The Loni Pipeline executes scientific workflows on a large cluster of Linux workstations. Each workflow is composed of multiple executables running on the cluster and accessing data on a network file system. The network file system is a bottleneck for I/O intensive jobs. We hope that the open source Hadoop project can provide a more scalable distributed file system and parallel processing infrastructure for the LONI pipeline cluster.
Your mission: Integrate Hadoop into the LONI Pipeline and demonstrate better performance, scalability and reliability. In addition, explore new parallel processing features such as parallelizing loops and control structures.
Currently, scientific workflows are visualized in the LONI Pipeline as two dimensional schematics that can be navigated by moving up, down, left, right and zooming in and out. However, these workflows and the neuroimaging data that they process can be large and complex. Therefore, visualizing and presenting workflows in three dimensions or using other visualization and HCI paradigms can vastly improve user experience and perception of the Pipeline workflows.
Your mission: Develop a Pipeline Navigator for browsing scientific workflows in 3D.
We have two Python libraries that provide Intelligence and Scripting capabilities to the Pipeline software. The amount of code is approximately 2000 lines including comments for the Scripting library and … for the Intelligence library.
Your mission: Develop a Java version of the Scripting and Intelligence libraries. This is a much shorter and easier project compared to Project 1 and Project 2, and is a great exercise in Object Oriented design and development.