Sciflow-Bridge: uma ferramenta para criação e distribuição de imagens de contêineres para workflows científicos

Abstract

A ScientificWorkflow defines how an experiment is characterized, determining in which order tasks are performed and which dataset each task must operate on. Each task have different environment dependencies, such as libraries, executables, and input data. Containers feature light virtualization and are capable of encapsulating multiple layers of software. In this context, the present work proposes to create a tool that is capable of executing these workflows and managing dependencies through the efficient creation and distribution of container images. It is hoped that by using this tool, researchers will be able to execute workflows easily and will also be able to take advantage of the processing resources available in the laboratories. Two bioinformatics workflows were executed with the created tool, the experiments proved that the tool is capable of performing scientific workflows, the use of cache has decreased the execution time of the workflows and was observed that different strategies for the positioning of the containers also influences the runtime. The use of the tool gives scientific workflows flexibility, scalability and reproducibility.