2010 Winter Project Week Slicer XNAT

Upload pipeline of tagged data to remote host

Download pipeline and automatic tagging of data types

Key Investigators

• Wendy Plesniak
• John Paulett
• Tim Olsen
• Dan Marcus

Objective

Exchanging data between XNAT and Slicer is currently very slow. We will identify and begin resolving bottlenecks in the current FetchMI module and Remote I/O pipeline.

Approach, Plan

• Profile data downloads in the FetchMI module and in Slicer's RemoteIO pipeline.
• Walk through the FetchMI code to see how the XNAT API is being utilized. Optimize.
• Determine ways to optimize Slicer's RemoteIO pipeline.

Progress

Progress will be rapid, possibly ferocious.

• Wendy found a bug in the upload pipeline! not yet fixed, but high priority.
• Instrumented the download pipeline with profiling methods to tease out where time is being spent.
• Reviewed with Tim and Dave -- transfer times seem appropriate.
• Tim and Dave reviewed Wendy's informatics code, appears that the code is appropriate.
• Profiling reveals that bulk of time is being spent in the Remote I/O pipeline...
• Discussed with Ron, Dan, Jim some ways to bypass cache in Slicer to speed up download.
• Profiling will continue and if necessary, we will extend the Remote I/O pipeline for improved performance.

Reference material