With traditional virtual machines, a datastore is either a LUN (VMFS) or a volume (NFS). VMDK is sitting on top of other file system ( VMFS or NFS ). VVOLs can be considered as individual LUNs and VMDK are not sitting on any file system. Only file system is of guest operating system. There are many great articles online explaining in detail the architecture of VVOL. Putting a link to article below from VMware. Other that stood out to me is VMworld 2019 presentation on VVOL by VMware technical marketing team.
https://kb.vmware.com/s/article/2113013
Outlining below important reasons I think VVOL is useful:
- Monitoring Performance : With VVOL, we can get insight of each VMDK of a VM in terms of IOPS, latency and Bandwidth. As a result of this , it is easy to identify VM or VMDK with high storage usage and identify reasons for storage performance of a particular VM. More visibility to storage administrator on individual VM and individual VMDK. In traditional environment, storage administrator have no visibility into what is on the LUN and hence on visibility on performance of a VM. They had visibility into overall LUN performance but not individual VM. This results in better coordination between vSphere admin and storage admin in performance troubleshooting.
- Snapshot : With VVOL, snapshot will be done at array level. Obviously it will be faster. In traditional method, vSphere snapshot is very robust too but it does have a performance impact. With vSphere snapshot , it locks the original vmdk file and create delta file. Any changes further on goes to delta file and it keeps growing. Performance is impacted because on the datastore there is file with extension .sh which is metadeta file . ESX host has to update the metadata file everytime the delta disk grows. Read goes to original baseline vmdk file. With VVOL, it creates the meta data when we create the snapshot of VM. There is no separate read and write disk like in vsphere snapshot. Deleting a snapshot is also a breeze , it just deletes the meta data. Overall, snapshot is much more efficient and fast with VVOL.
- Management: VVOL bring simplified management, scalability, smarter provisioing. Both vSphere admin and storage admin don’t have to create and keep track of all LUNs or datastores. Storage admin just create a storage container which is technically allocating logical quota of storage. It helps storage admins in tracking the storage use and capacity planning.
- Storage Profile : Last but not least, is use of storage profile on individual VM or VMDK. It provides immense flexibility. Though we don’t use storage profile at this point . I wouldn’t comment much on this. Something we are aiming to use in the future.
Virtualization blog 