, que es nuestra implementación de un sistema publicador/subscriptor basado en contenido e inspirado en SIENA [11, 9], resuelve dicho problema extendiendo el paradigma con dos elementos: el Contexto y la Función de Contexto. El contexto representa la información contextual propiamente dicha del mensaje por enviar o aquella requerida por el subscriptor para recibir notificaciones, mientras la función de contexto se evalúa usando el contexto del publicador y del subscriptor. Esto permite desacoplar la lógica de gestión del contexto de aquella de la función de contexto, incrementando de esta forma la flexibilidad de la comunicación entre varias aplicaciones. De hecho, al utilizar por defecto un contexto vacío, las aplicaciones clásicas y las que manejan el contexto pueden usar el mismo SilboPS, resolviendo de esta forma la incompatibilidad entre las dos categorías. En cualquier caso la posible incompatibilidad semántica sigue existiendo ya que depende de la interpretación que cada aplicación hace de los datos y no puede ser solucionada por una tercera parte agnóstica. El entorno IoT conlleva retos no sólo de contexto, sino también de escalabilidad. La cantidad de sensores, el volumen de datos que producen y la cantidad de aplicaciones que podrían estar interesadas en manipular esos datos está en continuo aumento. Hoy en día la respuesta a esa necesidad es la computación en la nube, pero requiere que las aplicaciones sean no sólo capaces de escalar, sino de hacerlo de forma elástica [22]. Desgraciadamente no hay ninguna primitiva de sistema distribuido de slicing que soporte un particionamiento del estado interno [33] junto con un cambio en caliente, además de que los sistemas cloud actuales como OpenStack u OpenNebula no ofrecen directamente una monitorización elástica. Esto implica que hay un problema bilateral: cómo puede una aplicación escalar de forma elástica y cómo monitorizar esa aplicación para saber cuándo escalarla horizontalmente. E-SilboPS es la versión elástica de SilboPS y se adapta perfectamente como solución para el problema de monitorización, gracias al paradigma publicador/subscriptor basado en contenido y, a diferencia de otras soluciones [5], permite escalar eficientemente, para cumplir con la carga de trabajo sin sobre-provisionar o sub-provisionar recursos. Además está basado en un algoritmo recientemente diseñado que muestra como añadir elasticidad a una aplicación con distintas restricciones sobre el estado: sin estado, estado aislado con coordinación externa y estado compartido con coordinación general. Su evaluación enseña como se pueden conseguir notables speedups, siendo el nivel de red el principal factor limitante: de hecho la eficiencia calculada (ver Figura 5.8) demuestra cómo se comporta cada configuración en comparación con las adyacentes. Esto permite conocer la tendencia actual de todo el sistema, para saber si la siguiente configuración compensará el coste que tiene con la ganancia que lleva en el throughput de notificaciones. Se tiene que prestar especial atención en la evaluación de los despliegues con igual coste, para ver cuál es la mejor solución en relación a una carga de trabajo dada. Como último análisis se ha estimado el overhead introducido por las distintas configuraciones a fin de identificar el principal factor limitante del throughput. Esto ayuda a determinar la parte secuencial y el overhead de base [26] en un despliegue óptimo en comparación con uno subóptimo. Efectivamente, según el tipo de carga de trabajo, la estimación puede ser tan baja como el 10 % para un óptimo local o tan alta como el 60 %: esto ocurre cuando se despliega una configuración sobredimensionada para la carga de trabajo. Esta estimación de la métrica de Karp-Flatt es importante para el sistema de gestión porque le permite conocer en que dirección (ampliar o reducir) es necesario cambiar el despliegue para mejorar sus prestaciones, en lugar que usar simplemente una política de ampliación. ABSTRACT The application of pervasive computing is extending from field-specific to everyday use. The Internet of Things (IoT) is the shiniest example of its application and of its intrinsic complexity compared with classical application development. The main characteristic that differentiates pervasive from other forms of computing lies in the use of contextual information. Some classical applications do not use any contextual information whatsoever. Others, on the other hand, use only part of the contextual information, which is integrated in an ad hoc fashion using an application-specific implementation. This information is handled in a one-off manner because of the difficulty of sharing context across applications. As a matter of fact, the application type determines what the contextual information is. For instance, for an imaging editor, the image is the information and its meta-data, like the time of the shot or camera settings, are the context, whereas, for a file-system application, the image, including its camera settings, is the information and the meta-data external to the file, like the modification date or the last accessed timestamps, constitute the context. This means that contextual information is hard to share. A communication middleware that supports context decidedly eases application development in pervasive computing. However, the use of context should not be mandatory; otherwise, the communication middleware would be reduced to a context middleware and no longer be compatible with non-context-aware applications. SilboPS, our implementation of content-based publish/subscribe inspired by SIENA [11, 9], solves this problem by adding two new elements to the paradigm: the context and the context function. Context represents the actual contextual information specific to the message to be sent or that needs to be notified to the subscriber, whereas the context function is evaluated using the publisher’s context and the subscriber’s context to decide whether the current message and context are useful for the subscriber. In this manner, context logic management is decoupled from context management, increasing the flexibility of communication and usage across different applications. Since the default context is empty, context-aware and classical applications can use the same SilboPS, resolving the syntactic mismatch that there is between the two categories. In any case, the possible semantic mismatch is still present because it depends on how each application interprets the data, and it cannot be resolved by an agnostic third party. The IoT environment introduces not only context but scaling challenges too. The number of sensors, the volume of the data that they produce and the number of applications that could be interested in harvesting such data are growing all the time. Today’s response to the above need is cloud computing. However, cloud computing applications need to be able to scale elastically [22]. Unfortunately there is no slicing, as distributed system primitives that support internal state partitioning [33] and hot swapping and current cloud systems like OpenStack or OpenNebula do not provide elastic monitoring out of the box. This means there is a two-sided problem: 1) how to scale an application elastically and 2) how to monitor the application and know when it should scale in or out. E-SilboPS is the elastic version of SilboPS. I t is the solution for the monitoring problem thanks to its content-based publish/subscribe nature and, unlike other solutions [5], it scales efficiently so as to meet workload demand without overprovisioning or underprovisioning. Additionally, it is based on a newly designed algorithm that shows how to add elasticity in an application with different state constraints: stateless, isolated stateful with external coordination and shared stateful with general coordination. Its evaluation shows that it is able to achieve remarkable speedups where the network layer is the main limiting factor: the calculated efficiency (see Figure 5.8) shows how each configuration performs with respect to adjacent configurations. This provides insight into the actual trending of the whole system in order to predict if the next configuration would offset its cost against the resulting gain in notification throughput. Particular attention has been paid to the evaluation of same-cost deployments in order to find out which one is the best for the given workload demand. Finally, the overhead introduced by the different configurations has been estimated to identify the primary limiting factor for throughput. This helps to determine the intrinsic sequential part and base overhead [26] of an optimal versus a suboptimal deployment. Depending on the type of workload, this can be as low as 10% in a local optimum or as high as 60% when an overprovisioned configuration is deployed for a given workload demand. This Karp-Flatt metric estimation is important for system management because it indicates the direction (scale in or out) in which the deployment has to be changed in order to improve its performance instead of simply using a scale-out policy.

