Dynamic data placement and tool assignment for big-data orchestrated bioinformatics workflows

Abstract

Big-data bioinformatics workflows are usually complex and data-intensive. They need to analyze large volumes of data using one or more analysis tools either from existing datasets or that from the new intermediate datasets generated during the workflow execution. Traditionally, the workflows are executed by moving the data to analysis tools. With the advent of big-data in bioinformatics workflows, moving the analysis tools to data is better solution for handling big-data. This is because the size of analysis tools is usually smaller than the size of data moving in a workflow. However, moving analysis tools to data may not be possible and cost effective at all times. Therefore, in this paper, we propose a dynamic approach based on dataset size, to move and place, datasets and tools during workflow execution. The approach moves tools to datasets and for those datasets whose movement is inevitable, based on their dependencies, groups them together using agglomerative clustering at the most appropriate data service before moving the tools to data. Simulations show that our approach reduces data movement and improves workflow performance compared to the traditional approach of workflow execution.