Page 1 of 1

Scratch space

Posted: Wed Jun 12, 2013 6:21 pm
by kengel
Hello:

My system admin has told me that Abinit 6.8.2 requires a common scratch space on our cluster if I run a job across multiple nodes. I'm wondering if the more recent version of Abinit parallelization allows a job to run across multiple nodes with just the local scratch space of each node separately? I think this would significantly speed up my large calculations, as right now all the scratch files from each node have to be written to a common NAS device.

Thank you,
Kale