Suporte para a adição de instâncias em tempo de execução no anthill

Detalhes bibliográficos
Ano de defesa: 2009
Autor(a) principal: Daniel Lacet de Faria Fireman
Orientador(a): Não Informado pela instituição
Banca de defesa: Não Informado pela instituição
Tipo de documento: Dissertação
Tipo de acesso: Acesso aberto
Idioma: por
Instituição de defesa: Universidade Federal de Minas Gerais
UFMG
Programa de Pós-Graduação: Não Informado pela instituição
Departamento: Não Informado pela instituição
País: Não Informado pela instituição
Palavras-chave em Português:
Link de acesso: http://hdl.handle.net/1843/SLSS-7WJNE6
Resumo: On one hand, the continuous evolution of technology in various areas of knowledge is leading to an increase of size of available data sets, reaching the order of petabytes. In this scenario, it is essential to use distributed resources to process data in a timely fashion. On the other hand, with the popularization of platforms composed by workstations, distributed computing systems have become inherently dynamic: resources may be added or are subjected to faults. However, the use of these resources as a platform for high performance computing is still restricted, since the development of efficient and scalable parallel systems remains a difficult task, even for experienced programmers. Anthill is a parallel programming environment based on the filter-stream paradigm. This paradigm allows to efficiently process large volumes of data, since it can exploit the parallelism in a simple and intuitive manner. However, Anthill has been built as a static environment and as such, it does not allow to modify the application's component distribution at runtime. This work presents a set of extensions that add support for augmenting the runtime platform of an Anthill application. Our solution exploits the locality of reference and the asynchrony which many Anthill application could make use. The results of our experimental evaluation show that this implementation allows the use of more computational power at runtime, keeping with low cost, the execution consistency and the asynchronous exploitation of different levels of parallelism.