Zugl.: Stuttgart, Univ., Diss., 2008.

Stuttgart, Univ., Diplomarb., 2003.

Zugl.: Potsdam, Universiẗat, Diss., 2008.

Universiẗat, Diss., 2009--Bayreuth.

Dissertação (mestrado)—Universidade de Brasília, Instituto de Ciências Exatas, Departamento de Ciência da Computação, 2007. / Submitted by Luis Felipe Souza (luis_felas@globo.com) on 2009-01-05T14:12:55Z No. of bitstreams: 1 Dissertacao_2007_GlauberScorsatto.pdf: 2122593 bytes, checksum: 53c097affeb6a78e961a465da8c004f1 (MD5) / Approved for entry into archive by Georgia Fernandes(georgia@bce.unb.br) on 2009-03-02T19:52:14Z (GMT) No. of bitstreams: 1 Dissertacao_2007_GlauberScorsatto.pdf: 2122593 bytes, checksum: 53c097affeb6a78e961a465da8c004f1 (MD5) / Made available in DSpace on 2009-03-02T19:52:14Z (GMT). No. of bitstreams: 1 Dissertacao_2007_GlauberScorsatto.pdf: 2122593 bytes, checksum: 53c097affeb6a78e961a465da8c004f1 (MD5) / Um sistema distribuído consiste em um conjunto de elementos computacionais que se comunicam através de uma rede de interconexão e que utilizam um software que permite que trabalhem de maneira cooperativa e coordenada. Visando o compartilhamento de recursos em larga escala, foi proposta a Computação em Grid, um tipo de sistema distribuído cujo objetivo é o compartilhamento de recursos geograficamente distribuídos, autônomos e pertencentes a diferentes domínios administrativos. Algumas aplicações executadas em grid podem necessitar de garantias de disponibilidade de recursos. Para que tais aplicações sejam satisfeitas, faz-se necessário o fornecimento de garantias de Qualidade de Serviço (QoS). QoS consiste em um conjunto de características quantitativas e qualitativas de um sistema necessárias para que seja alcançado o nível de serviço esperado pelas aplicações. Um meio de se fazer a negociação dos níveis de QoS fornecidos às aplicações é a utilização de acordos de serviço, pelos quais a parte fornecedora se compromete a prover o nível de serviço desejado. A presente dissertação propõe o GrAMoS (Grid Agreement Monitoring Service), um serviço capaz de monitorar acordos firmados em grid, assim como de detectar possíveis violações desses acordos e de tomar ações em caso de detecção de violações. A monitoração é feita em instantes randômicos dentro de intervalos fixos e as ações tomadas na situação de quebra de acordo são flexíveis e podem ser fornecidas pelo usuário. A análise do protótipo do GrAMoS, implementado no Globus Toolkit 4 (GT4), mostra que o overhead causado pelo monitoramento é baixo (abaixo de 2,8%), causando um impacto mínimo na aplicação de grid. _______________________________________________________________________________________________________________ ABSTRACT / A distributed system consists in a set of computational elements, which communicate through an interconnected network, making use of software that allows the elements to work in a cooperative and coordinate way. Aiming a large scale resource sharing, Grid Computing has emerged, a kind of distributed system which has as objective the sharing of autonomous, geographically distributed resources, belonging to different administrative domains. Some applications executed in a grid system may demand resource availability in some specific levels. For these applications to be satisfied, it is necessary to supply guarantees of Quality of Service (QoS). QoS consists of several quantitative and qualitative system characteristics which are necessary for the level of service demanded by applications to be achieved. A means of negotiating the QoS levels supplied for the applications is to use service agreements, by which the resource supplier compromises to supply the demanded level of service. The present dissertation proposes GrAMoS (Grid Agreement Monitoring Service), a service capable of monitoring agreements firmed in a grid, which can detect agreement violations and take actions. In our service, monitoring is done at random moments inside fixed intervals and the actions taken in the case of agreement violations are flexible and can be provided by the user. The evaluation of the GrAMoS prototype, which was implemented on top of the Globus Toolkit 4 (GT4), shows that the monitoring overhead is low (below 2.8%), causing a very small impact on the grid applications.