Differences in the oxidative balance of dispersing and non-dispersing individuals : an experimental approach in a passerine bird

Funding This work was supported by grants from the French Ministry of Research (PhD fellowship to CR), the University of Aberdeen (stipend to CR), the CNRS (PICS grant to BD), the L’Oréal Foundation-UNESCO “For Women in Science” program (fellowship to CR), the Région Rhône-Alpes (student mobility grant CMIRA Explora’doc to CR), the Rectors’ Conference of the Swiss Universities (mobility grant to CR), the Fédération de Recherche 41 BioEnvironnement et Santé (training grant to CR), and the Journal of Experimental Biology (travel grant to CR).

On the controllability of distributed systems

To “control” a system is to make it behave (hopefully) according to our “wishes,” in a way compatible with safety and ethics, at the least possible cost. The systems considered here are distributed—i.e., governed (modeled) by partial differential equations (PDEs) of evolution. Our “wish” is to drive the system in a given time, by an adequate choice of the controls, from a given initial state to a final given state, which is the target. If this can be achieved (respectively, if we can reach any “neighborhood” of the target) the system, with the controls at our disposal, is exactly (respectively, approximately) controllable. A very general (and fuzzy) idea is that the more a system is “unstable” (chaotic, turbulent) the “simplest,” or the “cheapest,” it is to achieve exact or approximate controllability. When the PDEs are the Navier–Stokes equations, it leads to conjectures, which are presented and explained. Recent results, reported in this expository paper, essentially prove the conjectures in two space dimensions. In three space dimensions, a large number of new questions arise, some new results support (without proving) the conjectures, such as generic controllability and cases of decrease of cost of control when the instability increases. Short comments are made on models arising in climatology, thermoelasticity, non-Newtonian fluids, and molecular chemistry. The Introduction of the paper and the first part of all sections are not technical. Many open questions are mentioned in the text.