opensource wide area san?

[Michael Butash]

I just started playing with openfiler recently, and I think for what you
want you'll need to end up using another protocol you can control,
either nfs or cifs (they're probably widest supported), and rsync behind
it if you want to use a wide-area optimization system on the network
between it.  If the bonded t1's are your ultimate solution, it'll be
necessary, as even cifs is wonky over a latent network (old tdm,
bandwidth-limited), and can benefit from tcp optimizations for such use.
Cisco and everyone else making them bank selling solutions that handle
"traffic optimization appliances" (read software on linux) for this, but
they do work for "enterprise" solutions and gives you a body to point a
finger at when they break.  

Look at this to see if it might be a good solution, supposedly it
supports cifs these days if windoze is your environment.  Rsync and
NFSv3/4 maybe as well being tcp-based for generic optimization and
compression. Might be worth running at least a local daemon on your OF
servers and let it tunnel the protocol itself via this as a proxy over
the wan.  I've considered the notion for a customer of mine, and maybe
testing it soon (I hope).

http://trafficsqueezer.sourceforge.net

Otherwise possibly consider a commercial appliance, as stated Cisco,
Riverbed, Juniper, and a gaggle of others sell products for this
purpose.  I can point you at some vendors or vars if you get to that
point.

-mb  


On Wed, 2009-07-29 at 14:06 -0700, Stephen wrote:
> We are useing an openfiler server here and it has been running great.
> 
> right now it is hosting several iSCSI connections to our servers
> however we need to replicate data on one of the iSCSI volumes between
> 2 sites.
> 
> internally using DRBD/heartbeat comes to mind as a new but no brainer
> solution. however these servers need to live in 2 different states
> (and eventually maybe 3-4 locations)
> 
> generally it would seem that rsync would be better however it is a
> file level replication and seems that it would not be able to
> replicate the iSCSI volume which is block level...
> 
> can DRBD/heartbeat sync across a t-1 or pair of bonded t-1's? or will
> it eat the pipe?
>