Replacing NFS.

[Jag]

On Mon, 09 Apr 2001, 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.  All nodes mount 
> their root filesystem and then a shared (read-only) /usr and shared 
> (read-write) /home exports.  So far I've tried to find documentation on 
> Network Block Device (NBD) and Extended NBD, but both seem to be in limbo as 
> far as development and current useability go.  Any recent info on either of 
> these?  Also, does anyone have experience with Coda and is it pretty much a 
> drop-in replacement for NFS?  I've also heard of AFS, but it seems to be a 
> bit limited.  Also, Global File System (GFS), but it requires special 

AFS is bad, but for the completely opposite reason.  It does to much,
thus requiring way more overhead than you want.  AFS is very good for
handling sites with > 5000 users with multiple fileservers and different
directories (volumes) being transparently stored on seperate servers, as
well as data replication and live backups.  It also does ACLs and a
bunch of other things that are extremely nice, but also add to extra
compute time to access your files, which is bad.

However, the main reason you don't want AFS is because AFS requires
tokens to access your files.  And the tokens expire, which means you
have to find a way to continually renew your tokens in order to run long
jobs.  This is a royal pain in the arse.  I speak from experience.


Jag
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 232 bytes
Desc: not available
URL: <http://www.beowulf.org/pipermail/beowulf/attachments/20010409/acf3bdc4/attachment.sig>