KTHFS is a highly available and scalable file system built from the version 0.24 of the Hadoop Distributed File system. It provides a platform to overcome the limitations of existing distributed file systems. These limitations include scalability of metadata server in terms of memory usage, throughput and its availability. This document describes KTHFS architecture and how it addresses these problems by providing a well coordinated distributed stateless metadata server (or in our case, Namenode) architecture. This is backed with the help of a persistence layer such as NDB cluster. Its primary focus is towards High Availability of the Namenode. It achieves scalability and recovery by persisting the metadata to an NDB cluster. All namenodes are connected to this NDB cluster and hence are aware of the state of the file system at any point in time. In terms of High Availability, KTHFS provides Multi-Namenode architecture. Since these namenodes are stateless and have a consistent view of the metadata, clients can issue requests on any of the namenodes. Hence, if one of these servers goes down, clients can retry its operation on the next available namenode. We next discuss the evaluation of KTHFS in terms of its metadata capacity for medium and large size clusters, throughput and high availability of the Namenode and an analysis of the underlying NDBcluster. Finally, we conclude this document with a few words on the ongoing and future work in KTHFS.

The cores of galaxy clusters contain the most massive galaxies in the Universe, the brightest cluster galaxies. These galaxies are unique compared to their counterpart galaxies outside of clusters as they have much brighter cores, and vast spatially- extended stellar envelopes. The theoretical picture of how they reached their huge masses relied on the idea of gradual stellar mass growth during the second half of the history of the Universe. However, recent observational evidence of highly-overdense protoclusters, the progenitors of these galaxies, demonstrates that some brightest cluster galaxies may have assembled within the first few billion years after the Big Bang – seemingly contradicting our theoretical predictions. I include my theoretical work that shows the short timescales over which these observed protoclusters trans- form into the brightest cluster galaxies and discuss the likelihood of finding these rare protoclusters in the early Universe. To push our understanding of the rapid evolution of these galaxies even further for- ward demands the use of numerical simulations due to the highly coupled, non-linear astrophysical processes that occur during the process. In this dissertation, I include improvements to our numerical models of hydrodynamical turbulence and supermas- sive black holes that I incorporated into a state-of-the-art hydrodynamical+gravity simulation code, in effort to provide the groundwork to improving our understanding of the build-up of the brightest cluster galaxies in the early Universe, and galaxy evolution in general. / Graduate

Cette thèse a pour objectif de développer des techniques permettant d'améliorer l'efficacité énergétique des réseaux de capteurs destinés à la surveillance d'une chaîne de froid. La première approche développée dans cette thèse concerne la phase de transport des produits alimentaires. Les protocoles proposés VBS, WaS et eVBS s'adressent à la problématique posée par l'absence de station de base permanente au sein du petit réseau déployé dans un camion où les nœuds sont en visibilité directe. Ces méthodes ont le triple avantage de présenter un faible overhead, d'accroître l'ergonomie de l'application et l'intérêt économique du réseau. Dans l'état de l'art, il n'existe pas de déploiement de très grands réseaux de capteurs sans fil destinés à la surveillance d'une chaîne de froid. Le cœur de cette thèse se positionne ainsi autour de la problématique du passage à l'échelle, en proposant plusieurs approches permettant d'améliorer l'efficacité énergétique du réseau. Ainsi des optimisations du routage et des techniques de formation de clusters multi-sauts sont proposées grâce à quelques idées originales d'utilisation de l'indicateur de la qualité de lien. Le protocole L2RP de routage par répartition de charge, le mécanisme SNCR de réduction des clusters singletons ainsi que le protocole LQI-DCP d'optimisation du positionnement des chefs de clusters sont des contributions de cette thèse exploitant le LQI afin d'accroître l'efficacité énergétique du réseau. De plus, nous avons montré que l'heuristique MaxMin de formation de clusters multi-sauts n'est pas compatible avec la topologie de déploiement en grille qui est la plus fréquente dans les architectures de réseaux de capteurs. / Wireless Sensor Networks (WSN) are used today in many applications that differ in their own objectives and specific constraints. However the common challenge in designing WSN applications comes from the specific constraints of micro-sensors because of their limited physical resources such as limited battery lifetime, weak computational capability and small memory capacity. This thesis aims to develop techniques to improve the energy efficiency of wireless sensor networks intended to a cold chain monitoring application. In such area, existing solutions consist of walled temperature recorders in warehouses and transport vehicles, which essentially control the room storage temperature which is not necessarily the product one. By integrating wireless micro-sensor devices with pallets and trays, one allows full real-time tracking of the cold chain originating from factories to the shelves of retailers. The first approach developed in this thesis concerns the phase of transporting food. The proposed VBS, WaS and eVBS protocols address the main issue resulting from the absence of a permanent base station in the small network deployed in a truck where nodes are generally in the same transmission range. These methods have the triple advantages of having a low overhead, increasing the ergonomics of the application and enhancing the economic interest of the network. In the state of art, there is no very large scale deployment of WSN for a cold chain monitoring system. Thus, the core of this thesis addresses the scalability issues by offering several ways to improve overall energy efficiency of the network. So, routing techniques improvement and efficient multi-hop clustering protocols are proposed through some original ideas using the Link Quality Indicator (LQI) provided by the MAC sublayer. The LQI is defined in the IEEE 802.15.4 standard in which its context of use is not specified. The link reliability based routing protocol (L2RP) which load balances the traffic between nodes, the single-node cluster reduction mechanism (SNCR) and the LQI-DCP multi-hop clusters formation protocol which improves clusterhead locations are some of contributions of this thesis which exploit the LQI to, significantly, increase the WSN efficiency. We also show that the MaxMin d-cluster formation heuristic does not support the grid deployment topology which is the more often used in WSN architectures.

