My Wish for dvFabric – a dvSwitch for Storage

This one has been sitting with me for a while. It would be nice to hear your comments on this.

The dvSwitch (Distributed Virtual Switch) – oh how it has changed our lives…

Do you all remember the days when you had to manually add the all the portgroups and network settings to a new host? We use scripts, we use mega-kickstart deployment scripts, gui applications.  Let us not forget that it is still the case for all those who do not have and Enterprise Plus vSphere license. You do have to pay some extra for the “premium features”, and if you ask me – it not really a premium feature but more of a necessity.

But since I have gotten used to a dvSwitch – those things are all in the past. Network configuration of a host – is really as simple as adding click click click (yes…. I know…. you could also automate it as well….), and I do not have to worry about configuration issues any more. This vmnic is used for vMotion, this one for FT, this for NFS/iSCSI – it is all so simple.

I think it is time that we asked for the next stage. It is time for a revolution!!

Maybe we should call it dvFabric?

Let me explain what I would like to see.

Today when you set up a host you have to setup your NFS/iSCSI/SAN connection more or less manually. Before you say – hey what about Host Profiles – yes you could do that with host profiles – but the way that I see this is – Host Pofiles comes as a band-aid to this issue – not solution to the problem. By applying a profile to a host – you are are running a process that runs a set of actions to set a configuration. You could so the same with a set of scripts (PowerCLI or Perl) but this is packaged very well in vCenter and easy to use.

So how do I envision this dvFabric? Essentially the same as a dvSwitch. A logical entity to which I attach my network cards (for iSCSI/NFS) or my HBA’s (for FcoE or FC). I define which uplink goes to which storage, what the multi-pathing policy is for this uplink, how many ports should be used, what is the failover policy for which NIC, which NFS volumes to mount, which LUNS to add – I gather you see what I am getting at. Here is a diagram trying to visualize this.


Once the dvFabric has been configured correctly – all you would need to do is add in a Host and all the storage will be configured automagically.

Today there are a number of additional benefits available with using a dvSwitch – just to name a few – port mirroring, Netflow, PVLANs, NIOC.

I am sure that there are additional benefits that VMware could come up with – have a distributed virtual Fabric switch. One that just came to me is to perform the zoning in the dvFabric itself – and not to each and every host. This way the storage administrator would only have to zone one entity and the rest will be up to the VI Admin. The same regarding NFS exports.

Will this ever come about? Who knows? It does seem to be to be the another natural stage in the abstraction of the layers which has been the trend of the last few years.

I guess one of the questions we should ask ourselves – and the storage vendors, Are we (and they) ready for it? Is this something that could / should / would happen in the future.

What are the risks? What are the benefits?

Perhaps some brilliant minds in some strange place in Palo Alto are, already thinking about this….?

I would appreciate your comments and thoughts.