I've had an ephiphany.....

The reason noone uses EFS is simple: the initial setup complexity and
infrastructure requirements are beyond the abilities of just about
everyone who has tried to get this thing up and running.   In
addition, it's only the large complex IT shops that really would
benefit from the NAS infrastructure, or that have the skills to build
something like this, and in most cases, they have existing solutions
and therefore don't need EFS.

Part of the epiphany is that the real value of EFS is NOT the NAS
infrastructure, but the namespace, and the management software for
that namespace.   If we could make it easy to install EFS on a single
machine, with a local database, and no NAS, then the barrier to
adoption would be all but eliminated, and people could get this
running on a single machine very easily.  Would it scale?  No, of
course, not but if you only have 10 machines, why would you look at
software designed to support 1000 anyway?

That's the first part of the epiphany.  The other half was looking at
the code, and realizing how EASY it will be to have the efs CLI bypass
the TPC/IP connection to efsd, and simply make the damn API calls
directly.   Honestly, I think I can get a branch of efs-core working
like this in a day of hacking.

The idea is to use the upload/download mechanism to maintain the state
of the machine.  Yes, this becomes a scale of one solution, but this
also means that with a bit more work, EFS should be able to support an
environment which has BOTH centralized NAS filers, and machines which
use the download machine to maintain a local subset of the /efs/dist
data.

The tricky part of figuring out how to make /efs/dev work locally,
but... I've got some sick ideas.

I intend to have this working by YAPC::NA 2012, where I will be
talking about cpan2efs, and demonstrating efsdeploy and cpan2efs
working on my laptop, standalone.

I am convinced at this point that until we make EFS scale *down*,
we're never going to get any traction in the Open Source world...

Something to think about.
_______________________________________________
EFS-dev mailing list
[email protected]
http://mailman.openefs.org/mailman/listinfo/efs-dev

Reply via email to