El fraude con tarjetas de crédito es uno de los problemas más importantes a los que se enfrentan actualmente las entidades financieras. Si bien la tecnología ha permitido aumentar la seguridad en las tarjetas de crédito con el uso de claves PIN, la introducción de chips en las tarjetas, el uso de claves adicionales como tokens y mejoras en la reglamentación de su uso, también es una necesidad para las entidades bancarias, actuar de manera preventiva frente a este crimen. Para actuar de manera preventiva es necesario monitorear en tiempo real las operaciones que se realizan y tener la capacidad de reaccionar oportunamente frente a alguna operación dudosa que se realice. La técnica de Clustering frente a esta problemática es un método muy utilizado puesto que permite la agrupación de datos lo que permitiría clasificarlos por su similitud de acuerdo a alguna métrica, esta medida de similaridad está basada en los atributos que describen a los objetos. Además esta técnica es muy sensible a la herramienta Outlier que se caracteriza por el impacto que causa sobre el estadístico cuando va a analizar los datos. / ---The credit card fraud is one of the most important problems currently facing financial institutions. While technology has enhanced security in credit cards with the use of PINs, the introduction of chips on the cards, the use of additional keys as tokens and improvements in the regulation of their use, is also a need for banks, act preemptively against this crime. To act proactively need real-time monitoring operations are carried out and have the ability to react promptly against any questionable transaction that takes place. Clustering technique tackle this problem is a common method since it allows the grouping of data allowing classifying them by their similarity according to some metric, this measure of similarity is based on the attributes that describe the objects. Moreover, this technique is very sensitive to Outlier tool that is characterized by the impact they cause on the statistic when going to analyze the data. / Tesis

No description available.

Cluster Organizations, as a means of promoting competition and innovation in industrial clusters, have become increasingly popular over the world. Cluster organizations aim to increase growth and competitiveness of clusters within a region, and have become a central part of economic policy-making across the world. Recently, the concept has been used to a larger extent as a tool for economic development and poverty alleviation. This thesis seeks to examine the cluster organizations that are part of the Sida funded program Innovation Systems and Cluster development in Tanzania (ISCP-Tz), by evaluating performance, goals and development of the program based on cluster facilitators perceptions, and assess linkages and actor collaborations between clustered actors. The data in this thesis is collected through a telephone-administered questionnaire, as well as interviews and visits to cluster sites. The results show a positive impact on cluster firms performance as assessed by cluster facilitators, but show that actor collaborations in many cases are inadequate and need to be improved.

No description available.

No description available.

No description available.

No description available.