[nycbug-talk] AFS (Andrew File System) Article published on Daemon News
ike at lesmuug.org
Thu May 11 14:45:46 EDT 2006
A humble question, but with regard to AFS/Coda 'where's the beef'?
On May 11, 2006, at 2:13 PM, Mikel King wrote:
> AFS: network filesystem beyond NFS weaknesses (Dreyfus, Emmanuel)
> This interview of Ty Sarna about AFS (Andrew File System) by Emmanuel
> Dreyfus, covers a lot of ground. This rare glimpse, explains what you
> need to do to properly deploy AFS in most environments. You will
> discover the advantages of deploying AFS in lieu of NFS or CIFS
> (Samba) and it is actually an enjoyable read.
I started going down the Coda/OpenAFS path a few months ago for a
project, and wound up retreating. The features were all extremely
compelling, as was the age (assumed software maturity) of the
filesystem, but alas- I found both Coda and AFS to be extremely
poorly implemented. For something as important as a filesystem, in
my particular application/context, both AFS/Coda were unacceptably
sloppy to work with- they felt like a houseplant occasionally tended
by graduate students at various universities...
With that, the concepts and features of AFS/Coda are all still very
intriguing, as they solve problems I've traditionally handled on
higher application levels combining network transacions and database
repication in various forms- i.e. I *want* it to work.
However, all that seems to really be great about AFS/Coda are glowing
conceptual advertisements from exited academics, much like the
interview posted here.
Has anyone on this list had good experiences with Coda/AFS?
Is anyone running it in some sort of production-grade enviornment?
Does anyone have anything except horror stories from the big
financial companies that have deployed it around town for years?
Of course I'm not in any way intending this email to be inflammatory,
but it's not just me, other trusted colleagues on various scales
report sub-par experiences with AFS/Coda.
More information about the talk