We have a customer that has NFS v4 as a requirement. It's a really big account. They also require 32TB NFS volumes, with autofailover. I've got something that starts out at about 192 TB and they require the capability to grow to over 1 PetaByte at least (not in the same volume, necessarily). They also need Infiniband support as an option so they can attach it to their supercomputer cluster.
So, I have it all figured out, but there are a couple things we need but Open-E does not yet support (but should support or has hinted that they will soon support) :
NFS v4
autofailover with NFS
32TB replication logical volumes? (This may already be supported, but I think there might be a 16TB limit?)
Ability to upgrade one server, then "failover" to another server to upgrade the primary one, then failback so that users never have to experience down-time at all (This HAS to be supported by Open-E. It's ridiculous that it isn't already supported, since autofailover for iscsi has been out for a while now).
Optional:
IPoIB-CM (IPoIB is supported, but not the "connected mode" version. IPoIB is as slow as gigabit ethernet unless you use IPoIB-CM, so it kind of sucks right now)
Ability to instantly roll-back a volume to a previous snapshot (I know there's a patch out there for LVM that allows this, but it is a beta-quality patch)... otherwise, snapshots are not very useful.
How far is Open-E to doing these things? We would have to engineer the solution ourselves using linux or opensolaris boxes if this isn't possible. We'd MUCH rather be able to use Open-E for this.