On Apr 21, 2011 3:05 PM, "Gilles Mocellin" <gilles.mocellin@free.fr> wrote:
>
> Le Thursday 21 April 2011 13:29:23 shawn wilson, vous avez écrit :
> [...]
> > now, i am (obviously) considering building a sub-$1k san. what i'm
> > wondering if if there are any better or cheaper options? it would be
> > real nice if, instead of spending $900 + (big ass) power supply +
> > disks, i could buy tons of cheap consumer hardware and do internal
> > software raid and then some type of network raid and spend that money
> > on 4 or so servers that had ~4 bays and have (even smaler / cheaper)
> > redundant servers saying, get those few bits from server 1, then those
> > few bits from server 2, etc. since this would require fast switching,
> > i suppose such a think would run mainly on a network layer.
> >
> > i understand that these are some of the features of nfs4 / pnfs.
> > however, i was wondering if there was any way to do this currently
> > (with semi-stable software)?
> >
> > also, are there any *free* / open source low delta replication
> > solutions yet? lastly, is there any project for doing san stuff with
> > consumer hardware? (last i checked freenas looked more of a joke to
> > me)
>
> [...]
>
> I'd say you can look at some distributed filesystems.
> I was looking lately at glusterfs and moosefs.
>
> MooseFS is simpler to setup, and despite lake of High Availibillity of the
> metada server, it will soon be a good choice.
>
> In log term, I hope Ceph (already in upstream kernel) will be the real one.
>
> Please, report back if you make some tests !
>
Well, someone (literally 15 min ago) told me that I needed to pick up 3x p3 servers with 3x 72gig discs. So, tests will commence.
I don't have access to any commercial san hardware to compare. But I should be able to provide data on my end and hope someone can let me what those tests look on their enterprise stuff..
Then, if everything goes well I can drop some cash on a few of what I mentioned earlier.