We are witnessing a proliferation of video in the Internet; YouTube is the most bandwidth intensive service of today’s Internet. It accounts for 20 - 35% of the Internet traffic with 35 hours of videos uploaded every minute and more than 700 billion playbacks in 2010. Netflix, a web service that streams premium contents such as TV series, shows, and movies, consumes 30% of the network bandwidth in North America at peak time. Recently, leveraging the content distribution networks (CDNs), a new paradigm for video streaming on the Internet has emerged, namely, Adaptive HTTP Streaming (AHS). AHS has become the industry standard for video streaming over the Internet adopted by broadcast networks as well as VoD services such as YouTube, Netflix, Hulu, etc. In the 90’s and early 2000’s, Internet-based video streaming for high-bitrate video was challenging due to hardware limitations. In that era, to cover the hardware limitations, every software component of a video server needed to be carefully optimized to support the real-time guarantees for jitter-free video delivery. However, most of the software solutions have become less important with the remarkable hardware improvements over the past two decades. There is 100× speedup in CPU speeds; RAM capacity has increased by 1,000×; hard disk drive (HDD) capacity has grown by 10,000×. Today, CPU is no longer a bottleneck for video streaming. On the other hand, storage bandwidth and network bandwidth are still serious bottlenecks for large scale on-demand video streaming. In this dissertation, we aim at a scalable video content distribution system that addresses both storage bottleneck and network bottleneck. The first part of the dissertation pertains to the storage system on the server side: A multi-tiered storage system that exploits a flash memory solid-state drive (SSD) can meet the bandwidth needs in a much more cost- effective way than a traditional two-tier storage system. We first identify the challenges in architecting such a system given the performance quirks of flash-based SSDs, and the lim- itations of state-of-the-art multi-tiered storage systems for video streaming. Armed with the knowledge of these challenges, we show how to construct such a storage system and implement a real web server with multi-tiered storage, evaluate the system with AHS work- loads, and demonstrate significant performance gains while reducing the TCO. The second part of the dissertation pertains to the network system on the client side: Integrating peer- to-peer (P2P) technology with the client-server paradigm results in a much more scalable video content distribution system. AHS is a paradigm for client-driven video streaming; its philosophy matches well with that of P2P video streaming. An adaptation mechanism is the most important component of AHS that determines overall video streaming quality and user experience. We show a throughput-smoothing-based adaptation mechanism that is designed for a client-server architecture does not work well for a P2P architecture. We pro- vide a buffer-based adaptation mechanism, evaluate our solution with OMNeT++/INET simulator, and demonstrate significant performance gains.

