On Mon, Feb 22, 2010 at 8:35 AM, Mag Gam
<magawake@xxxxxxxxx> wrote:
Currently we get all of our files from 1 NFS server but it seems like
that is the bottleneck. On all of our 500 simulation boxes we have
100GB of local disk space available for scratch space. I would like to
place pieces of the data into the 100GB disks and access it like that
instead of the NFS server. Is it possible to do this?
Mag, I don't know if Stork can do this, but have you talked to the crew at Cycle Computing? They have a distributed filesystem product that does exactly what you're asking for: it ties together unused disk on your compute nodes into one big, redundant file system.
- Ian
![[Computer Systems Lab]](http://www.cs.wisc.edu/pics/csl_logo.gif){kind=logo}