The avalanche of data from scientific instruments and the ensuing interest from geographically distributed users to analyze and interpret it accentuates the need for efficient data dissemination. An optimal data distribution scheme will find the delicate balance between conflicting requirements of minimizing transfer times, minimizing the impact on the network, and uniformly distributing load among participants. We identify several data distribution techniques, some successfully employed by today's peer-to-peer networks: staging, data partitioning, orthogonal bandwidth exploitation, and combinations of the above. We use simulations to explore the performance of these techniques in contexts similar to those used by today's data-centric scientific collaborations and derive several recommendations for efficient data dissemination. Our experimental results show that the peer-to-peer solutions that offer load balancing and good fault tolerance properties and have embedded participation incentives lead to unjustified costs in today's scientific data collaborations deployed on over-provisioned network cores. However, as user communities grow and these deployments scale, peer-to-peer data delivery mechanisms will likely outperform other techniques. / Applied Science, Faculty of / Electrical and Computer Engineering, Department of / Graduate

The Grid has revolutionized the way computations are done on the Internet. Access to remote computational resources and ad hoc creation of virtual organizations across administrative domains opens new opportunities on the Grid. The newly developed web services based Open Grid Services Architecture makes the Grid more accessible by allowing the Grid to be constructed from distinct platform independent components. Together they provide an environment for application sharing (or trading), collaborations and access to remote data repositories. The application marketplace is a natural extension to this application sharing environment. The marketplace addresses the fact that the existing infrastructure is still incomplete without provisions for publishing and discovering applications and resources, including the application descriptors that must be moved between the market participants. This work demonstrates a web service instance-based infrastructure, the application market that allows the sellers, the application and the CPU providers to publish their applications for the users to find and use. The application market uses a portal architecture built on top of Globus toolkit 3.0 that interacts with the providers and the users. The market services provide distinct interfaces that allow providers to advertise applications and users to select, configure, and run these applications. The applications themselves are modeled as stateful objects represented using XML which can be exchanged between the providers and users when required. The marketplace, through its interfaces, effectively hides the compute resource and application complexity thus allowing end users to explore and use applications unfamiliar to them with ease.

No description available.

The use of the distributed file system, AFS, as a solution to the “input/output sandbox” problem in grid computing is studied. A computational grid middleware, primarily to accommodate the environment of the BaBar Computing Model, has been designed, written and is presented. A summary of the existing grid middleware and resources is discussed. A number of benchmarks (one written for this thesis) are used to test the performance of the AFS over the wide area network and grid environment. The performance of the AFS is also tested using a straightforward BaBar Analysis code on real data. Secure web-based and command-line interfaces created to monitor job submission and grid fabric are presented.

Epigenetics is the study of factors that can change DNA and passed to next generation without change to DNA sequence. DNA methylation is one of the categories of epigenetic change. DNA methylation is the attachment of methyl group (CH3) to DNA. Most of the time it occurs in the sequences that G is followed by C known as CpG sites and by addition of methyl to the cytosine residue. As science and technology progress new data are available about individual’s DNA methylation profile in different conditions. Also new features discovered that can have role in DNA methylation. The availability of new data on DNA methylation and other features of DNA provide challenge to bioinformatics and the opportunity to discover new knowledge from existing data. In this research multiple data series were used to identify classes of methylation DNA to CpG sites. These classes are a) Never methylated CpG sites,b) Always methylated CpG sites, c) Methylated CpG sites in cancer/disease samples and non-methylated in normal samples d) Methylated CpG sites in normal samples and non-methylated in cancer/disease samples. After identification of these sites and their classes, an analysis was carried out to find the features which can better classify these sites a matrix of features was generated using four applications in EMBOSS software suite. Features matrix was also generated using the gUse/WS-PGRADE portal workflow system. In order to do this each of the four applications were grid enabled and ported to BOINC platform. The gUse portal was connected to the BOINC project via 3G-bridge. Each node in the workflow created portion of matrix and then these portions were combined together to create final matrix. This final feature matrix used in a hill climbing workflow. Hill climbing node was a JAVA program ported to BOINC platform. A Hill climbing search workflow was used to search for a subset of features that are better at classifying the CpG sites using 5 different measurements and three different classification methods: support vector machine, naïve bayes and J48 decision tree. Using this approach the hill climbing search found the models which contain less than half the number of features and better classification results. It is also been demonstrated that using gUse/WS-PGRADE workflow system can provide a modular way of feature generation so adding new feature generator application can be done without changing other parts. It is also shown that using grid enabled applications can speedup both feature generation and feature subset selection. The approach used in this research for distributed workflow based feature generation is not restricted to this study and can be applied in other studies that involve feature generation. The approach also needs multiple binaries to generate portions of features. The grid enabled hill climbing search application can also be used in different context as it only requires to follow the same format of feature matrix.