Replacing NFS.

Rob Latham rlatham at
Mon Apr 9 17:44:57 PDT 2001

On Mon, Apr 09, 2001 at 05:20:58PM -0400, Georgia Southern Beowulf Cluster Project wrote:
> Hello,
> I'm trying to replace NFS on my cluster and I'm running into a dead end.  
> I'm using 15 diskless nodes with a single master node.  

of course, it depends on your situation, but aren't you pretty much
running a hardware environment begging for Scyld?

> Also, does anyone have experience with Coda and is it pretty much a 
> drop-in replacement for NFS?  

please speak up otherwise, coda fans, but it seems to be in a
perpetual state of devleoment.  i usually check the option  when
building kernels at home. not that i've actually used it, though :>

> I've also heard of AFS, but it seems to be a 
> bit limited.  

limited how?  if anything it's overkill, especially since you won't
have a local disk for cache ( non-broken caching is one of the best
reasons to use AFS in a cluster of workstations )

> Also, Global File System (GFS), but it requires special 
> hardware.  

not strictly true.  you can use the network block device to do a
poorly-performing GFS setup.  You'll need kernel patches and utilities
from ( or you can try out the latest plogic kernel, though
i'm not done testing it just yet )

> PVFS (Parallel Virtual File System) seems like the most logical, 
> but it only appears to be a kernel module and not a static compile.

again, not /strictly/ true, though yes if you want a VFS-like
interface to pvfs you'll need the kernel module.  I think PVFS for
your needs is very *il*logical.  PVFS is great for making use of the
extra space on compute node disks in a *parallel* file system.  since
you have no disks,  you (obviously ) have no extra space, nor could
you really make use of the P in PVFS.  

> Any words of wisdom or experience?  All help will be appreciated.

summary: use scyld. (and i'm sure the scyld guys would love to hear
why you can't :> )


[ Rob Latham <rlatham at>         Developer, Admin, Alchemist ]
[ Paralogic Inc. -                                     ]
[                                                                     ]
[ EAE8 DE90 85BB 526F 3181                   1FCF 51C4 B6CB 08CC 0897 ]

More information about the Beowulf mailing list