Uma Estratégia para Versionamento dos Dados de Workflows Científicos Executados em Nuvem

Scientific experiments usually run hundreds or thousands of times, generating a huge amount of data that requires to be managed. Analizing and comparing the results of such experiments is na extremely complex task. This becomes even more complex for workflows running in the cloud because the data is scattered across multiple virtual machines. In order to alleviate this proble, previous work proposed the use of a version control system to manage the data consumed and generated by scientific experiments. However, they add considerable overhead to the experiment, increasing the processing time and the use of disk space. In this article, we propose an alternative strategy to reduce time and space. Our initial experiments show that the time overhead of our approach is still high, but disk overhead was 5 times smaller than the approaches in the literature.