> The Open Source AFS project, Arla, may arrive in time to replace
> IBM / Transarc AFS, but so far Arla only has clients. In order
> to keep AFS alive there will need to be an Arla server.
Word on the street is that KTH has hired someone full time to work on Milko,
the Arla fileserver.
Milko is currently very much in an Alpha-stage, enough so that generally when
someone posts to the Arla-Drinkers list (the Arla discussion group)
compliaining about problems with it, the general response is, "Don't run
Milko, it's not ready yet for even beta-testing."
> Are there any other similar wide-area filesystem technologies?
> DFS, sure maybe, but I have never understood why our support
> guys consistently refuse to look that way.
Uh.
DFS was designed to be "AFS done right" and never really got there. Never
forget that a lot of the original AFS code was done by post-docs along with
experienced filesystems types. They were inventing the wheel as the drove the
car.
DFS was started largely by the folks who had done the filesystems internals for
AFS. However just as it started to get speed there was a sizeable exodus of
personnel.
DFS, which I've run little of in my time, is clunky. Even if you have years of
AFS experience the ramp-up time for learning it is very steep. The
administration is not always easy. The set up is a riot, and I mean as in
Watts.
Let's not forget that Transarc used to strongly encourage sites to go to DFS
from AFS. A lot of sites had major problems with migrating their cell to DFS.
By last December Transarc stopped supporting the AFS->DFS migration kit, and by
that time many of the biggest DFS sites gave up and went back to AFS or, worse,
NFS.
Occasionally some people make mumbles about another CMU project, CODA, being a
potential AFS replacement, but I've heard internal-CMU rumblings that it'll
never be a polished product, either.
e.