Dy-incorporated HfO₂ gate oxide with TaN gate electrode nMOS device has been developed for high performance CMOS applications in 22nm node technology. DyO /HfO bi-layer structure shows thin EOT with reduced leakage current and less charge trapping compared to HfO₂. Excellent electrical performance of the DyO-capped HfO₂ oxide n-MOSFET such as lower V[subscript TH], higher drive current, and improved channel electron mobility are reported. DyO/HfO samples also show better immunity for V[subscript TH] instability and less severe charge trapping characteristics. Its charge trapping characteristics, conduction mechanisms and dielectric reliability have been investigated in this work. As an application to memory device, HfON charge trapping layered NAND flash memory is developed and characterized. First, temperature-dependent Dy diffusion and the diffusion-driven Dy dipole formation process are discussed to clarify the origin of V[subscript TH] shift, and eventually modulate the effective work function in Dy-Hf-O/SiO₂ system. The Dy-induced dipoles are closely related to the Dy-silicate formation at the high-k/SiO₂ interfaces since the V[subscript FB] shift in Dy₂O₃ is caused by the dipole and coincides with the Dy-silicate formation. Dipole formation is a thermally activated process, and more dipoles are formed at a higher temperature with a given Dy content. The Dy-silicate related bonding structure at the interface is associated with the strength of the Dy dipole moment, and becomes dominant in controlling the V[subscript FB]/V[scubscript TH] shift during high temperature annealing in the Dy- Hf-O/SiO₂ gate oxide system. Dy-induced dipole reduces the degradation of the electron mobility. Second, to understand the reduced leakage current of the DyO/HfO sample, the effective barrier height of Dy₂O₃ was calculated from FN tunneling models, and the band diagram was estimated. The higher effective barrier height of Dy₂O₃, which is around 2.32 eV calculated from the F-N plot, accounts for the reduced leakage current in Dy incorporated HfO₂ nMOS devices. The lower barrier height of HfO₂ result in increased electron tunneling currents enhanced by the buildup of hole charges trapped in the oxide, which causes a severe increase of stress-induced leakage current (SILC), leading to oxide breakdown. However, the increased barrier height in Dy incorporated HfO₂ inhibits a further increase of the electron tunneling from the TaN gate, and trapped holes lessen the hole tunneling currents, resulting in a negligible SILC. The lower trap generation rate by the reduced hole trap density and the reduced hole tunneling of the Dy-doped HfO₂ dielectric demonstrates the high dielectric breakdown strength by weakening the charge trapping and defect generation during the stress. Based on these fundamental studies of the dielectric breakdown, modeling of time-dependent dielectric breakdown (TDDB) was done. The intrinsic TDDB of the Dy-doped HfO₂ gate oxide having 1 nm EOT is characterized by the progressive breakdown (PBD) model. At high temperature, the PBD becomes severe, since thermal energy causes carrier hopping between the localized weak spots. The voltage acceleration factor derived from the power law shows a realistic prediction in comparison with those from the 1/E model. The increase of the voltage acceleration factor at lower stress voltage is due to the lower trap generation rate in Dy- incorporated HfO₂. This voltage acceleration factor can be easily extended to include temperature dependency, and the effective activation energy derived from the power law is voltage dependent. Lastly, I studied the device characteristics of thin HfON charge-trap layer nonvolatile memory in a TaN/Al₂O₃/HfON/SiO₂/p-Si (TANOS) structure. A large memory window and fast erase speed, as well as good retention time, were achieved by using the NH₃ nitridation technique to incorporate nitrogen into the thin HfO₂ layer, which causes a high electron-trap density in the HfON layer. The higher dielectric constant of the HfON charge-trap layer induces a higher electric field in the tunneling oxide at the same voltage compared to non-nitrided films and, thus, creates a high Fowler-Nordheim (FN) tunneling current to increase the erase and programming speed. The trap-level energy in the HfON layer was calculated by using an amphoteric model. / text

Chemical treatments using solvents and surfactants can be used to increase the productivity of gas-condensate wells with condensate banks. CMG’s compositional simulator GEM was used to simulate such treatments to gain a better understanding of design questions such as how much treatment solution to inject and to predict the benefits of such treatments. GEM was used to simulate treatments in vertical wells with and without hydraulic fractures and also horizontal wells. However, like other commercial compositional simulators, the flash calculations used to predict the phase behavior is limited to two phases whereas a three-phase flash is needed to accurately model the complex phase behavior that occurs during and after the injection of treatment solutions. UTCOMP is a compositional simulator with three-phase flash routine and attempts were made to use it to simulate such well treatments. However, this is a very difficult problem to simulate and all previous attempts failed because of numerical problems caused by inconsistent phase labeling (so called phase flipping) and the discontinuities this causes in the relative permeability values. In this research, a new relative permeability model based on molar Gibbs free energy was developed, implemented in a compositional simulator and applied to several difficult three-phase flash problems. A new way of modeling the residual saturations was needed to ensure a continuous variation of the residual saturations from the three-phase region to the two-phase region or back and was included in the new model. The new relative permeability model was implemented in the compositional reservoir simulator UTCOMP. This new relative permeability model makes it is unnecessary to identify and track the phases. This method automatically avoids the previous phase flipping problems and thus is physically accurate as well as computationally faster due to the improved numerical performance. The new code was tested by running several difficult simulation problems including a CO2 flood with three-hydrocarbon phases and a water phase. A new framework for doing flash calculations was also developed and implemented in UTCOMP to account for the multiple roots of the cubic equation-of-state to ensure a global minimum in the Gibbs free energy by doing an exhaustive search for the minimum value for one, two and three phases. The purpose was to determine if the standard method using a Gibbs stability test followed by a flash calculation was in fact resulting in the true minimum in the Gibbs free energy. Test problems were run and the results of the standard algorithm and the exhaustive search algorithm compared. The updated UTCOMP simulator was used to understand the flow back of solvents injected in gas condensate wells as part of chemical treatments. The flow back of the solvents, a short-term process, affects how well the treatment works and has been an important design and performance question for years that could not be simulated correctly until now due to the limitations of both commercial simulators and UTCOMP. Different solvents and chase gases were simulated to gain insight into how to improve the design of the chemical treatments under different conditions. / text

