On Apr 3, 2013, at 5:42 PM, Jeff Darcy wrote: > The flippant answer would be to move the computation to the data instead > of vice versa, like Hadoop is designed to do. ;) This promise of being able to move the computation to the data was one of the reasons for my recommending glusterfs as the distributed fs for an in-house HPC cluster. However, since the necessary xattr's (which is how you figure out where the file "has landed") are privileged they can't be read by non-root users. Is there a solution to this or an alternative that you can offer? Thank you, Vlad
