Closed sina-mansour closed 2 years ago
Hi Rob,
The UKB connectivity mapping pipeline script accepts input for a temporary file storage directory. The automation script uses this input to write all temporary files on the scratch space. This ensures that the code uses the fast RAM file system when possible.
Cool; as long as it's a performance issue that you're aware of and have accounted for.
Following on the suggestion by @Lestropie (this commit):
RS: If you can guarantee adequate space, creating the scratch directory in /tmp/ would result in using a RAM file system, which will be way faster than a shared network filesystem. For re-executing data for which some files are pre-existing, you would need to explicitly manage the scratch directory location. My Python API provides command-line options for doing this.