Distributed models of the land surface are essential for global climate models because of the importance of land-atmosphere exchanges of water, energy, momentum. They are also used for high resolution hydrologic simulation because of the need to capture non-linear responses to spatially variable inputs. Continued improvements to these models, and the data which they use, is especially important given ongoing changes in climate and land cover. In hydrologic models, important aspects are sometimes neglected due to the need to simplify the models for operational simulation. For example, operational flash flood models do not consider the role of snow and are often lumped (i.e. do not discretize a watershed into multiple units, and so do not fully consider the effect of intense, localized rainstorms). To address this deficiency, an overland flow model is coupled with a subsurface flow model to create a distributed flash flood forecasting system that can simulate flash floods that involve rain on snow. The model is intended for operational use, and there are extensive algorithms to incorporate high-resolution hydrometeorologic data, to assist in the calibration of the models, and to run the model in real time. A second study, which is designed to improve snow simulation in forested environments, demonstrates the importance of explicitly representing a near canopy environment in snow models, instead of only representing open and canopy covered areas (i.e. with % canopy fraction), as is often done. Our modeling, which uses canopy structure information from Aerial Laser Survey Mapping at 1 meter resolution, suggests that areas near trees have more net snow water input than surrounding areas because of the lack of snow interception, shading by the trees, and the effects of wind. In addition, the greatest discrepancy between our model simulations that explicitly represent forest structure and those that do not occur in areas with more canopy edges. In addition, two value-added Land Cover products (land cover type and maximum green vegetation fraction; MGVF) are developed and evaluated. The new products are good successors to current generation land cover products that are used in global models (many of which rely on 20 year old AVHRR land cover data from a single year) because they are based on 10 years of recent MODIS data. There is substantial spurious interannual variability in the MODIS land cover type data, and the MGVF product can vary substantially from year to year depending on climate conditions, suggesting the importance of using climatologies for land cover data. The new land cover type climatology also agrees better with validation sites, and the MGVF climatology is more consistent with other measures of vegetation (e.g. Leaf Area Index) than the older land cover data.

This dissertation reports the results of a field based study examining sediment transport and bed mobility in a low-ordered, ephemeral watershed. Runoff and sediment transport concentrations were sampled at the watershed outlet to determine flow discharge and sediment flux during approximately 21 flow events, from 1998 - 2007. Sediment collected in flow was measured for grain-size distribution to determine if specific grain-size fractions behave differently while in transport. The coarse sediment yield was measured for mass and grain-size distribution at the watershed outlet for two years, 2005 - 2006. Further, the arrangement and composition of the channel bed material was comprehensively mapped using terrestrial-based photogrammetry for the years, 2005 - 2006. Results show that patterns of sediment transport are complex, controlled in part by flow hydraulics but also by other phenomena. Some of the variation in sediment transport is determined by grain-size. Grain-sizes with different sources within the watershed and that transported by different transport modes were observed to follow different patterns of transport. Also, the channel bed, which serves as the source for the coarse fraction of the sediment transport, was observed to change in grain composition during periods of flow. This tendency for the bed material to evolve in time likely affected the amount and composition of the sediment grains that were entrained from it.An additional objective of this dissertation was to determine how unique the observed patterns of sediment transport were to low-order ephemeral channels. Sediment transport and yield were modeled using bed load transport formulae designed to capture the physical mechanics of transport as observed in perennial streams. Results show that contemporary transport models predict transport within the field site with similar accuracy as that in many perennial systems but not well enough to rely on their predictions for many engineering applications.

