As I continue to gather data to assist with our migration to 6.5, I read a few articles on virtual volumes and how they apparently make things so much better but as far as i can see, they appear to offer few real benefits??? I'm no expert but it seems like one of the main benefits is the ability to provision targeted storage for VM's with a particular need for say fast iops and high availability. This is obviously useful. Has anyone identified any other real benefits? Be good to get a conversation going
Biggest benefits I see are:
Granular storage policies per VM, so you could for example assign a policy to a VM to be on flash storage and another VM or VMDK to be on slower storage without having to worry about datastores at all.
From a management perspective, the Datastore concept is old. Mapping 1 LUN to 1 Datastore is just a waste of time. Datastores run out of space too, so getting rid of this LUN to Datastore mapping makes a lot of sense to me, no longer do we need to "Make sure there is enough space on the datastore to provision a new VM, or to storage vMotion"
The VVOL concept is not new, but the integration into production storage arrays is relatively new so I would be very careful when selecting a vendor.
Thanks for your input virtualg_uk! I agree, the idea of being free from a set number of luns at a set size will free things up a bit. Possibly lead to complacency though if the admin is not disciplined enough to routinely check storage levels on the arrays.
The main reasons i can see to go towards vvols would be to move away from array based snapshots, currently having to snap a whole volume is a pain, individual Vvol snaps make this easier, as you are targeting a specific VM rather than all on the datastore. This extends again if you are using these snaps for backups to another device or tape etc.
Another reason would be reclaiming space inside the volume, while i haven't looked at this yet, id assume it would be easier than SCSI unmap on an entire datastore.
Also as others have said above, not having to provision datastore volumes, but a number of solutions dont support vvols so you ll likely end up using both anyhow.
Here is the Interoperability for vmware products for 6.0.x, i can t find an updated document for 6.5
One or two papers I read suggested that VVOL's might present more benefits to larger operations with hundreds of sites. Possibly cloud based. Smaller estates like my own (120 VM's and two sites) might not see a huge benefit but as has been demonstrated above, that might not be the case...