Suporte a fluxos de trabalho de aplicações intensivas em dados

Abstract

The increase of the demand of computation and data have forced the scientific applications to use distributed and shared resources. The scientific workflow systems have been introduced in response to the demand of researcher from several domainsof science who need to process and analyse this increasingly larger experimental datasets.The introduction of the workflow systems is based on the observation that scientific applications are constructed by the composition of multiple computation stages as a standard pipeline that need to be executed on very large data collection. In such a way, the scientific workflow systems had allowed the computation stages to be mapped into workflow stages, which can be efficiently executed in distributed systems. In this work we present scientific workflow system that is unique in sence that it have been developed to facilitate the execution of scientific applications in distributed systems using databases to store scientifc data. Our system is optimized for data-intesive workflows, meaning that we are very concerned with data management issues. The experimental results with our system have shown that we can achieve linear speedups for fairly sophisticated application, created from multiple components.