[OpenAFS] AFS on NFS?

[Tino Schwarze]

On Fri, Jul 25, 2003 at 01:15:40AM -0600, Michael Loftis wrote:
> And while we're on that subject, performance (which is better/worse)?  And 
> how does one 'ask' for an inode based partition?

You 'ask' on compile time. On Linux, NAMEI is default. NAMEI is probably
a bit slower than inode-based (but I for myself don't like the idea of
performing behind-the-back operations on a filesystem - if you run the
original fsck by accident, you lose all your data).

> As I said in my other message I'm also gearing up to roll AFS out on a 
> fairly large web cluster, that is only going to grow, and I do not want to 
> revisit the "aww hell XYZ networked filesystem now sucks" problem that 
> we're having with NFS and bind mounts (this is what I'm left with from a 
> previous SysAdmin actually, not my personal choices...I would have been 
> much more likely to use OpenAFS even back when this thing was started).
> 
> Obviously with AFS it's not such a big deal that decisions like that be 
> made right up front (vos move mmmmmmmm tasty.......sorry daydreaming again 
> of a manageable system as compared to what I've got today....)

I think, AFS is quite manageable. It has some disadvantages but they are
mostly of the "you cannot solve this cleanly anyway" nature (like
RW-replication).

Bye, Tino.

-- 
             * LINUX - Where do you want to be tomorrow? *
                  http://www.tu-chemnitz.de/linux/tag/