Railroad bridges and culverts in the United States are often subject to extreme floods, which have been known to washout sections of track and ultimately lead to derailments. The potential for these events is particularly high in the western U.S. due to the lack of data, inadequate radar coverage, and the high spatial and temporal variability of storm events and terrain. In this work, a hydrologic model is developed that is capable of effectively describing the rainfall-runoff relationship of extreme thunderstorms in arid and semi-arid regions. The model was calibrated and validated using data from ten storms at the semi-arid Walnut Gulch Experimental Watershed. A methodology is also proposed for reducing the amount of raingages required to provide acceptable inputs to the hydrologic model, and also determining the most appropriate placement location for these gages. Results show that the model is capable of reproducing peak discharges, peak timings, and total volumes to within 22.1%, 12 min, and 32.8%, respectively. Results of the gage reduction procedure show that a decrease in the amount of raingages used to drive the model results in a disproportionally smaller decrease in model accuracy. Results also indicate that choosing gages using the minimization of correlation approach that is described herein will lead to an increase in model accuracy as opposed to selecting gages on a random basis.

Internet in recent years has changed quite a lot, usual static user interfaces poorly satisfy user's requirements or do not at all. That is why there is a need to analyze a new breed of internet presentation-level technique – rich internet applications (RIA). Main RIA technologies and several internet projects were analyzed in this thesis. Macromedia Flash technology was chosen because of its flexibility, multiplatform implementation, rich user experience, good integration among various server technologies. An experimental system named Ferry transport booking system was built using Macromedia Flash technology and implemented with Microsoft server technologies. Experimental research of this project is described; a comparative analysis of rich internet application and simple static interface advantages is provided.

L'archivage de données industrielles est un problème complexe : il s'agit de stocker un grand nombre de données sur plusieurs décennies, tout en supportant la charge des insertions temps réel et des requêtes d'extraction et d'analyse. Pour ce type d'application, des produits " de niche " se sont spécialisés pour ce segment du marché : les progiciels d'historisation. Il s'agit de solutions propriétaires avec des coûts de licence de l'ordre de plusieurs dizaines de milliers d'euros, et dont le fonctionnement interne n'est pas dévoilé. Nous avons donc dans un premier temps mis en évidence les spécificités de ces progiciels d'historisation, tant au niveau des fonctionnalités que des performances. Néanmoins, l'archivage de données industrielles peut s'appliquer à des contexte très différents. L'IGCBox par exemple est un mini PC industriel utilisant MySQL pour l'archivage à court terme des données de production des centrales hydrauliques d'EDF. Ce matériel présente quelques spécificités, la principale étant son système de mémoire non volatile basé uniquement sur la technologie flash, pour sa fiabilité importante en milieu industriel et sa faible consommation d'énergie. Les SGBD possèdent pour des raisons historiques de nombreuses optimisations spécifiques aux disques durs, et le manque d'optimisation adaptée aux mémoires flash peut dégrader significativement les performances. Le choix de ce type de mémoire a donc eu des répercussions notables sur les performances pour l'insertion, avec une dégradation importante par rapport aux disques durs. Nous avons donc proposé Chronos, un SGBD dédié à l'historisation de données sur mémoires flash. Pour cela, nous avons en particulier identifié un algorithme d'écriture " quasi-séquentiel " efficace pour accéder à la mémoire, ainsi que des mécanismes de bufferisation et de mise à jour d'index optimisés pour les charges typiques de l'historisation. Les résultats expérimentaux montrent un gain significatif pour les insertions par rapport à des solutions équivalentes, d'un facteur 20 à 54. Chronos est donc une solution compétitive lorsque les insertions correspondent à une proportion importante de la charge soumise au SGBD. En particulier pour les charges typiques des IGCBox, Chronos se distingue en proposant des performances globales améliorées d'un facteur 4 à 18 par rapport aux autres solutions.

This thesis presents Illusion SDK: a general, extensible framework for augmented reality (AR) applications. Illusion provides loosely coupled or decoupled abstractions of sensors, trackers, and compositors. Implementations are optimized for particular use cases. Illusion’s architecture depends on only an event system and a 3D scene graph, so it is highly portable. Wrapping of third-party trackers is supported. Illusion’s current implementation targets Flash 11.4 and integrates with the Alternativa3D 8 graphics engine. To our knowledge, Illusion’s support for wrapping third-party trackers is unique among toolkits targeting the GPU-accelerated Web. Illusion performs well on MacBook Pro 13" mid-2010, where an intensive camera application can exceed 45 FPS. Generally, Illusion should perform well on hardware that uses shared video memory. Optimizations are needed for hardware that uses dedicated video memory. These optimizations are problematic in Flash 11.4 but should not generally be problematic in ports to other platforms.

Thèse numérisée par la Division de la gestion de documents et des archives de l'Université de Montréal